DNA methyltransferase-mediated transcriptional silencing ... - CiteSeerX

Oncogene (2009) 28, 2667–2677

& 2009 Macmillan Publishers Limited All rights reserved 0950-9232/09 $32.00 www.nature.com/onc

ORIGINAL ARTICLE

DNA methyltransferase-mediated transcriptional silencing in malignant glioma: a combined whole-genome microarray and promoter array analysis G Foltz1,2, J-G Yoon1, H Lee1, TC Ryken3, Z Sibenaller3, M Ehrich4, L Hood2 and A Madan1,2 1 Center for Advanced Brain Tumor Treatment, Swedish Neuroscience Institute, Seattle, WA, USA; 2Institute for Systems Biology, Seattle, WA, USA; 3Department of Neurosurgery, University of Iowa, Iowa City, IA, USA and 4Sequenome Inc., San Diego, CA, USA

Epigenetic inactivation of tumor suppressor genes is a common feature in human cancer. Promoter hypermethylation and histone deacetylation are reversible epigenetic mechanisms associated with transcriptional regulation. DNA methyltransferases (DNMT1 and DNMT3b) regulate and maintain promoter methylation and are overexpressed in human cancer. We performed whole-genome microarray analysis to identify genes with altered expression after RNAi-induced suppression of DNMT in a glioblastoma multiforme (GBM) cell line. We then identified genes with both decreased expression and evidence of promoter CpG island hypermethylation in GBM tissue samples using a combined whole-genome microarray transcriptome analysis in conjunction with a promoter array analysis after DNA immunoprecipitation with anti-5-methylcytidine. DNMT1 and 3b knockdown resulted in the restored expression of 308 genes that also contained promoter region hypermethylation. Of these, 43 were also found to be downregulated in GBM tissue samples. Three downregulated genes with hypermethylated promoters and restored expression in response to acute DNMT suppression were assayed for methylation changes using bisulfite sequence analysis of the promoter region after chronic DNMT suppression. Restoration of gene expression was not associated with changes in promoter region methylation, but rather with changes in histone methylation and chromatin conformation. Two of the identified genes exhibited growth suppressive activity in in vitro assays. Combining targeted genetic manipulations with comprehensive genomic and expression analyses provides a potentially powerful new approach for identifying epigenetically regulated genes in GBM. Oncogene (2009) 28, 2667–2677; doi:10.1038/onc.2009.122; published online 25 May 2009 Keywords: epigenetics; histone modifications; methylation; glioblastoma

Correspondence: Dr A Madan or Dr G Foltz, Center for Advanced Brain Tumor Treatment, Swedish Neuroscience Institute, James Tower Suite #570, 500 17th Ave, Seattle, WA 98122, USA. E-mail: [email protected] or [email protected] Received 14 October 2008; revised 6 March 2009; accepted 15 April 2009; published online 25 May 2009

Introduction Glioblastoma multiforme (GBM) is a highly aggressive tumor with extremely poor prognosis. In the United States, it results in approximately 13 000 deaths per year despite surgical, chemotherapy and radiotherapeutic treatments (Behin et al., 2003). Although overall mortality remains high, recent developments have provided a better understanding of key molecular mechanisms, which hold promise for improved therapeutic approaches (Choe et al., 2003; Mischel et al., 2003; Singh et al., 2004; Hegi et al., 2005; Mellinghoff et al., 2005; Hambardzumyan et al., 2006; Beier et al., 2007). Recent studies have also more clearly defined the role of epigenetic modifications in the aberrant silencing of genes associated with cell proliferation, tumor progression, apoptosis, angiogenesis and astrocyte motility in GBM (Foltz et al., 2006; Kim et al., 2006b; Mueller et al., 2007). In these studies, global genomic screens were employed to identify genes upregulated in GBM cells following short-term exposure to histone deacetylase (HDAC) and/or DNA methyltransferase (DNMT) inhibitors. These reports suggest that promoter region hypermethylation of CpG islands and associated histone modifications play a role in the aberrant silencing of genes required to maintain normal cellular proliferation and differentiation. It has been previously established that a significant number of genes may require prolonged treatment with DNMT inhibitors to reverse the repressive effects of aberrant epigenetic silencing (McGarvey et al., 2006; Fontijn et al., 2007). Unfortunately, prolonged treatment often results in cell death due to potent drug toxicity even at low concentrations. As a result, a significant cohort of epigenetically silenced genes is likely to be hidden in these analyses. In mammals, DNMTs are required for cytosine methylation and its maintenance. Aberrant DNMT expression has been observed in primary tumor specimens (including GBM) relative to normal tissue samples (Ahluwalia et al., 2001; Kanai et al., 2001; Mizuno et al., 2001; Sato et al., 2002; Girault et al., 2003; Fang et al., 2004; Arai et al., 2006; Park et al., 2006; Kim et al., 2006a; Lorente et al., 2008). Genetic approaches that disrupt DNMT activity have been recently employed to unmask epigenetic programs that integrate genesilencing networks in tumor cells (Rhee et al., 2002; Leu

