Human mammary epithelial cells (HMEC) grown under standard cell culture conditions enter a growth phase referred to as selection, but a subpopulation is able to escape from arrest and continue to proliferate. These cells, called post-selection or variant HMECs, may be derived from progenitor cells found in normal mammary epithelium that subsequently acquire premalignant lesions, including p16INK4A promoter hypermethylation. Epigenetic silencing of tumor suppressor genes through DNA methylation and histone modification is an early event in tumorigenesis. A major challenge is to find genes or gene pathways that are commonly silenced to provide early epigenetic diagnostic and therapeutic cancer targets. To identify very early epigenetic events that occur in breast cancer, we used microarrays to screen for gene pathways that were suppressed in post-selection HMECs but reactivated after treatment with the demethylation agent 5-aza-2′-deoxycytidine. We found that several members of the transforming growth factor β (TGF-β) signaling pathway were consistently down-regulated in the post-selection HMEC populations, and this was associated with a marked decrease in Smad4 nuclear staining. Gene suppression was not associated with DNA methylation but with chromatin remodeling, involving a decrease in histone H3 lysine 27 trimethylation and an increase in histone H3 lysine 9 dimethylation and deacetylation. These results show for the first time that TGF-β2, its receptors TGF-βR1 and TGF-βR2, and activator thrombospondin-1 are concordantly suppressed early in breast carcinogenesis by histone modifications and indicate that the TGF-β signaling pathway is a novel target for gene activation by epigenetic therapy. [Cancer Res 2007;67(24):11517–27]

Major progress in controlling mortality and morbidity from cancer requires an understanding of the molecular mechanisms that are responsible for initiating the disease. Analysis of these early aberrant events is complex because by the time the tumor is detected, the cancer progenitor cells have already undergone multiple genetic and epigenetic changes that may be a consequence, rather than a cause, of malignancy. Genetic changes include DNA mutations and changes in DNA copy number due to gene deletions, rearrangements, and amplification, whereas epigenetic lesions involve alterations to both the chromatin structure and the DNA methylation pattern of the cancer genome (1). Changes in the DNA methylation pattern in cancer include global hypomethylation of CpG dinucleotides in DNA repeated regions (2) in conjunction with hypermethylation of CpG island promoter–associated genes (1, 3). CpG islands (4) are ∼200 bp to several kilobases in length and typically are unmethylated in a normal cell and span the proximal promoter and first exon of most tumor suppressor genes. Repressive chromatin modifications are also associated with tumor suppressor gene silencing and hypermethylation of CpG island promoters. Common chromatin modifications leading to repression include histone H3 lysine 9 (H3K9) methylation, H3K9 deacetylation (5), and histone H3 lysine 27 (H3K27) trimethylation, all of which contribute to the inaccessibility of the promoter elements to transcription factor binding (6). Although epigenetic changes are a common feature of human cancer cells compared with their normal counterparts, it is not clear at what stage in the transformation process these epigenetic changes occur and whether they occur concurrently or independently.

Human mammary epithelial cells (HMEC) isolated from disease-free breast tissue provide an excellent system to study early events in tumorigenesis. When cultured in serum-free medium, HMECs can exhibit two phases of growth (7, 8). The first growth phase, referred to as pre-selection, lasts for several passages, after which growth ceases (termed selection). However, some cells with enhanced growth capacity are able to escape this plateau and continue into a second growth phase where they proliferate for another 20 to 40 generations before entering a second growth plateau referred to as agonescence (9). Cells in the second phase of growth are termed post-selection (10) or variant HMECs (11, 12) and exhibit various preneoplastic characteristics including DNA methylation and silencing of the p16INK4A tumor suppressor gene (9, 10, 1215). We and others have previously shown that the silencing and DNA methylation of the p16INK4A locus occurs during selection (9, 10, 1215), and it has been suggested that epigenetic deregulation of p16INK4A occurs commonly in premalignant breast lesions. Indeed, rare foci of morphologically normal epithelial cells with hypermethylation of the tumor suppressor gene p16INK4A have been identified in vivo in disease-free breast tissue (12) and these foci have been hypothesized to be precursors to cancer (16).

The transforming growth factor β (TGF-β) family of growth factors comprises three TGF-β isoforms (TGF-β1, TGF-β2, and TGF-β3) and plays a pivotal role in the regulation of many cellular processes such as cell division, differentiation, motility, adhesion, and death (17, 18). Deregulation of TGF-β has been shown to be involved in the pathogenesis of cancer, including that of the breast, but the precise role it plays in cancer biology is complex. In normal cells of epithelial, endothelial, and hematopoietic cell lineages, TGF-β functions as a potent tumor suppressor (19). At some point along the neoplastic continuum, cells become unresponsive to TGF-β growth inhibition, and as malignant disease progresses, TGF-β can actually promote tumor progression by enhancing tumor cell motility and invasiveness (17, 19, 20). Several members of the TGF-β pathway, including TGF-β receptor 1 (TGF-βR1), TGF-β receptor 2 (TGF-βR2), and members of the downstream Smad signaling family, are often mutated and/or functionally inactivated in human cancers (21). In addition to genetic changes, epigenetics may also play a role in regulating individual members of the TGF-β pathway, including TGF-βR1 and TGF-βR2 (2226), the Smad molecules (27), and thrombospondin (THBS1), an activator of TGF-β (28); however, there are no reports of coordinate epigenetic suppression of these genes in breast cancer.

In this study, we used HMECs to identify genes or gene pathways that may succumb to epigenetic down-regulation in the early stages of breast cancer. Using microarrays, we identified gene suppression of transforming growth factor β2 (TGF-β2), its receptors TGF-βR1 and TGF-βR2, and its activator THBS1 in post-selection HMECs. Bisulfite methylation analysis and chromatin immunoprecipitation assays showed that the down-regulation was not under the control of DNA hypermethylation but was associated with methylation and deacetylation of H3K9 residues and demethylation of H3K27 residues in post-selection HMECs and in the breast cancer cell line MDAMB453. Decreased expression of TGF-β2 and its associated genes was also observed in breast tumor samples. This study shows (a) the epigenetic regulation of multiple members in a gene pathway; (b) that this can be controlled by histone methylation and deacetylation rather than DNA methylation; and (c) that epigenetic deregulation of the TGF-β gene pathway members is likely to be an early event in breast cancer formation.

Cells, cell culture, and tissue samples. Breast tissue removed at reduction mammoplasty was obtained with institutional ethics committee approval and informed donor consent. HMEC cultures were prepared from normal breast tissues as previously described (10) and further details about the HMEC samples are provided in Supplementary Methods. The breast cancer samples were obtained with institutional ethics committee approval and were previously described (29). Breast cancer cell lines (T47D, MDAMB453, MDAMB468, MDAMB231, MDAMB157, SKBR3, MCF-7, KPL1, DU4475, and MCF-10A) were obtained from American Type Culture Collection and cultured in RPMI 1640 with 10% fetal bovine serum.

