DNA Methylation and Complete Transcriptional Silencing of Cancer Genes Persist after Depletion of EZH2

The polycomb group (PcG) proteins in the EED-EZH2 complex are involved in gene silencing in multiple settings including X-chromosome inactivation and developmental gene silencing (1). This silencing is associated with a specific lysine methylation pattern where histone 3 Lysine 27 is trimethylated (H3K27me3) by the histone methyltransferase EZH2 (2). The role of PcG proteins and H3K27me3 has particularly been stressed for maintenance of embryonic stem cell fate through controlling a key subset of genes that contains promoter CpG islands and for which expression must be delayed until it is required for proper cell commitment and specification of cell lineages (3–6). This above role for PcG proteins in embryonic stem cells has recently assumed great relevance for human cancer in that several groups, including our own, have shown that many genes that become DNA hypermethylated for promoter CpG islands and silenced in adult cancers are marked at their promoters, even in the absence of DNA methylation, by PcG components in embryonic stem cells (7–9).

EZH2 is overexpressed in several types of cancer, and the levels of expression correlate with cancer aggressiveness (1, 10, 11). Recently, we have shown that trimethylation of H3K27 and EZH2 binding are enriched along multiple DNA hypermethylated and silenced gene promoters in colon and breast cancer cells (12). All of these above data support a link between EZH2-mediated H3K27 trimethylation and aberrant DNA methylation. However, experimental evidence connecting the actual functional link between these PcG components and aberrant epigenetic cancer gene silencing is limited. To this effect, there is evidence of a modest decrease in global DNA methylation and gene re-expression after introduction of a dominant-negative H3K27R mutant in ovarian cancer cells, although a reduction in promoter DNA methylation is not observed (13). Importantly, the effects of a lysine-to-arginine substitution may be less specific than a discrete loss of H3K27me3, as evidenced by a greater reduction in global DNA methylation in the K27R mutant compared with EZH2 knockdown (13). Another recent study used small interfering RNA (siRNA) to reduce EZH2 levels in an osteosarcoma cell line and proposed that EZH2 directly controls both initiation and maintenance of DNA methylation (14). However, we now present findings suggesting that although EZH2 may function to hold genes in a basally low transcription state in the relative absence of DNA methylation, this protein and the H3K27me3 mark that it catalyzes are not solely responsible for maintenance of transcriptional repression at heavily DNA hypermethylated tumor suppressor genes.

Cell culture. SW480 and U2OS cells were maintained in McCoy's 5A modified medium, and RKO cells were maintained in MEM. All media (Invitrogen) were supplemented with 10% fetal bovine serum (Gemini Bio-Products) and 1% penicillin/streptomycin (Invitrogen) and grown at 37°C in 5% CO₂ atmosphere.

Transient siRNA transfection. RKO or U2OS cells were transfected with either a non-targeting control or EZH2 targeting siRNA (Dharmacon D-001210-01 and D-004218-01) using LipofectAMINE 2000 (Invitrogen). Cells were transfected at a 25 nmol/L siRNA concentration once and then transfected again 24 h later and continued to be passaged and transfected every other day for up to 8 days.

Stable siRNA knockdown. A non-targeting or EZH2 targeting shRNA was cloned into the pSuper vector (OligoEngine), and RKO cells were transfected with LipofectAMINE 2000 (Invitrogen). The transfectants were selected with Puromycin, and individual clones were isolated.

Western blot analysis. Nuclear extracts were isolated using the NE-PER kit (Pierce). Western blot analysis was conducted using antibodies against EZH2 (Upstate 07-400), glyceraldehyde-3-phosphate dehydrogenase (Abcam), β-actin (Sigma), H3K9me2, or H3K27me3 (15).

Reverse transcription-PCR and methylation-specific PCR. Reverse transcription-PCR (RT-PCR) and methylation-specific PCR (MSP) were done as previously described (16). Primer sequences are available upon request.

Chromatin immunoprecipitation. Chromatin immunoprecipitation analysis was done as previously described (16). Antibodies against EZH2 (Upstate), monomethyl, dimethyl, or trimethyl H3K27; trimethyl H3K9 (15); RNA Polymerase (Pol) II (Epiquick); or IgG (Santa Cruz) were used.

PCR amplification and analysis. Previously designed primers targeting the promoter region of the hMLH1 (16), MYT1 (17), WNT1 (17), and p16^INK4a (18) genes were used to analyze promoter occupancy. Primers were purchased from Integrated DNA Technologies. All PCR reactions were done in a total volume of 25 μL, using 2 μL of either immunoprecipitated (bound) DNA, a 1:100 dilution of non-immunoprecipitated (input) DNA, or a no antibody or IgG control; 10 μL of PCR product were size fractionated by PAGE and were quantified using Kodak Digital Science 1D Image Analysis software. Enrichment was calculated by taking the ratio between the net intensity of the gene promoter PCR products from each primer set for the bound, immunoprecipitated sample and the net intensity of the PCR product for the non-immunoprecipitated input sample. Values for enrichment were calculated as the average from at least two independent chromatin immunoprecipitation experiments and multiple independent PCR analyses (three PCR reactions for each primer set used per independent chromatin immunoprecipitation).