Epigenetic regulation of gene expression in GBM G Foltz et al

2668

et al., 2003; Paz et al., 2003; Robert et al., 2003; Yan et al., 2003; Milutinovic et al., 2004; Suzuki et al., 2004; James et al., 2006). These studies indicate that RNA interference (RNAi)-mediated knockdown of DNMT results in the reexpression of epigenetically silenced genes, some of which inhibit in vitro tumor growth. In our recent study, we used whole-genome microarray analysis to identify aberrantly expressed genes in GBM. We found both DNMT1 and DNMT3b to be expressed at higher levels (DNMT1 >2-fold and DNMT3b >3fold) in histologically confirmed GBM tissue samples relative to nontumor brain tissue samples. The expression levels of DNMT3a were found to be similar in tumor and nontumor tissue samples. After confirming these microarray results with real-time PCR (Supplementary Figure S1), we sought to investigate whether silencing DNMT1 or DNMT3b has any effect on epigenetic transcriptional regulation in GBM. In this study, we investigate the role of promoter region DNA hypermethylation as a mediator of gene silencing in GBM using RNAi directed against DNMT1 and DNMT3b. First, we use promoter-specific DNA microarrays in combination with gene expression microarrays to define a subset of genes that are likely to be regulated by promoter methylation in GBM. We then explore the effects of DNMT1 and/or DNMT3b knockdown on transcriptional silencing of target genes using whole-genome microarray analysis. We next determine the epigenetic determinants associated with the promoter regions of selected candidate genes activated by DNMT knockdown using bisulfite sequence analysis and chromatin immunoprecipitation (ChIP) analysis. Our results indicate that changes in histone modifications rather than CpG island methylation are associated with the promoter regions of reactivated genes. These changes result in increased chromatin accessibility, suggesting the opening of repressive chromatin in aberrantly silenced genes. Two of these epigenetically regulated genes exhibited growth inhibitory effects when reexpressed in GBM cells in vitro.

Results Correlation between DNA methylation and gene expression level in GBM tissue samples To characterize the scope of promoter region DNA hypermethylation in GBM tissue samples, we profiled immunoprecipitated methylated DNA derived from histologically confirmed GBM tissue samples and nontumor brain samples. GeneChIP human promoter 1.0R microarrays (Affymetrix, Santa Clara, CA, USA) were probed with immunoprecipitated DNA fragments obtained using methylated cytosine-specific antibodies (meC) and DNA from tumor and nontumor brain samples (10 each) in a standard ChIP assay. The promoter microarray contained 25 bp probes spaced at a high density (over 4.5 million) representing approximately 25 500 promoter regions (7.5 kb upstream to 2.45 Oncogene

downstream of the 50 transcriptional start site of associated genes). Data were analysed using Affymetrix Tiling Array Software Version 1.0.05 as described (Hayashi et al., 2007). Briefly, raw data were quantile normalized and for each probe a dataset was generated with a 200 bp length sliding window. For each individual tumor or nontumor brain sample, hybridization intensities were compared between the meC array and the input array (meC antibody/input ratio). Differential enrichment relative to the input array was then tested for statistical significance. To identify a common cohort across the majority of patient samples, the 10 tumor and nontumor samples were then treated as biological replicates allowing for the identification of a subset of probes enriched in at least 70% of specimens. Selection of promoters enriched in at least 7 of 10 samples optimized our ability to identify common epigenetically silenced genes although allowing for the expected variation between individual specimens. To account for nonspecific binding, promoters that showed enrichment also in immunoglobulin G (IgG) microarrays (control) were eliminated from further analysis. This analysis led to the identification of 5504 differentially enriched promoters (3103 hypermethylated and 2401 hypomethylated) in tumor samples relative to nontumor brain samples. A comprehensive methylome representing the differentially methylated promoter regions in GBM is shown in Figure 1. We next investigated if these promoters were associated with differential gene expression in the tissue samples. We performed whole-genome microarray analysis of RNA isolated from tumor and nontumor brain samples using the ABI 1700 Human Genome Expression Microarray (Applied Biosystems, Foster City, CA, USA). We identified 2570 differentially expressed genes, 1502 of which were downregulated twofold or more (qo0.05). By comparing the two datasets, we identified 665 downregulated genes that also exhibited promoter hypermethylation in tumor samples relative to normal brain specimens. Conversely, we found 323 genes upregulated in tumor samples that also displayed hypomethylated promoter regions. This suggested that changes in the methylation state of the promoter regions of these genes might regulate their transcriptional activity. Providing further validation of our approach, we identified several genes in our cohort, including BEX1, BEX2, MGMT, p16, DAPK, TMS-1, RASSF1A, CDH1 (Costello et al., 1996; Esteller et al., 1999; Gonzalez-Gomez et al., 2003; Stone et al., 2004; Foltz et al., 2006; Martinez et al., 2007), which have been previously shown to be expressed at lower levels in GBM due to promoter hypermethylation. We further validated our microarray analysis by using bisulfite sequence analysis to confirm hypermethylation of promoter region CpG islands in selected genes (see below). Loss of DNMTs expression results in transcriptional reactivation of silenced genes We conducted small interfering RNA (siRNA)mediated knockdown in T98 GBM cells to determine