Immunocytochemistry. The localization of Smad4 was detected by immunofluorescence staining with a primary Smad4 (B-8) mouse monoclonal antibody (Santa Cruz Biotechnology, Inc.) and a secondary goat anti-mouse IgG (H+L) Alexa Fluor 488 antibody (Molecular Probes, Invitrogen). For more details, see Supplementary Methods.

5-Aza-2′-deoxycytidine and trichostatin A treatment. MDAMB453 and T47D cells were treated with 5-aza-2′-deoxycytidine (5-aza-dC) and trichostatin A (TSA) essentially as previously described (30) with the modifications described in Supplementary Methods.

Microarray experiment and analysis. The details are described in Supplementary Methods.

Gene expression analysis. Total RNA was extracted from pre- and post-selection HMEC, breast cancer cell lines, and fresh frozen normal breast tissue using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. cDNA was reverse transcribed using SuperScript III RNase H- Reverse Transcriptase (Invitrogen Life Technologies) and random hexamers (Roche) according to the manufacturers' instructions. Primer sequences are described in Supplementary Table S1. Expression of TGF-β2, TGF-βR1, TGF-βR2, THBS1, and p16INK4A was quantitated with a fluorogenic real-time detection method using the ABI Prism 7900HT Sequence Detection System and normalized against endogenous control 18S rRNA mix (Applied Biosystems). For more details, see Supplementary Methods.

DNA isolation and DNA methylation analysis. Details of DNA isolation and the bisulfite protocol (31) are described in Supplementary Methods. Location of the bisulfite PCR amplicons, in relation to the CpG island and start of transcription, is summarized in Supplementary Fig. S1. The primer sequences are described in Supplementary Table S1.

Chromatin immunoprecipitation assays. Chromatin immunoprecipitation assays were carried out according to the manufacturer's instructions (Upstate Biotechnology) using Bre-40 and Bre-12 pre- and post-selection HMECs or MDAMB453 cells that were either untreated or treated with 3 μmol/L 5-aza-dC and 50 nmol/L TSA. The complexes were immunoprecipitated with antibodies specific for acetylated histone H3K9 (Upstate), dimethyl-histone H3K9 (Upstate), and trimethyl-histone H3K27 (Upstate), and DNA yield was measured by quantitative real-time PCR as previously described (32). Chromatin immunoprecipitation amplification primers are described in Supplementary Table S1, and the position of the chromatin immunoprecipitation PCR amplicons in relation to the CpG island and start of transcription is summarized in Supplementary Fig. S1. For each sample, the average CT value was obtained for the immunoprecipitated material and for the input chromatin. The difference in CT values (δCT) reflects the difference in the amount of material that was immunoprecipitated relative to the amount of input chromatin or the internal control genes GAPDH and 16-Cen (33) as described in ABI PRISM 7700 Sequence Detection System User Bulletin no. 2, P/N 4303859.

Down-regulation of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 gene expression in post-selection HMECs. To identify genes or gene pathways that are epigenetically down-regulated in the early stages of breast cancer, we grew normal HMECs in culture to select for post-selection HMECs. We used microarrays to compare the gene expression profiles between pre- and post-selection HMECs from two HMEC strains (designated Bre-60 and Bre-80) and from the post-selection cells following treatment with 5-aza-dC treatment. Three hundred eighty-two of 10,242 informative genes on the Compugen array were found to be down-regulated >1.5 fold in both Bre-60 and Bre-80 post-selection HMECs compared with the equivalent pre-selection HMECs (Supplementary Fig. S2). Forty-three of these genes, including several TGF-β associated transcripts, were up-regulated >1.5-fold in both Bre-60 and Bre-80 post-selection HMECs after 5-aza-dC treatment, indicating that the gene expression may be controlled epigenetically. TGF-β2 was of direct interest because disruption of the TGF-β signaling pathway due to mutation is implicated in the pathogenesis of breast cancer (19). We were therefore interested to determine if epigenetic deregulation of the TGF-β signaling pathway also occurred in oncogenesis. Interestingly, the TGF-β2 receptors, TGF-βR1 and TGF-βR2, and THBS1, an activator of TGF-β2, were also generally down-regulated in both Bre-60 and Bre-80 post-selection cells, unlike TGF-β1 and TGF-β3, which were not suppressed (Supplementary Fig. S2).

To validate if TGF-β2 mRNA expression was down-regulated in different post-selection HMEC strains to the ones used in the array studies, we carried out quantitative reverse transcription-PCR (RT-PCR) and found that TGF-β2 mRNA expression was also suppressed (8.1- and 5.4-fold, respectively) in two independent post-selection HMEC strains designated Bre-40 and Bre-12 (Fig. 1A). This suggests that TGF-β2 gene suppression is a common and early event in post-selection HMECs that escape senescence and continue to proliferate. We also observed gene suppression of other CpG island–associated TGF-β pathway genes, TGF-β receptors TGF-βR1 and TGF-βR2, and THBS1 in post-selection Bre-40 HMECs by quantitative RT-PCR and modest reduction in mRNA expression of TGF-βR1 and THBS1 in post-selection Bre-12 cells (Fig. 1A), suggesting that the TGF-β associated genes may be coordinately regulated. The p16INK4A CpG island promoter, which we and others have reported to be methylated in post-selection HMECs (9, 10, 1215, 34), was also found to be completely inactivated in both post-selection Bre-40 and Bre-12 HMECs (Fig. 1A).

Figure 1.

Suppression of TGF-β gene expression and TGF-β signaling in pre- and post-selection HMECs. A, mRNA levels of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 in Bre-40 and Bre-12 pre- and post-selection HMECs were determined by quantitative RT-PCR. After normalizing expression to 18S rRNA, the fold change in expression levels was made relative to pre-selection HMECs. p16INK4A is shown as a positive control for gene suppression in post-selection HMECs. B, Smad4 immunostaining in Bre-38 pre-selection cells showing predominant Smad4 nuclear localization. C, Smad4 immunostaining of Bre-38 post-selection cells showing predominantly cytoplasmic localization. Blue signal corresponds to 4′,6-diamidino-2-phenylindole (DAPI) and the green signal represents Smad4 expression. Arrows, examples of nuclear high staining (NH), nuclear low staining (NL), and cytoplasmic staining (C). D, quantification of Smad4 immunostaining pattern shown in B and C. The percentages of cells in which Smad4 staining was high or low in the nucleus or evenly distributed throughout the cytoplasm were determined by randomly counting 260 and 226 individual pre- and post-selection Bre-38 HMECs in >40 independent fields, respectively. ***, P < 0.0001, Smad4 localization found to be significantly different between pre- and post-selection HMECs.

Figure 1.

