A stem cell-like chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing

Abstract

Adult cancers may derive from stem or early progenitor cells. Epigenetic modulation of gene expression is essential for normal function of these early cells but is highly abnormal in cancers, which often show aberrant promoter CpG island hypermethylation and transcriptional silencing of tumor suppressor genes and pro-differentiation factors. We find that for such genes, both normal and malignant embryonic cells generally lack the hypermethylation of DNA found in adult cancers. In embryonic stem cells, these genes are held in a 'transcription-ready' state mediated by a 'bivalent' promoter chromatin pattern consisting of the repressive mark, histone H3 methylated at Lys27 (H3K27) by Polycomb group proteins, plus the active mark, methylated H3K4. However, embryonic carcinoma cells add two key repressive marks, dimethylated H3K9 and trimethylated H3K9, both associated with DNA hypermethylation in adult cancers. We hypothesize that cell chromatin patterns and transient silencing of these important regulatory genes in stem or progenitor cells may leave these genes vulnerable to aberrant DNA hypermethylation and heritable gene silencing during tumor initiation and progression.

Figure 1. Genes that are frequently DNA hypermethylated and silenced in adult cancers remain unmethylated in embryonal carcinoma (EC) and embryonic stem (ES) cells

A panel of 29 tumor suppressor, and candidate tumor suppressor genes were selected that are known to be frequently hypermethylated in various cancer cell lines and primary tumor samples (right panel) from review of the literature and from studies in our own labs (all utilized refs are given in Supplementary Table 1 - methylation was considered notable when greater than 5% of the human cell lines or patient samples surveyed were methylated). Methylation specific PCR was used to determine the promoter DNA methylation status of these genes in colon cancer HCT-116 cells, in WA01 human ES cells, and two EC cell lines: Tera-1 and Tera-2 cells. Genes were characterized as unmethylated (empty square), fully methylated (black square), or partially methylated (shaded grey squares).

Figure 2. EC cells retain a plasticity of expression that is lacking in adult cancer cells

a. Real-time qRT/PCR analysis of genes frequently hypermethylated in adult cancers following treatment EC cells with all-trans retinoic acid (ATRA, 2uM) for 0 (untreated), 1, 3, 6, 9, and 12 days. PCR reactions were performed in triplicate, and average threshold cycle for altered gene expression with ATRA treatment was normalized to GAPDH. Fold change (log scale) for each gene over untreated Tera-2 cells was calculated using the formula fold change = -log[c^t(treatment)-c^t(untreated)]. Representative results of two or more independent experiments are shown. Genes are divided into two groups by threshold cycle: low to medium expression (top panel - note high cycle threshold number for the PCR, average; 30.8)) and genes with high basal expression (note low threshold cycle number, average; 21.4). b. Immunohistochemistry of teratocarcinoma tumor grown in NOD/SKID mice. Immunostaining was performed by the Johns Hopkins Immunopathology Department according to established protocols on paraffin embedded section using antibodies to CD34 (40x):endothelial cells, chromogranin (40x); neuroendocrine, cytokeratin (10x): epithelial, alpha fetoprotein (AFP; 20x): yolk sac development, Glial fibrillary acidic protein (GFAP; 20x): glial cells, and Myogenin (40x): muscle. c. qRT/PCR was performed for RNA from 5×10⁶ Tera-2 cells grown as xenographs in nod-skid mice until tumors reached approximately 1.5 centimeters in diameter. Fold expression change was calculated as described above, and results from Tera2 cells treated with ATRA (2uM) for 12 days are shown for comparison.