2669 hypo-methylated CpG islands hyper-methylated CpG islands

Chr 1

Chr 2

Chr 3

Chr 4

Chr 5

Chr 6

Chr 7

Chr 8

Chr 9

Chr 10

Chr 11

Chr 12

Chr 13

Chr 14

Chr 15

Chr 16

Chr 17

Chr 18

Chr 19

Chr 20

Chr 21

Chr 22

Chr X

Chr Y

Figure 1 Distribution of hyper- and hypomethylated promoter regions across all chromosomes. The methylation status of promoterassociated CpG islands in histologically confirmed glioblastoma multiforme (GBM) tumor or nontumor brain tissue samples (n ¼ 10 each) was investigated by profiling immunoprecipitated DNA using Affymetrix gene chip human promoter 1.0 R arrays. A total of 5504 promoters were found to carry differentially methylated promoters (3103 hypermethylated and 2401 hypomethylated) in 70% of the tumor samples relative to nontumor brain samples. The coordinates for these promoters regions were used to view data using USCS genome browser and correlating the positions of the enriched regions with the transcriptional start sites of the known gene. Chromosome views were generated using dCHIP software.

whether suppression of DNMT levels would modulate promoter hypermethylation and associated changes in transcriptional activity. T98 cells were transfected with three independent siRNA oligonucleotides specific for DNMT1 and/or DNMT3b transcripts along with off-target siRNA controls. Real-time PCR and western blot analysis demonstrated that two of the three siRNA molecules were effective in significantly reducing specific DNMT levels (28–40%; Figure 2a). RNA extracted from cells demonstrating efficient knockdown of DNMT levels was profiled using expression microarrays as described previously (Foltz et al., 2006). We only selected those genes demonstrating differential expression (>2-fold, qo0.05) in response to treatment with every DNMT-specific siRNA but exhibiting no change after treatment with control off-target siRNA molecules. This resulted in elimination of gene expression changes due to off-target effects of siRNA molecules. 1341 genes demonstrated increased expression after DNMT1 knockdown at either the 5, 10 or 30 daytime period relative to control transfections. By comparing this dataset with the 3103 genes previously identified to have hypermethylated promoters (see above), we found 228 genes both upregulated by DNMT1 suppression and also exhibiting potential promoter hypermethylation. Similarly, of the 1377 genes that were upregulated by DNMT3b suppression, 247

were also found to exhibit promoter hypermethylation. Interestingly, a significant overlap (73%) was observed in these two sets of potential DNMT1 and DNMT3b targets (Figure 2b). A complete list of genes activated by DNMT knockdown is provided in Supplementary Table S1. Expression analysis of RNA extracted from T98 cells where both DNMT1 and DNMT3b were silenced using four different combinations of specific siRNA molecules did not yield any additional upregulated genes. Real-time PCR was used to confirm changes in gene expression, which proved to be in excellent agreement with microarray data (Supplementary Figure S2). Reexpression of DNMT activated genes suppresses the growth of glioma cells in culture We then selected the subset of 43 genes that showed reduced expression in GBM tissues, hypermethylation of promoter regions and transcriptional activation in response to DNMT inhibition. We confirmed the relative expression levels of 10 of these genes in tumor and nontumor brain tissue samples using real-time PCR (Figure 3). Hypothesizing that epigenetically silenced genes might play important roles in growth inhibition, we investigated the functional significance of three newly identified genes, DUSP5, SDC2 and TMTC1 by Oncogene