Suppression of TGF-β gene expression and TGF-β signaling in pre- and post-selection HMECs. A, mRNA levels of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 in Bre-40 and Bre-12 pre- and post-selection HMECs were determined by quantitative RT-PCR. After normalizing expression to 18S rRNA, the fold change in expression levels was made relative to pre-selection HMECs. p16INK4A is shown as a positive control for gene suppression in post-selection HMECs. B, Smad4 immunostaining in Bre-38 pre-selection cells showing predominant Smad4 nuclear localization. C, Smad4 immunostaining of Bre-38 post-selection cells showing predominantly cytoplasmic localization. Blue signal corresponds to 4′,6-diamidino-2-phenylindole (DAPI) and the green signal represents Smad4 expression. Arrows, examples of nuclear high staining (NH), nuclear low staining (NL), and cytoplasmic staining (C). D, quantification of Smad4 immunostaining pattern shown in B and C. The percentages of cells in which Smad4 staining was high or low in the nucleus or evenly distributed throughout the cytoplasm were determined by randomly counting 260 and 226 individual pre- and post-selection Bre-38 HMECs in >40 independent fields, respectively. ***, P < 0.0001, Smad4 localization found to be significantly different between pre- and post-selection HMECs.

Close modal

To investigate if gene suppression of TGF-β2 and associated genes resulted in impaired TGF-β signaling, we examined the nuclear localization of Smad4 fluorescence in the pre- and post-selection HMECs because Smad4 is only translocated to the nucleus if the TGF-β receptor is activated. As previously reported, the cell morphology is different and heterogeneous in pre- and post-selection HMECs (10, 35). We found a marked and significant (P < 0.0001, χ2 test) difference in the number of cells with intense Smad4 nuclear staining in the post-selection HMECs in comparison with the pre-selection HMECs (Fig. 1B–D). In the post-selection HMECs, Smad4 fluorescence was evenly distributed and predominantly cytoplasmic, supporting the gene expression results indicating that the TGF-β signaling pathway is indeed disrupted in post-selection cells.

Concordant epigenetic suppression of TGF-β2 and associated TGF-β genes in post-selection HMECs. Because TGF-β2 was identified on the microarray as a gene that was up-regulated after treatment with the demethylation agent 5-aza-dC, we asked whether suppression of TGF-β2 and associated TGF-β genes in post-selection HMECs was due to DNA hypermethylation. We carried out bisulfite methylation assays on DNA from four independent pre- and post-selection HMEC strains (Bre-40, Bre-60, Bre-70, and Bre-80) and DNA from two normal breast tissue samples (Bre-12 and Bre-13; Fig. 2A and Supplementary Fig. S3). After bisulfite treatment, PCR amplification within the CpG island promoters was done in triplicate and methylation of the pooled PCR products was determined by PCR heat dissociation curve analysis (Supplementary Fig. S3A), direct bisulfite sequencing (Supplementary Fig. S3B), and/or bisulfite clonal sequencing (Supplementary Fig. S3C). Figure 2A summarizes the methylation studies and shows that TGF-β2, TGF-βR1, TGF-βR2, and THBS1 were unmethylated in all the pre- and post-selection HMEC strains. In contrast, p16INK4A was unmethylated in the normal uncultured breast tissue samples (Bre-12 and Bre-13) and in the pre-selection cells (Bre-40, Bre-60, Bre-70, and Bre-80) but was extensively methylated in all four post-selection HMEC strains studied, as determined previously (10). We concluded that unlike p16INK4A, DNA methylation of the CpG island promoter is not responsible for the suppression of TGF-β2, TGF-βR1, TGF-βR2, or THBS1 gene expression in post-selection HMEC cultures.

Figure 2.

DNA methylation and chromatin immunoprecipitation analysis in HMECs. A, summary of DNA methylation status of TGF-β2, TGF-βR1, TGF-βR2, THBS1, and p16INK4A in Bre-40, Bre-60, Bre-70, and Bre-80 pre- and post-selection (shaded) HMECs and normal Bre-12 and Bre-13 breast tissue samples as determined by bisulfite methylation analysis (see Supplementary Fig. S3). −, unmethylated; +, methylated; NA, not amplified. Chromatin from HMECs was immunoprecipitated with dimethylated H3K9 (B), diacetylated H3K9 (C), and trimethylated H3K27 (D) antibodies. The amount of immmunoprecipitated DNA (relative binding) was quantified by real-time PCR and was calculated as a ratio of immunoprecipitated DNA to the total amount of input DNA or as a ratio relative to the amount of immunoprecipitated 16-Cen (B:Bre-40) or GAPDH (C:Bre-40). B and C, data are relative to pre-selection HMECs.

Figure 2.

DNA methylation and chromatin immunoprecipitation analysis in HMECs. A, summary of DNA methylation status of TGF-β2, TGF-βR1, TGF-βR2, THBS1, and p16INK4A in Bre-40, Bre-60, Bre-70, and Bre-80 pre- and post-selection (shaded) HMECs and normal Bre-12 and Bre-13 breast tissue samples as determined by bisulfite methylation analysis (see Supplementary Fig. S3). −, unmethylated; +, methylated; NA, not amplified. Chromatin from HMECs was immunoprecipitated with dimethylated H3K9 (B), diacetylated H3K9 (C), and trimethylated H3K27 (D) antibodies. The amount of immmunoprecipitated DNA (relative binding) was quantified by real-time PCR and was calculated as a ratio of immunoprecipitated DNA to the total amount of input DNA or as a ratio relative to the amount of immunoprecipitated 16-Cen (B:Bre-40) or GAPDH (C:Bre-40). B and C, data are relative to pre-selection HMECs.

Close modal

To determine if the down-regulation of TGF-β2 and associated genes was due to chromatin remodeling and repression, we carried out chromatin immunoprecipitation assays on pre- and post-selection Bre-40 and Bre-12 HMECs with antibodies to dimethylated and diacetylated H3K9. First, we found that the TGF-β2 CpG island promoter selectively bound methylated H3K9 antibody in the post-selection versus pre-selection Bre-40 and Bre-12 HMECs (Fig. 2B). TGF-β2 receptors (TGF-βR1 and TGF-βR2) also selectively bound more methylated H3K9 antibody in the post-selection versus pre-selection Bre-40 HMECs, as did THBS1 and p16INK4A (Fig. 2B), indicating that suppression of TGF-β2 and associated genes in post-selection cells correlates with repressed chromatin, regardless of DNA hypermethylation. Indeed, the level of promoter-associated H3K9 methylation for Bre-12 associated with the p16INK4A, TGF-β2, TGF-βR1, and THBS1 CpG island promoters increased with successive passages (passages 3–16; Fig. 2B), indicating an enrichment of chromatin modification with each cell division in the post-selection HMECs. Interestingly in Bre-12 HMECs, there was no change in the TGF-βR2 CpG island H3K9 methylation, and this also correlated with a lack of TGF-βR2 suppression (Fig. 1A). Second, we found that the TGF-β2 CpG island promoter selectively bound diacetylated H3K9 antibody in the pre-selection Bre-40 HMECs, indicating deacetylation of the H3K9 associated histones in the post-selection cells (Fig. 2C). Similarly, H3K9 deacetylation of p16INK4A, receptors TGF-βR1 and TGF-βR2, and THBS1 was also observed during post-selection growth, correlating again with gene suppression. These results show that TGF-β2 and associated pathway members TGF-βR1, TGF-βR2, and THBS1 are concordantly down-regulated in post-selection HMECs by repressive chromatin remodeling associated with H3K9 deacetylation and methylation rather than DNA hypermethylation.