Figure 2. EC cells retain a plasticity of expression that is lacking in adult cancer cells

a. Real-time qRT/PCR analysis of genes frequently hypermethylated in adult cancers following treatment EC cells with all-trans retinoic acid (ATRA, 2uM) for 0 (untreated), 1, 3, 6, 9, and 12 days. PCR reactions were performed in triplicate, and average threshold cycle for altered gene expression with ATRA treatment was normalized to GAPDH. Fold change (log scale) for each gene over untreated Tera-2 cells was calculated using the formula fold change = -log[c^t(treatment)-c^t(untreated)]. Representative results of two or more independent experiments are shown. Genes are divided into two groups by threshold cycle: low to medium expression (top panel - note high cycle threshold number for the PCR, average; 30.8)) and genes with high basal expression (note low threshold cycle number, average; 21.4). b. Immunohistochemistry of teratocarcinoma tumor grown in NOD/SKID mice. Immunostaining was performed by the Johns Hopkins Immunopathology Department according to established protocols on paraffin embedded section using antibodies to CD34 (40x):endothelial cells, chromogranin (40x); neuroendocrine, cytokeratin (10x): epithelial, alpha fetoprotein (AFP; 20x): yolk sac development, Glial fibrillary acidic protein (GFAP; 20x): glial cells, and Myogenin (40x): muscle. c. qRT/PCR was performed for RNA from 5×10⁶ Tera-2 cells grown as xenographs in nod-skid mice until tumors reached approximately 1.5 centimeters in diameter. Fold expression change was calculated as described above, and results from Tera2 cells treated with ATRA (2uM) for 12 days are shown for comparison.

Figure 3. Genes frequently DNA hypermethylated in cancer are marked by “bivalent” chromatin in ES cells, including the polycomb group (PcG) protein associated histone modification, H3K27me

a. Chromatin IP was performed using antibodies to H3K4me2, H3K9me27, H3K9me2, H3K9me3, on WA09 human ES cells. PCR was performed for regions within the CpG islands, and near the transcription start sites of p16, GATA4, GATA5, RASSF1, sFRP1, and sFRP5. Representative results of two independent experiments are shown. A zero-antibody negative control is included for comparison. b. Top panel: Polycomb repressive complex 1, 2/3 and 4 are shown. Lower panel: Genes frequently DNA hypermethylated in adult cancers are targeted by PcG target genes in human embryonic stem (ES) and embryonic fibroblast (EF) cells. We compiled the list of genes marked by by PcG proteins and estimated in two recent studies to represent between ~8% (PRC2; ES cells) and ~14% (PRC1/PRC2; EF cells) of the annotated genome. We identified, in these lists, our genes frequently DNA hypermethylated in adult cancers (top left) and find a remarkable 68% were associated with PcG in either ES or EF cells in the above studies. An additional 14% had related proteins under polycomb control. Similar results were seen using 23 genes newly identified by microarray studies (manuscript submitted) to be DNA hypermethylated in HCT-116 cells (top right).

Figure 4. Genes studied in EC cells share the bivalent chromatin pattern seen in ES cells, but add additional repressive marks characteristic of these same genes when they are induced to DNA de-methylate in adult cancers

a. Chromatin IP was performed using antibodies to H3K4me2, H3K9me27, H3K9me2, H3K9me3, on Tera-2 cells. PCR was performed for regions within the CpG islands, and near the transcription start sites of CDH1, sFRP1, sFRP2, p15, p16, GATA4, GATA5, sFRP5 and RASSF1. Representative results of two independent experiments are shown. A zero-antibody negative control is included for comparison. b. Quantitation of representative gels shown in Fig. 4a showing the ratio of H3K4me2 (active mark) and H3K27me3 (repressive mark) to input DNA in Tera-2 cells for CDH1, sFRP1, sFRP2, p15, p16, GATA4 and GATA5, and as an average for four of these genes (sFRP2, sFRP5, GATA4, GATA5) from previous studies of HCT-116 DKO and wild type cells (far right two columns). Error bars indicate standard errors. c. Ratio of H3K9me2 and H3K9me3 to input DNA in Tera-2 cells for CDH1, sFRP1, sFRP2, p15, p16, GATA4 and GATA5, and as an average for four of these genes (sFRP2, sFRP5, GATA4, GATA5) from previous studies in HCT116 wild type cells, where each of the genes is DNA hypermethylated and lacks any basal transcription, and HCT116 DKO cells where each gene has become fully DNA demthylated and is re-expressed (far right two columns). d. Western blot analysis for the repressive histone modifications, H3K9me2 and H3K9me3 in Tera2 cells and the human ES cell line WA01. Lamin B was used as a loading control.