siRNA2 siRNA3

0.6

siRNA3

0.9

siRNA2

siRNA1

siRNA1

1.2

Control

Relative Levels

2670

250 kDa 150 kDa

DNMT1

100 kDa

DNMT3b

Actin 0.3

Actin

0 DNMT1

DNMT3b Genes associated with hyper-methylated CpGs

2824

61

80 167

346

784

Genes activated on DNMT3b inhibition

329 Genes activated on DNMT1 inhibition

Figure 2 (a) Real-time PCR and western blot analysis show two of three siRNA molecules effectively reduce expression of specific DNA methyltransferases (DNMTs) in T98 cells. T98 cells were harvested after 48 h of treatment with siRNA molecules targeting different regions of DNMT1, DNMT3b or nonsilencing siRNA. Samples were analysed using Assay on Demand reagents with housekeeping gene human glutathione synthetase (hGUS) as endogenous control (real-time PCR) or antibodies specific for DNMTs with actin as a control (western blot). (b) Whole-genome microarray analysis to identify epigenetically regulated genes in malignant glioma. Genes upregulated twofold or more (false discovery rate of 5%) by microarray analysis after inhibition of DNMT1 or DNMT3b were compared with those carrying hypermethylated promoters regions in tumor tissue samples (promoter microarray analysis). A total of 308 genes carrying hypermethylated promoters were found to be activated by DNMTs inhibition.

5

Normal Tumor

Relative Expression

4

3

2

1

TMTC1

TAF7L

SDC2

FKBPIB

FKBPIB

DUSP5

DGCR6L

c9orf3

c2orf23

AP1S1

0

Figure 3 Relative transcriptional silencing of epigenetically regulated genes in histologically confirmed glioblastoma multiforme (GBM) tissue samples (n ¼ 30) relative to nontumor brain samples (n ¼ 9; Po0.001). Total RNA was extracted from tissue samples and relative expression level of genes in tumor samples with respect to expression in a nontumor brain sample was determined using real-time PCR. Arbitrary expression units (y axis) were used to normalize fold change relative to expression in one nontumor brain sample. Oncogene

determining the effect of their expression on the growth of T98 cells in vitro. Expression plasmids coding for these genes were transfected into T98 cells and cells were selected in the presence of G418 selection media for 3 weeks. As shown in Figure 4, expression of two genes (SDC2 and TMTC1) resulted in a marked decrease in colony formation compared to control-transfected cells in T98 cells. Chromatin rearrangement rather than DNA methylation correlates with increased transcriptional activity due to DNMT knockdown To gain insight into the mechanisms leading to epigenetic regulation of the three genes, DUSP5, SDC2 and TMTC1, we examined the methylation status of promoter-associated CpG islands in tumor tissue samples using methylation-specific PCR. The promoter regions of all three genes were found to be methylated (Supplementary Figure S3). We then examined the possible changes in promoter methylation of these three target genes after DNMT knockdown. We used two different techniques, both MALDI-TOF MS as well as direct sequencing of cloned PCR products, to confirm the results of bisulfite sequence analysis. This analysis


2671

a

Control

down resulted in relative opening of chromatin and increased accessibility in the promoter region of these genes as measured by a DNA accessibility assay (Figure 7).

DUSP5

Discussion

SDC2

b

TMTC1

120

%age control

90

60

30

0 Control

DUSP5

SDC2

TMTC1

Figure 4 Effect of increased expression of three genes, identified as transcriptionally activated in response to DNA methyltransferase (DNMT) inhibition, on the growth of T98 cells (a). T98 cells were transfected with expression plasmids coding for DUSP5, SDC2 and TMTC1 after which G418 selected colonies were quantified in three independent experiments relative to empty vector control (b).

revealed no significant changes in promoter methylation in response to chronic DNMT silencing (Figure 5 and Supplementary Figure S4). We then asked whether the increased transcriptional activity in response to DNMT suppression was due to changes in histone modifications, which affect chromatin structure. We used ChIP assays to determine the changes associated with methylation levels of histone side chains (H3-K4, H3-K9 and H3-K27) in the promoter regions of these three target genes following DNMT knockdown. Each of these histone modifications affects the access of various elements of transcriptional machinery to chromatin and therefore influences transcriptional activity of associated genes (Shilatifard, 2006). We found a decrease in H3-K9 di- and trimethylation, and in H3-K27 trimethylation levels in the promoters of each of these three genes in response to DNMT suppression (Figure 6). ChIP analysis showed no significant difference in H3-K4 methylation, after DNMT suppression (data not shown). Consistent with the observed changes in histone modifications, we found that DNMT knock-