Polycomb-mediated epigenetic suppression of TGF-β2 and associated TGF-β genes in post-selection HMECs. It has been proposed that CpG islands that undergo de novo methylation in cancer cells may be marked by polycomb-mediated H3K27 trimethylation in early development, which is then maintained in differentiated cell types by the presence of an EZH2-containing polycomb complex (36, 37). Therefore, to investigate if the p16INK4A gene, which is hypermethylated in post-selection cells, is associated with polycomb-mediated repression, we carried out chromatin immunoprecipitation on pre- and post-selection Bre-40 HMECs with antibodies to trimethylated H3K27. We found that the p16INK4A CpG island unmethylated promoter was enriched for trimethylated H3K27 in Bre-40 pre-selection cells; however, this enrichment was substantially reduced in the post-selection cells when p16INK4A was methylated (Fig. 2D). Interestingly, the unmethylated CpG island promoters associated with TGF-β2, TGF-β2R1, and TGF-β2R2 also showed a reduction in H3K27 trimethylation in the Bre-40 post-selection cells despite the fact these genes did not become DNA methylated following selection. Notably, the level of H3K27 trimethylation in the pre-selection cells was greater in p16INK4A than the level found in the TGF-β associated genes, suggesting that the density of modification may be an important factor in promoting DNA methylation versus histone methylation.

Epigenetic deregulation of TGF-β2, TGF-β receptors, and THBS1 in breast cancer cell lines. Because post-selection HMECs may mimic the early epigenetic lesions that occur in breast cancer initiation, we next asked if the coordinate epigenetic gene suppression of TGF-β2, TGF-βR1 and TGF-βR2 receptors, and THBS1 that was observed in post-selection HMECs was also observed in breast cancer cell lines. A summary of the expression data for eight breast cancer cell lines (T47D, MDAMB453, MDAMB468, MDAMB231, MDAMB157, SKBR3, MCF-7, and MCF-10A) is shown in Fig. 3A and Supplementary Fig. S4. By comparison with pre-selection HMECs, reduced gene expression was observed for TGF-β2 in 6 of 8 (75%) breast cancer cell lines, TGF-βR1 in 3 of 8 (38%), TGF-βR2 in 7 of 8 (88%), and THBS1 in 8 of 8 (100%; Fig. 3B). The level of concordant gene suppression varied between the breast cancer lines. In MDAMB453 and MCF7, all four genes (TGF-β2, TGF-βR1, TGF-βR2, and THBS1) were suppressed relative to expression in pre-selection Bre-40 cells; expression of three genes (TGF-β2, TGF-βR1 or TGF-βR2, and THBS1) was reduced in T47D, MDAMB157, SKBR3, and MCF10-A; and expression of two genes (TGF-βR2 and THBS1) was down-regulated in MDAMB468 and MDAMB231. There was no apparent correlation with estrogen receptor status.

Figure 3.

Gene expression in breast cancer cell lines. A, mRNA levels of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 in eight breast cancer cell lines and Bre-40 and Bre-12 HMECs were determined by quantitative RT-PCR. After normalizing expression to 18S rRNA, the fold change in expression levels was made relative to Bre-40 pre-selection HMECs. B, summary table of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 mRNA expression status in eight breast cancer cell lines. Dotted boxes, reduced gene expression relative to expression in Bre-40 pre-selection HMECs. Percentage of cell lines with reduced gene expression is indicated. Estrogen receptor (ER) status is also noted (54).

Figure 3.

Gene expression in breast cancer cell lines. A, mRNA levels of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 in eight breast cancer cell lines and Bre-40 and Bre-12 HMECs were determined by quantitative RT-PCR. After normalizing expression to 18S rRNA, the fold change in expression levels was made relative to Bre-40 pre-selection HMECs. B, summary table of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 mRNA expression status in eight breast cancer cell lines. Dotted boxes, reduced gene expression relative to expression in Bre-40 pre-selection HMECs. Percentage of cell lines with reduced gene expression is indicated. Estrogen receptor (ER) status is also noted (54).

Close modal

To determine if the reduced expression of TGF-β2 and associated genes in the breast cancer cell lines was associated with epigenetic suppression, we treated MDAMB453 and T47D cells with 5-aza-dC, TSA, or a combination of both. In MDAMB453 cells, expression of TGF-β2 mRNA increased 30-fold following treatment with 5-aza-dC, but showed limited activation after TSA alone (Fig. 4A). In T47D cells, expression of TGF-β2 mRNA increased ∼3-fold following treatment with 5-aza-dC and showed ∼6-fold activation after TSA alone (Fig. 4B). In both MDAMB453 and T47D cells, TGF-βR1, TGF-βR2, THBS1, and p16INK4A all showed minimal gene activation after 5-aza-dC treatment (1.2- to 3-fold) or TSA treatment (Fig. 4A and B). However, when the two breast cancer cell lines were treated with 5-aza-dC in conjunction with TSA, all genes commonly exhibited an enhanced induction ranging from 1.5- to 167-fold, most notably in MDAMB453 cells.

Figure 4.

Gene reactivation and DNA methylation studies in breast cancer cell lines. mRNA levels of TGF-β2, TGF-βR1, TGF-βR2, THBS1, and p16INK4A mRNA in MDAMB453 (A) and T47D (B) were determined by quantitative RT-PCR in untreated cells and cells treated with 5-aza-dC (Aza) alone, TSA alone, or 5-aza-dC/TSA in combination. After normalizing expression to 18S rRNA, the fold change in expression levels was made relative to untreated cells. C, summary of DNA methylation status of TGF-β2, TGF-βR1, TGF-βR2, THBS1, and p16INK4A in 10 breast cancer cell lines as determined by bisulfite methylation analysis (see Supplementary Fig. S4). −, unmethylated; +, methylated; NA, not amplified; Deleted, p16INK4A is deleted.

Figure 4.

Gene reactivation and DNA methylation studies in breast cancer cell lines. mRNA levels of TGF-β2, TGF-βR1, TGF-βR2, THBS1, and p16INK4A mRNA in MDAMB453 (A) and T47D (B) were determined by quantitative RT-PCR in untreated cells and cells treated with 5-aza-dC (Aza) alone, TSA alone, or 5-aza-dC/TSA in combination. After normalizing expression to 18S rRNA, the fold change in expression levels was made relative to untreated cells. C, summary of DNA methylation status of TGF-β2, TGF-βR1, TGF-βR2, THBS1, and p16INK4A in 10 breast cancer cell lines as determined by bisulfite methylation analysis (see Supplementary Fig. S4). −, unmethylated; +, methylated; NA, not amplified; Deleted, p16INK4A is deleted.