Bisulfite sequencing. Genomic DNA was extracted from RKO and U2OS cells that were treated with non-targeting or EZH2 siRNA; 1 μg of DNA was bisulfite modified as previously described (19), and MYT1, WNT1, p16^INK4a, and hMLH1 promoters were analyzed (primer sequences are available upon request). All PCRs were done using JumpStart Red Taq DNA Polymerase. PCR products were run on 1% agarose gels, and bands were excised using QIAquick Gel Extraction kit following the manufacturer's instructions (Qiagen). Purified bands were cloned using TOPO-TA cloning kit following the manufacturer's instructions (Invitrogen). Colonies were selected and grown overnight in Luria-Bertani medium containing ampicillin (100 μg/mL) with shaking at 37°C. Plasmid DNA was isolated using QIAprep Spin Miniprep kit following the manufacturer's instructions (Qiagen). Plasmids were screened for inserts by EcoRI digestion and sequenced using the M13 universal reverse primer (Invitrogen).

EZH2 knockdown is not sufficient for loss of DNA methylation or gene re-expression of the tumor suppressor gene hMLH1. We have previously shown that EZH2 is localized to the hMLH1 promoter when it is silent, and DNA hypermethylated, in RKO colorectal cancer cells (12). EZH2 was not found at the unmethylated, active hMLH1 promoter in SW480 colorectal cancer cells or at the demethylated, reactivated hMLH1 promoter in RKO cells treated with 5-aza-2′deoxycytidine (12). To first examine the role of EZH2 in hypermethylated tumor suppressor gene silencing, we used siRNA to transiently knockdown EZH2 in RKO cells. Knockdown of EZH2 was associated with a global decrease in H3K27 trimethylation, with no effect on dimethylation of H3K9 (Fig. 1A). EZH2 knockdown additionally caused depletion of the protein at the hMLH1 gene promoter (Fig. 1B). Chromatin immunoprecipitation analysis also showed that at the promoter, H3K27me3 sharply decreased after EZH2 knockdown, whereas H3K27me1 actually increased, perhaps suggesting that H3K27me1 is a substrate for EZH2 (Fig. 1B). Importantly, even with the loss of EZH2 and the H3K27me3 mark at the hMLH1 promoter, there was no gene re-expression or loss of DNA methylation associated with these changes (Fig. 1C and D).

To test whether the transient knockdown allowed enough time for gene re-expression or for sufficient depletion of the protein to fully affect its catalyzed mark, we also stably knocked down EZH2 in RKO colorectal cancer cells using a retroviral pSuper vector with a short hairpin loop targeting EZH2. Again, knockdown of EZH2 caused a global decrease of EZH2 and H3K27me3 (Fig. 3A; data not shown) and a significant decrease in EZH2 and H3K27me3 at the hMLH1 promoter (Supplementary Fig. S1) similar to the transient approach. Despite long-term EZH2 depletion at the hMLH1 promoter, gene re-expression or DNA methylation changes were not observed (Fig. 3C; Supplementary Fig. S2).

Comparison of EZH2 target genes shows differing basal states in two cell lines. Vire et al. recently reported that knockdown of EZH2 in U2OS cells led to activation of the MYT1 gene and loss of DNA methylation at some CpG sites in both the MYT1 and WNT1 gene promoters (14). To address the discrepancies between our observation with the hMLH1 gene and this recent data suggesting that EZH2 is necessary for gene silencing, and to understand the role that EZH2 has in a cell and gene specific manner, we compared the U2OS and RKO cells. Interestingly, RT-PCR showed that although MYT1 and WNT1 were completely silent in RKO colorectal cancer cells, they were basally expressed in U2OS cells (Fig. 2A). We also observed that the tumor suppressor gene p16^INK4a was completely silent in both cell lines and could thus serve as an excellent control gene for our studies (Fig. 2A).

To determine if chromatin status was responsible for the differences in basal expression of MYT1 and WNT1 in the two cell lines, we did chromatin immunoprecipitation at the MYT1, WNT1, and p16^INK4a gene promoters in both cell lines. We first found that whereas EZH2 is equally enriched at all three gene promoters in both cell lines, H3K27me3 is distinctly less abundant at these promoters in U2OS cells compared with RKO cells (Fig. 3B and C). Thus, the chromatin marks are consistent among all three genes. This suggests that the lower H3K27me3 level is not responsible for active transcription of MYT1 and WNT1 in U2OS cells because p16^INK4a shares this chromatin signature but is silent. We also examined these gene promoters for RNA Pol II association, and consistent with the differences in basal transcription state, RNA Pol II is strongly associated with MYT1 and WNT1 but not p16^INK4a in the U2OS cells (Fig. 2B and C) and not with any of the three genes in the RKO cells (Fig. 2B and C).