Figure 4. Genes studied in EC cells share the bivalent chromatin pattern seen in ES cells, but add additional repressive marks characteristic of these same genes when they are induced to DNA de-methylate in adult cancers

a. Chromatin IP was performed using antibodies to H3K4me2, H3K9me27, H3K9me2, H3K9me3, on Tera-2 cells. PCR was performed for regions within the CpG islands, and near the transcription start sites of CDH1, sFRP1, sFRP2, p15, p16, GATA4, GATA5, sFRP5 and RASSF1. Representative results of two independent experiments are shown. A zero-antibody negative control is included for comparison. b. Quantitation of representative gels shown in Fig. 4a showing the ratio of H3K4me2 (active mark) and H3K27me3 (repressive mark) to input DNA in Tera-2 cells for CDH1, sFRP1, sFRP2, p15, p16, GATA4 and GATA5, and as an average for four of these genes (sFRP2, sFRP5, GATA4, GATA5) from previous studies of HCT-116 DKO and wild type cells (far right two columns). Error bars indicate standard errors. c. Ratio of H3K9me2 and H3K9me3 to input DNA in Tera-2 cells for CDH1, sFRP1, sFRP2, p15, p16, GATA4 and GATA5, and as an average for four of these genes (sFRP2, sFRP5, GATA4, GATA5) from previous studies in HCT116 wild type cells, where each of the genes is DNA hypermethylated and lacks any basal transcription, and HCT116 DKO cells where each gene has become fully DNA demthylated and is re-expressed (far right two columns). d. Western blot analysis for the repressive histone modifications, H3K9me2 and H3K9me3 in Tera2 cells and the human ES cell line WA01. Lamin B was used as a loading control.

Figure 5. Changes in histone modifications and localization of known polycomb group (PcG) proteins to the gene promoters in Tera-2 cells with ATRA induced differentiation

a. ChIP reactions were performed using antibodies to Suz12, EZH2 and SIRT1 in Tera-2 cells for genes that are expressed at a low to medium basal state in undifferentiated Tera-2 and up-regulated with differentiation (p16, GATA4, GATA5, p15), and genes that are expressed at a high basal degree in Tera-2 and down-regulated with differentiation (CDH1, sFRP1, and sFRP2). Representative results of two independent experiments are shown. A zero antibody negative control is included for comparison. b. RT/PCR was performed as described in Figure 2 for Bmi1, Suz12, EZH2, Sfmbt and SIRT1 during ATRA induced differentiation of Tera-2 cells. Fold expression change (log scale) is shown following 0, 1, 3, 6, 9, and 12 days of differentiation. c. Real time ChIP PCR shows a reduction in PcG protein localization to the promoters of CDH1, sFRP2, p15, and GATA-4 following ATRA (2 uM, 10 days) induced differentiation of Tera-2 cells. Fold change compared to undifferentiated cells is shown. A representative experiment from two or more PCR determinations is shown. d. Quantitative ChIP PCR showing fold change of H3K4me2/H3K27me3 ratio following ten days of ATRA treatment in genes that are down-regulated with ATRA treatment (CDH1 and sFRP2) and genes that are up-regulated with ATRA (p15 and GATA4). A representative experiment from two or more PCR determinations is shown.