Close modal

To ascertain whether the combined 5-aza-dC and TSA treatment was inducing expression via DNA demethylation, TGF-β2, TGF-βR1, TGF-βR2, THBS1, and p16INK4A were assayed for DNA methylation in MDAMB453 cells and nine other breast cancer cell lines (T47D, MDAMB468, SKBR3, KPL1, MDAMB231, DU4475, MCF-7, MDAMB157, and MCF-10A). A summary of the methylation status in the 10 breast cancer cell lines is shown in Fig. 4C. Interestingly, the CpG islands spanning TGF-β2, TGF-βR1, THBS1, and p16INK4A promoters were all unmethylated in MDAMB453 cells, and TGF-βR2 showed a low level of DNA methylation by direct PCR bisulfite sequencing (Fig. 4B). TGF-β2, TGF-βR1, TGF-βR2, and THBS1 were either unmethylated or only partially methylated in the other nine breast cancer cell lines examined (Fig. 4C and Supplementary Fig. S5). Therefore, DNA methylation is not responsible for the suppression of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 in either the breast cancer cells or post-selection HMECs.

To determine if the epigenetic regulation of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 in MDAMB453 cells is due to chromatin remodeling rather than DNA hypermethylation, chromatin immunoprecipitation assays were done with dimethylated H3K9, diacetylated H3K9, and trimethylated H3K27 antibodies on untreated MDAMB453 cells versus cells treated with 5-aza-dC in conjunction with TSA (Fig. 5). Treatment with 5-aza-dC/TSA resulted in a loss of dimethylated H3K9 (Fig. 5A) and a gain of diacetylated H3K9 (Fig. 5B) in all four genes (TGF-β2, TGF-βR1, TGF-βR2, and THBS1). Interestingly a loss of H3K9 methylation and a gain of H3K9 acetylation were also observed for p16INK4A, which is unmethylated in MDAMB453 cells. We also examined the chromatin state of connexin 26 (38) because it is reported to be methylated in MDAMB453 cells. We found that after treatment with 5-aza-dC/TSA, there was a substantial loss of H3K9 methylation and a moderate increase in H3K9 acetylation. Interestingly, the four TGF-β pathway genes as well as p16INK4A bound more trimethylated H3K27 antibody in MDAMB453 cells treated with 5-aza-dC/TSA, despite these genes being unmethylated in MDAMB453 cells. H3K27 trimethylation was also elevated in the methylated connexin 26 after treatment with 5-aza-dC/TSA. These results suggest that EZH2 polycomb modification could mediate the repression of these genes in the cancer cell, and the polycomb mark is reduced in the cancer cell when the histones are repressed with H3K9 methylation and deacetylation marks regardless of the DNA methylation status. Moreover, the results indicate that there is a similar epigenetic-based repression of the TGF-β pathway genes in the post-selection HMECs and the breast cancer cell line MDAMB453, which is associated with histone H3K9 methylation and deacetylation and loss of H3K27 trimethylation.

Figure 5.

Chromatin immunoprecipitation studies in MDAMB453 cells after 5-aza-dC/TSA treatment. Chromatin from MDAMB453 cells treated with 5-aza-dC/TSA was immunoprecipitated with dimethylated H3K9 (A), diacetylated H3K9 (B), and trimethylated H3K27 (C) antibodies. The amount of immmunoprecipitated DNA (relative binding) was quantified by real-time PCR and was calculated as a ratio of immunoprecipitated DNA to the total amount of input. A and B, data are relative to untreated cells. Cx-26, connexin 26.

Figure 5.

Chromatin immunoprecipitation studies in MDAMB453 cells after 5-aza-dC/TSA treatment. Chromatin from MDAMB453 cells treated with 5-aza-dC/TSA was immunoprecipitated with dimethylated H3K9 (A), diacetylated H3K9 (B), and trimethylated H3K27 (C) antibodies. The amount of immmunoprecipitated DNA (relative binding) was quantified by real-time PCR and was calculated as a ratio of immunoprecipitated DNA to the total amount of input. A and B, data are relative to untreated cells. Cx-26, connexin 26.

Close modal

Concordant gene suppression of TGF-β2, TGF-βR1 and TGF-βR2 receptors, and THBS1 in breast cancer. To determine if coordinate suppression of the TGF-β pathway genes is common in primary breast cancer cells, we carried out quantitative RT-PCR to measure the expression of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 in 18 breast tumor samples and compared this with the expression in four normal breast tissue samples (Bre-43, Bre-56, Bre-59, and Bre-67; Fig. 6 and Supplementary Fig. S6). We found that, consistent with the breast cancer cell lines, TGF-β2, TGF-βR2, and THBS1 expression is commonly reduced [P < 0.03, P < 0.002, and P < 0.003, respectively (two-tailed Mann-Whitney U test)] in the cancer samples relative to the normal tissue, whereas TGF-βR1 expression was more variable (Fig. 6A). Levels of expression in the pre-selection HMECs were also consistent with that of the primary normal breast tissue (Fig. 6B). Coordinate suppression of all four genes was observed in 10 of 18 (55%) breast tumor samples, and 18 of 18 (100%) showed a reduction of expression of TGF-β2, TGF-βR2, and THBS1 relative to the average normal expression level. We analyzed the CpG island promoters associated with TGF-β2 in all of the cancer samples and found DNA methylation only in donor 37 (Supplementary Fig. S5C). These results support the conclusions from the post-selection HMEC studies that suppression of the TGF-β signaling pathway is common in cancer, is an early event, and is associated with chromatin repression rather than DNA methylation.

Figure 6.

Concordant gene suppression in breast cancer. A, scatter plots depicting TGF-β2, TGF-βR1, TGF-βR2, and THBS1 mRNA expression in 18 breast tumors (T; tumors 70, 39, 46, 30, 15, 28, 9, 37, 22, 32, 45, 10, 23, 25, 43, 24, 6, and 47) and four normal breast tissue samples (N; Bre-43, Bre-56, Bre-59, and Bre-67). B, mRNA expression of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 mRNA expression in the 18 breast tumors and 4 normal breast tissue samples and Bre-40 and Bre-12 pre-selection HMECs was determined by quantitative RT-PCR. Expression levels were normalized to 18S rRNA and made relative to Bre-40 pre-selection HMECs.

Figure 6.