We additionally examined whether other repressive chromatin marks might associate with the genes being examined in the two cell lines. MYT1 and WNT1 were originally identified as EZH2 target genes in the SW480 colorectal cancer cell line (17). In the SW480 cell line, the promoters of MYT1 and WNT1 were reported to associate with H3K27me3 but were lacking another silencing mark, H3K9me3 (17). We found that all DNA hypermethylated tumor suppressor genes we have previously examined had promoters enriched in both of these silencing marks (12). We examined the level of H3K9me3 at each of the promoters in both cell lines and found that MYT1, WNT1, and p16^INK4a all had equal levels of H3K9me3 in both RKO and U2OS, suggesting that it was not the H3K9me3 mark that contributed to the different expression status of these genes.

EZH2 knockdown results in increased expression of unmethylated and basally expressing genes but not of completely silenced and hypermethylated tumor suppressor genes. To further probe the role of EZH2 for the genes under study, we next examined the effect of EZH2 knockdown in both cell lines. Although we were able to significantly decrease levels of EZH2 in both cell lines (Fig. 3A), varying downstream effects were observed. U2OS cells were collected after 3 days of EZH2 knockdown, and the associated depletion of EZH2 caused an increase in expression of MYT1 and WNT1, consistent with the Vire et al. data, but did not re-express fully silenced and hypermethylated p16^INK4a in these same cells (Fig. 3B). Alternatively, expression of MYT1 and WNT1 were unaffected by EZH2 knockdown in RKO cells (Fig. 3C). Thus, although all three genes share a similar chromatin pattern in the U2OS cells, only the basally expressed genes show any expression changes after EZH2 knockdown.

Finally, we employed bisulfite sequencing to examine whether the DNA methylation status, either before or after EZH2 knockdown, might account for the different basal expression states of the genes and differing effects of decreasing EZH2 levels. Contrary to data from the previously published work where they showed 60% and 80% DNA methylation of selected sites in the MYT1 and WNT1 promoters, respectively, we find that both the promoters show minimal basal DNA methylation in U2OS cells (Fig. 4A and B). In sharp contrast, these promoters are fully methylated in RKO cells, correlating with their fully repressed transcription state (Fig. 4A and B). Examination of the silenced p16^INK4a gene reveals that it is also fully DNA methylated in U2OS (Fig. 4C). In addition, contrary to previously published work, we failed to observe a significant reduction in DNA methylation at the MYT1, WNT1, or p16^INK4a promoters in either cell line after EZH2 knockdown (Fig. 4A–C).

In light of the present data, from examination of hMLH1, p16^INK4a, and especially MYT1 and WNT1 in two cancer cell lines, we conclude that EZH2 is not required for the maintenance of dense promoter DNA methylation and the transcriptional silencing of genes associated with this promoter change. This conclusion is particularly compelling from the comparative studies of MYT1 and WNT1 genes in U2OS versus RKO cells. When these genes are not truly regulated by DNA methylation, and when basal expression is evident, as in U2OS cells, depletion of EZH2 can induce increased transcription. Additionally, we now show that reduction of EZH2 does not largely affect levels of promoter DNA methylation. It is important to note that our data do not rule out, as suggested by Vire et al., a role for EZH2 and its H3K27me3 mark in initial recruitment of DNA methylation (14).

Our data indicate that when EZH2 functions to mediate acute gene changes, it may act as a modulator when the promoter is not associated with dense DNA methylation, as in the case of MYT1 and WNT1 in U2OS cells. Our results for the DNA methylation state of these genes in U2OS cells are contrary to previously published data for reasons that are not clear (14). However, our findings showing minimal DNA methylation of these genes are based on extensive sequencing of multiple individual alleles for their promoter regions and contrasts sharply with their densely methylated state for these same regions in RKO cells.

Our hypotheses are consistent with our recent findings for multiple genes in teratocarcinoma cells, where we found simultaneous enrichment for EZH2, H3K27me3, and the additional repressive H3K9me3 chromatin mark, in the absence of DNA methylation (7). In the teratocarcinoma cells, such genes can be induced by differentiation cues to increase expression (7). Thus, all of our findings stress the dominant role of dense CpG island DNA methylation over a variety of associated repressive chromatin states for tightly maintaining the heritable silencing states of genes. This seems to be an important distinction in the control of certain tumor suppressor genes in normal cells, such as in embryonic stem cells, where the gene promoters are associated with PcG proteins, versus the added feature of aberrant DNA hypermethylation-associated tight transcriptional silencing in adult cancer cells.

Grant support: National Institute of Environmental Health Sciences grant ES11858 (K. McGarvey, E. Greene, J. Fahrner, and S. Baylin). T. Jenuwein is sponsored by the IMP through Boehringer Ingelheim, the European Union (NoE network The Epigenome LSHG-CT-2004-503433), and the Austrian GEN-AU initiative financed by funds from the Austrian Federal Ministry for Education, Science and Culture.

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 Joyce Ohm, Angela Ting, and Helai Mohammad for supplying primers and James Herman for technical advice.