Figure 5. Changes in histone modifications and localization of known polycomb group (PcG) proteins to the gene promoters in Tera-2 cells with ATRA induced differentiation

a. ChIP reactions were performed using antibodies to Suz12, EZH2 and SIRT1 in Tera-2 cells for genes that are expressed at a low to medium basal state in undifferentiated Tera-2 and up-regulated with differentiation (p16, GATA4, GATA5, p15), and genes that are expressed at a high basal degree in Tera-2 and down-regulated with differentiation (CDH1, sFRP1, and sFRP2). Representative results of two independent experiments are shown. A zero antibody negative control is included for comparison. b. RT/PCR was performed as described in Figure 2 for Bmi1, Suz12, EZH2, Sfmbt and SIRT1 during ATRA induced differentiation of Tera-2 cells. Fold expression change (log scale) is shown following 0, 1, 3, 6, 9, and 12 days of differentiation. c. Real time ChIP PCR shows a reduction in PcG protein localization to the promoters of CDH1, sFRP2, p15, and GATA-4 following ATRA (2 uM, 10 days) induced differentiation of Tera-2 cells. Fold change compared to undifferentiated cells is shown. A representative experiment from two or more PCR determinations is shown. d. Quantitative ChIP PCR showing fold change of H3K4me2/H3K27me3 ratio following ten days of ATRA treatment in genes that are down-regulated with ATRA treatment (CDH1 and sFRP2) and genes that are up-regulated with ATRA (p15 and GATA4). A representative experiment from two or more PCR determinations is shown.

Figure 6. Over-expression of Bmi1 can cause progressive promoter DNA methylation of the SFRP5 gene in EC cells

a. Tera2 cells were stably infected with a Bmi1 expressing lentivirus, and qRT/PCR for Bmi1 transcript is shown in Bmi1 infected pools of cells at passage 5, 10, and 15 post-infection. A representative experiment from two PCR determinations is shown. b. Wild type Tera2 (left panel) and Bmi1 over-expressing Tera2 cells (right panel) in culture (10×). Arrow (left panel) indicates small infrequent cluster of cells in Tera2 which demonstrate increased proliferation and loss of contact inhibition. c. Methylation analysis by MSP was performed for p16, GATA4, GATA5, and sFRP5 for Tera2 and Bmi1 over-expressing pools passage 5, 10, and 15 and five individual clones at passage 15 and 20. For clones, representative results for two of five clones are shown. For all MSP, representative results of two independent experiments are shown. M= methylated signal; U = unmethylated. Normal lymphocytes (NL) and in-vitro methylated DNA (IVD) are included as positive and negative controls for methylated DNA. d. Chromatin IP was performed at the sFRP5 promoter in Tera2 cells, pooled Bmi1 infected cells at passage10 post-infection, and a single clone of Bmi1 infected cells at passage 20. ChIP was performed using antibodies to H3K4me2, H3K9me27, H3K9me2, H3K9me3 as described in Fig. 4.

Figure 6. Over-expression of Bmi1 can cause progressive promoter DNA methylation of the SFRP5 gene in EC cells

a. Tera2 cells were stably infected with a Bmi1 expressing lentivirus, and qRT/PCR for Bmi1 transcript is shown in Bmi1 infected pools of cells at passage 5, 10, and 15 post-infection. A representative experiment from two PCR determinations is shown. b. Wild type Tera2 (left panel) and Bmi1 over-expressing Tera2 cells (right panel) in culture (10×). Arrow (left panel) indicates small infrequent cluster of cells in Tera2 which demonstrate increased proliferation and loss of contact inhibition. c. Methylation analysis by MSP was performed for p16, GATA4, GATA5, and sFRP5 for Tera2 and Bmi1 over-expressing pools passage 5, 10, and 15 and five individual clones at passage 15 and 20. For clones, representative results for two of five clones are shown. For all MSP, representative results of two independent experiments are shown. M= methylated signal; U = unmethylated. Normal lymphocytes (NL) and in-vitro methylated DNA (IVD) are included as positive and negative controls for methylated DNA. d. Chromatin IP was performed at the sFRP5 promoter in Tera2 cells, pooled Bmi1 infected cells at passage10 post-infection, and a single clone of Bmi1 infected cells at passage 20. ChIP was performed using antibodies to H3K4me2, H3K9me27, H3K9me2, H3K9me3 as described in Fig. 4.