Concordant gene suppression in breast cancer. A, scatter plots depicting TGF-β2, TGF-βR1, TGF-βR2, and THBS1 mRNA expression in 18 breast tumors (T; tumors 70, 39, 46, 30, 15, 28, 9, 37, 22, 32, 45, 10, 23, 25, 43, 24, 6, and 47) and four normal breast tissue samples (N; Bre-43, Bre-56, Bre-59, and Bre-67). B, mRNA expression of TGF-β2, TGF-βR1, TGF-βR2, and THBS1 mRNA expression in the 18 breast tumors and 4 normal breast tissue samples and Bre-40 and Bre-12 pre-selection HMECs was determined by quantitative RT-PCR. Expression levels were normalized to 18S rRNA and made relative to Bre-40 pre-selection HMECs.

Close modal

HMECs grown in culture provide a unique in vitro system to dissect the very early molecular pathways leading to carcinogenesis. Post-selection HMECs represent a subpopulation of cells that proliferate rapidly and have p16INK4A silencing and promoter methylation and cyclooxygenase-2 overexpression. In the later stages of post-selection growth, there is an accumulation of chromosomal abnormalities and expression phenotypes that are conducive to malignant progression (9, 10, 1215, 34, 39). Recently, it has also been shown that in post-selection HMECs, loss of p16INK4A activity is associated with polycomb-mediated silencing and DNA methylation of the HOXA9 gene, which raises the possibility that these cells may also share characteristics of stem or progenitor cells (40).

To define early epigenetic lesions in breast cancer, we used primary HMECs grown in culture to generate post-selection HMECs as an in vitro model to study mammalian carcinogenesis. We identified concordant epigenetic suppression of multiple members of the TGF-β signaling pathway in the post-selection HMECs. This involved lowered expression of the receptors TGF-βR1 and TGF-βR2, THBS1 (an activator of TGF-β2), and the TGF-β2 gene itself; expression levels of the other TGF-β genes (isoforms), TGF-β1 and TGF-β3, were not suppressed. Functional TGF-β signaling results in nuclear localization of Smad4 in pre-selection HMECs, whereas altered expression of TGF-β2 and TGF-β receptors in post-selection HMECs was associated with a significant reduction in Smad4 nuclear localization, indicating deregulation of a functional TGF-β pathway. Gene suppression of TGF-β2, TGF-β receptors, and THBS1 was also commonly found in breast cancer cell lines and breast cancer tissue samples. Surprisingly, gene suppression was associated with a decrease in polycomb-directed H3K27 histone methylation and an increase in histone H3K9 methylation and deacetylation, but suppression was not associated with DNA methylation. These results indicate that epigenetic suppression of TGF-β signaling is an early lesion that promotes cell proliferation in the process of breast oncogenesis.

Contemporary ideas of carcinogenesis envisage a series of stochastic genetic and epigenetic changes that confer a selective growth advantage over healthy cells. These changes collectively lead to the disruption of coordinated networks of intercellular communication and cause a fundamental change in cellular behavior that affects processes such as proliferation, differentiation, and apoptosis, with progressive dysregulation and acquisition of a malignant phenotype. The TGF-β signaling pathway is involved in many different biological processes during embryonic development and also plays an important role in tissue homeostasis (18). TGF-β is an important negative growth regulator that has antiproliferative effects on a range of epithelial cells, and disruption of TGF-β signaling is strongly implicated in many cancers including breast cancer (20, 41). Interestingly, it was previously reported (35, 42) that transformed HMECs commonly showed resistance to TGF-β growth inhibition compared with normal HMECs isolated from reduction mammoplasty, suggesting that suppression of the TGF-β pathway contributes to proliferation of these cells. Genetic lesions of individual members of the TGF-β signaling pathway are also frequently reported in human cancer. For example, TGF-βR1 and TGF-βR2 are inactivated by mutations in numerous cancers including breast and ovarian cancer (43, 44). In addition to genetic lesions in the TGF-β signaling pathway, epigenetic modifications have also been reported in some cancers. For example, THBS1, which is an activator of TGF-β (20), is down-regulated in many tumors (28) and methylated in glioblastoma cell lines and in 33% of examined glioblastoma multiforme tumors. The TGF-β receptors TGF-βR1 and TGF-βR2, which are essential for TGF-β signaling (20), are epigenetically regulated in many cancers; TGF-βR1 is hypermethylated in gastric cancer (22), and treatment of breast cancer cells with DNA methylation (23) or histone deacetylase (25) inhibitors induces TGF-βR1 and TGF-βR2 expression, indicating a role of chromatin remodeling in epigenetic repression. TGF-β has been observed to have dual and opposing roles in breast cancer: in healthy tissues it acts as a tumor suppressor gene, but in the advanced stages of cancer it can switch to be a tumor promoter (20). The answer to this apparent paradox may lie in the early stages of epigenetic silencing of the TGF-β signaling pathway due to chromatin remodeling and not DNA methylation, therefore permitting later gene reactivation in advanced breast cancer.

It is of interest to speculate what triggers chromatin remodeling and repression of TGF-β2 and associated genes in post-selection HMECs and yet DNA methylation, and therefore irreversible epigenetic silencing, is not triggered, as is the case for p16INK4A. Our results shed light on a number of possibilities. First, the degree of expression may be an important factor because p16INK4A is completely inactivated in HMECs whereas expression of TGF-β2 and related genes is commonly suppressed but not completely silenced. This supports the hypothesis that prior gene silencing is an important factor that initiates aberrant DNA methylation and subsequent chromatin remodeling (32, 45). Second, the degree of chromatin modification in the post-selection HMECs may not be sufficient to promote DNA methylation. For example, the CpG island promoter region associated with p16INK4A was enriched >40-fold for H3K9 dimethylation after 16 passages in comparison with a 6-fold enrichment for H3K9 dimethylation across the TGF-β2 CpG island promoter region. Conversely, p16INK4A was associated with a 36-fold enrichment of acetylated H3K9 histones in pre-selection HMECs in comparison with a 2-fold enrichment observed for TGF-β2. This observation supports the hypothesis that it is histone methylation and deacetylation that direct subsequent DNA methylation in cancer cells (46). Lastly, an intriguing possibility is that it is the degree of EZH2 polycomb–directed H3K27 methylation in the pre-selection HMECs that plays a crucial role in determining which genes are susceptible to subsequent DNA methylation. For example, there was a 16-fold enrichment in trimethylated H3K27 in the CpG island promoter region associated with p16INK4A in comparison with a 4-fold difference for TGF-β2 in pre-selection HMEC cells. Recently, Schlesinger et al. (37) reported that genes methylated in cancer are potentially marked in early development by trimethylated H3K27, and in cancer, unlike in normal cells, the presence of the EZH2-containing polycomb complex recruits DNA methyltransferases that lead to de novo methylation (47). Ohm et al. (48) and Widschwendter et al. (49) find that genes enriched for trimethylated H3K27 in stem and progenitor cells are also common cancer-specific promoter DNA methylation targets. Members of TGF-β signaling pathway also were recently identified as genes in embryonic cells commonly bound by polycomb proteins in a genome-wide screen (50). It is therefore interesting to speculate that post-selection HMECs are derived from a small subset of progenitor cells found in normal mammary epithelial cells and one of the first predisposing steps toward malignancy, which occurs early in post-selection, is associated with a change in expression and concordant chromatin remodeling of polycomb-associated genes, but not all are susceptible to subsequent DNA methylation and permanent repression.

Our study is the first to report that multiple members of the TGF-β pathway are commonly down-regulated in post-selection HMECs that have acquired an enhanced proliferative capacity in culture and exhibit some of the early events in breast carcinogenesis. We show that repression of these genes is not caused by DNA hypermethylation in either post-selection HMECs or breast cancer cells but by chromatin repression. Interestingly, treatment of breast cancer cells with the demethylation and deacetylation agents was able to induce expression of these pathway genes despite the genes being unmethylated. Gene activation and changes in histone methylation of unmethylated genes in cancer cells after treatment with 5-aza-dC have been previously reported, suggesting that this may be a common phenomenon (30, 5153). Therefore, the use of epigenetic therapy in cancer for gene reactivation should be approached with caution because the activation of the TGF-β pathway may have deleterious effects in stimulating tumor progression. Our findings of coordinate down-regulation of multiple members of a gene pathway through epigenetic remodeling will have major implications in the future understanding of breast cancer initiation and progression.

Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).

Grant support: NSW Cancer Institute, National Breast Cancer Foundation (NBCF), and National Health and Medical Research Council (NH&MRC) project grants; Dora Lush Biomedical Postgraduate scholarship from the NH&MRC, Cancer Institute NSW Research Scholar Award, and NBCF Excellence Award (R.A. Hinshelwood); and NH&MRC fellowships (S.J. Clark and R.R. Reddel).

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

We thank Prof. R. Sutherland and the Garvan Cancer program for breast cancer samples, Drs. Cholmundely (Sydney Adventist Hospital, Australia) and Curtin (Westmead Private Hospital, Australia) for providing reduction mammoplasty samples, Dr. Will Hughes for invaluable technical assistance with microscopy, and the Australian Cancer Research Foundation Facility for DNA sequencing.

1
Baylin SB, Herman JG. DNA hypermethylation in tumorigenesis: epigenetics joins genetics.
Trends Genet
2000
;
16
:
168
–74.
2
Wilson AS, Power BE, Molloy PL. DNA hypomethylation and human diseases.
Biochim Biophys Acta
2007
;
1775
:
138
–62.
3
Jones PA. The DNA methylation paradox.
Trends Genet
1999
;
15
:
34
–7.
4
Gardiner-Garden M, Frommer M. CpG islands in vertebrate genomes.
J Mol Biol
1987
;
196
:
261
–82.
5
Rice JC, Briggs SD, Ueberheide B, et al. Histone methyltransferases direct different degrees of methylation to define distinct chromatin domains.
Mol Cell
2003
;
12
:
1591
–8.
6
Richon VM, Sandhoff TW, Rifkind RA, Marks PA. Histone deacetylase inhibitor selectively induces p21WAF1 expression and gene-associated histone acetylation.
Proc Natl Acad Sci U S A
2000
;
97
:
10014
–9.
7
Hammond SL, Ham RG, Stampfer MR. Serum-free growth of human mammary epithelial cells: rapid clonal growth in defined medium and extended serial passage with pituitary extract.
Proc Natl Acad Sci U S A
1984
;
81
:
5435
–9.
8
Stampfer MR, Bartley JC. Induction of transformation and continuous cell lines from normal human mammary epithelial cells after exposure to benzo[a]pyrene.
Proc Natl Acad Sci U S A
1985
;
82
:
2394
–8.
9
Romanov SR, Kozakiewicz BK, Holst CR, Stampfer MR, Haupt LM, Tlsty TD. Normal human mammary epithelial cells spontaneously escape senescence and acquire genomic changes.
Nature
2001
;
409
:
633
–7.
10
Huschtscha LI, Noble JR, Neumann AA, et al. Loss of p16INK4 expression by methylation is associated with life span extension of human mammary epithelial cells.
Cancer Res
1998
;
58
:
3508
–12.
11
Tlsty TD, Romanov SR, Kozakiewicz BK, Holst CR, Haupt LM, Crawford YG. Loss of chromosomal integrity in human mammary epithelial cells subsequent to escape from senescence.
J Mammary Gland Biol Neoplasia
2001
;
6
:
235
–43.
12
Holst CR, Nuovo GJ, Esteller M, et al. Methylation of p16(INK4a) promoters occurs in vivo in histologically normal human mammary epithelia.
Cancer Res
2003
;
63
:
1596
–601.
13
Brenner AJ, Stampfer MR, Aldaz CM. Increased p16 expression with first senescence arrest in human mammary epithelial cells and extended growth capacity with p16 inactivation.
Oncogene
1998
;
17
:
199
–205.
14
Foster SA, Wong DJ, Barrett MT, Galloway DA. Inactivation of p16 in human mammary epithelial cells by CpG island methylation.
Mol Cell Biol
1998
;
18
:
1793
–801.
15
Crawford YG, Gauthier ML, Joubel A, et al. Histologically normal human mammary epithelia with silenced p16(INK4a) overexpress COX-2, promoting a premalignant program.
Cancer Cell
2004
;
5
:
263
–73.
16
Tlsty TD, Crawford YG, Holst CR, et al. Genetic and epigenetic changes in mammary epithelial cells may mimic early events in carcinogenesis.
J Mammary Gland Biol Neoplasia
2004
;
9
:
263
–74.
17
Kretzschmar M. Transforming growth factor-β and breast cancer: transforming growth factor-β/SMAD signaling defects and cancer.
Breast Cancer Res
2000
;
2
:
107
–15.
18
Massague J. TGF-β signal transduction.
Annu Rev Biochem
1998
;
67
:
753
–91.
19
Yingling JM, Blanchard KL, Sawyer JS. Development of TGF-β signalling inhibitors for cancer therapy.
Nat Rev Drug Discov
2004
;
3
:
1011
–22.
20
Benson JR. Role of transforming growth factor β in breast carcinogenesis.
Lancet Oncol
2004
;
5
:
229
–39.
21
Donovan J, Slingerland J. Transforming growth factor-β and breast cancer: cell cycle arrest by transforming growth factor-β and its disruption in cancer.
Breast Cancer Res
2000
;
2
:
116
–24.
22
Kang SH, Bang YJ, Im YH, et al. Transcriptional repression of the transforming growth factor-β type I receptor gene by DNA methylation results in the development of TGF-β resistance in human gastric cancer.
Oncogene
1999
;
18
:
7280
–6.
23
Ammanamanchi S, Brattain MG. 5-AzaC treatment enhances expression of transforming growth factor-β receptors through down-regulation of Sp3.
J Biol Chem
2001
;
276
:
32854
–9.
24
Osada H, Tatematsu Y, Masuda A, et al. Heterogeneous transforming growth factor (TGF)-β unresponsiveness and loss of TGF-β receptor type II expression caused by histone deacetylation in lung cancer cell lines.
Cancer Res
2001
;
61
:
8331
–9.
25
Ammanamanchi S, Brattain MG. Restoration of transforming growth factor-β signalling through receptor RI induction by histone deacetylase activity inhibition in breast cancer cells.
J Biol Chem
2004
;
279
:
32620
–5.
26
Osada H, Tatematsu Y, Sugito N, Horio Y, Takahashi T. Histone modification in the TGFβRII gene promoter and its significance for responsiveness to HDAC inhibitor in lung cancer cell lines.
Mol Carcinog
2005
;
44
:
233
–41.
27
Cheng KH, Ponte JF, Thiagalingam S. Elucidation of epigenetic inactivation of SMAD8 in cancer using targeted expressed gene display.
Cancer Res
2004
;
64
:
1639
–46.
28
Li Q, Ahuja N, Burger PC, Issa JP. Methylation and silencing of the Thrombospondin-1 promoter in human cancer.
Oncogene
1999
;
18
:
3284
–9.
29
Hui R, Ball JR, Macmillan RD, et al. EMS1 gene expression in primary breast cancer: relationship to cyclin D1 and oestrogen receptor expression and patient survival.
Oncogene
1998
;
17
:
1053
–9.
30
Frigola J, Song J, Stirzaker C, Hinshelwood RA, Peinado MA, Clark SJ. Epigenetic remodeling in colorectal cancer results in coordinate gene suppression across an entire chromosome band.
Nat Genet
2006
;
38
:
540
–9.
31
Clark SJ, Harrison J, Paul CL, Frommer M. High sensitivity mapping of methylated cytosines.
Nucleic Acids Res
1994
;
22
:
2990
–7.
32
Stirzaker C, Song JZ, Davidson B, Clark SJ. Transcriptional gene silencing promotes DNA hypermethylation through a sequential change in chromatin modifications in cancer cells.
Cancer Res
2004
;
64
:
3871
–7.
33
Zhao W, Soejima H, Higashimoto K, et al. The essential role of histone H3 Lys9 dimethylation and MeCP2 binding in MGMT silencing with poor DNA methylation of the promoter CpG island.
J Biochem (Tokyo)
2005
;
137
:
431
–40.
34
McDermott KM, Zhang J, Holst CR, Kozakiewicz BK, Singla V, Tlsty TD. p16(INK4a) prevents centrosome dysfunction and genomic instability in primary cells.
PLoS Biol
2006
;
4
:
e51
.
35
Hosobuchi M, Stampfer MR. Effects of transforming growth factor β on growth of human mammary epithelial cells in culture.
In Vitro Cell Dev Biol
1989
;
25
:
705
–13.
36
Keshet I, Schlesinger Y, Farkash S, et al. Evidence for an instructive mechanism of de novo methylation in cancer cells.
Nat Genet
2006
;
38
:
149
–53.
37
Schlesinger Y, Straussman R, Keshet I, et al. Polycomb-mediated methylation on Lys27 of histone H3 pre-marks genes for de novo methylation in cancer.
Nat Genet
2007
;
39
:
232
–6.
38
Tan LW, Bianco T, Dobrovic A. Variable promoter region CpG island methylation of the putative tumor suppressor gene Connexin 26 in breast cancer.
Carcinogenesis
2002
;
23
:
231
–6.
39
Berman H, Zhang J, Crawford YG, et al. Genetic and epigenetic changes in mammary epithelial cells identify a subpopulation of cells involved in early carcinogenesis.
Cold Spring Harb Symp Quant Biol
2005
;
70
:
317
–27.
40
Reynolds PA, Sigaroudinia M, Zardo G, et al. Tumor suppressor p16INK4A regulates polycomb-mediated DNA hypermethylation in human mammary epithelial cells.
J Biol Chem
2006
;
281
:
24790
–802.
41
Elliott RL, Blobe GC. Role of transforming growth factor β in human cancer.
J Clin Oncol
2005
;
23
:
2078
–93.
42
Stampfer MR, Bodnar A, Garbe J, et al. Gradual phenotypic conversion associated with immortalization of cultured human mammary epithelial cells.
Mol Biol Cell
1997
;
8
:
2391
–405.
43
Chen T, Carter D, Garrigue-Antar L, Reiss M. Transforming growth factor β type I receptor kinase mutant associated with metastatic breast cancer.
Cancer Res
1998
;
58
:
4805
–10.
44
Lynch MA, Nakashima R, Song H, et al. Mutational analysis of the transforming growth factor β receptor type II gene in human ovarian carcinoma.
Cancer Res
1998
;
58
:
4227
–32.
45
Clark SJ, Melki J. DNA methylation and gene silencing in cancer: which is the guilty party?
Oncogene
2002
;
21
:
5380
–7.
46
Bachman KE, Park BH, Rhee I, et al. Histone modifications and silencing prior to DNA methylation of a tumor suppressor gene.
Cancer Cell
2003
;
3
:
89
–95.
47
Vire E, Brenner C, Deplus R, et al. The Polycomb group protein EZH2 directly controls DNA methylation.
Nature
2006
;
439
:
871
–4.
48
Ohm JE, McGarvey KM, Yu X, et al. A stem cell-like chromatin pattern may predispose tumor suppressor genes to DNA hypermethylation and heritable silencing.
Nat Genet
2007
;
39
:
237
–42.
49
Widschwendter M, Fiegl H, Egle D, et al. Epigenetic stem cell signature in cancer.
Nat Genet
2007
;
39
:
157
–8.
50
Bracken AP, Dietrich N, Pasini D, Hansen KH, Helin K. Genome-wide mapping of Polycomb target genes unravels their roles in cell fate transitions.
Genes Dev
2006
;
20
:
1123
–36.
51
Nguyen CT, Weisenberger DJ, Velicescu M, et al. Histone H3-lysine 9 methylation is associated with aberrant gene silencing in cancer cells and is rapidly reversed by 5-aza-2'-deoxycytidine.
Cancer Res
2002
;
62
:
6456
–61.
52
Hellebrekers DM, Melotte V, Vire E, et al. Identification of epigenetically silenced genes in tumor endothelial cells.
Cancer Res
2007
;
67
:
4138
–48.
53
Wozniak RJ, Klimecki WT, Lau SS, Feinstein Y, Futscher BW. 5-Aza-2′-deoxycytidine-mediated reductions in G9A histone methyltransferase and histone H3 K9 dimethylation levels are linked to tumor suppressor gene reactivation.
Oncogene
2007
;
26
:
77
–90.
54
Neve RM, Chin K, Fridlyand J, et al. A collection of breast cancer cell lines for the study of functionally distinct cancer subtypes.
Cancer Cell
2006
;
10
:
515
–27.