Recent evidence indicates that inappropriate epigenetic silencing of tumor suppressor genes is as common as mutation-induced disruption in human cancer (Baylin, 2005; Chen and Baylin, 2005; Curtin et al., 2005; Dahl and Guldberg, 2007; Gronbaek et al., 2007; Schuebel et al., 2007). In GBM, both promoter hypermethylation and histone modifications have been shown to modulate the local chromatin environment and mediate expression of several genes critical for cell proliferation, response to therapy, tumor progression, apoptosis, angiogenesis and migration (Hegi et al., 2005; Foltz et al., 2006; Wang et al., 2008). These epigenetically silenced genes are attractive therapeutic targets as their expression can be restored using pharmacological inhibitors. We have recently identified a cohort of genes activated in primary GBM cells following exposure to DNMT and HDAC inhibitors (Foltz et al., 2006). Two other studies have also demonstrated that pharmacological inhibition of DNMT enhances the transcriptional activity of genes exhibiting hypermethylated promoters in GBM (Kim et al., 2006b; Mueller et al., 2007). Unfortunately, pharmacological inhibition of DNMT is limited by cell toxicity and other nonspecific drug side effects with chronic treatment. In this study, we sought to circumvent these limitations by exploring changes in transcriptional activity after specific and chronic suppression of DNMT1 and DNMT3b using siRNA. We have identified a cohort of 43 genes epigenetically silenced in GBM with associated promoter hypermethylation. These promoter hypermethylation patterns have also been shown to be stable, tissue- and tumor specific, and readily detectable by sensitive PCR methods (Dulaimi et al., 2004; Ibanez de Caceres et al., 2004; Tan et al., 2007; Zhang et al., 2007). This comprehensive identification of epigenetically regulated genes in GBM holds considerable promise for early detection, diagnosis and, potentially, prognosis in response to therapy. To gain further knowledge about the function of genes unmasked by targeted DNMT suppression, we utilized GO classification pathway analysis to identify the functional associations of these genes (Supplementary Table S1). This analysis revealed that suppression of DNMT activity results in wide spread regulation of genes across several signaling pathways implicated in tumorigenesis. Some of these genes have already been shown to play an important role in GBM biology such as BAX, BEX1, BMP4, NDRG2, p27Kip1 and TSC22D1 (Deng et al., 2003; Shostak et al., 2003; Foltz et al., 2006; Piccirillo et al., 2006; Abdullah et al., 2007; Schiappacassi et al., 2008). Of note, a significant fraction of these genes (65%) were also found to be upregulated by 5-AzaC, a pharmacological inhibitor of DNMT, in our earlier study. A brief review of our pathway analysis Oncogene


2672 DUSP5 Control DNMT1-siRNA1 DNMT1-siRNA2 DNMT3b-siRNA2 DNMT3b-siRNA3

SDC2 Control DNMT1-siNA1R DNMT1-siRNA2 DNMT3b-siRNA2 DNMT3b-siRNA3

TMTC1

DNMT3b-siRNA3

≈ ≈ ≈ ≈≈ ≈

Control DNMT1-siRNA1 DNMT1-siRNA2 DNMT3b-siRNA2

≈ 0

mCpG

100

Figure 5 Assessment of methylation status of promoter-associated CpG islands following DNA methyltransferases (DNMT1 and DNMT3b) inhibition in T98 cells. DNA samples extracted from T98 cells were bisulfite treated after 30 days of exposure to siRNA molecules or control siRNAs (nonsilencing). Methylation-sensitive PCR was used to amplify CpG islands associated with promoter regions of DUSP5, SDC2 and TMTC1. The resulting products were analysed using Maldi-TOF MS-based method and the extent of methylation for each CpG site was quantitated. Methylation patterns did not change significantly in response to DNMT inhibition in T98 cells.

reveals that many of the genes identified in this study possess the potential for antitumor activity in vivo. Two of the three genes selected from this cohort, SDC2 (FAM129A) and TMTC1, suppressed the growth of T98 GBM cells in vitro. SDC2 is a member of syndecan proteoglycan family and has been shown to be expressed differentially in several different types of cancers (Park et al., 2002; Kim et al., 2003; Han et al., 2004; Kusano et al., 2004; Modrowski et al., 2005; Essner et al., 2006; Fears et al., 2006). These reports have implicated SDC2 as a potential tumor suppressor gene with growth inhibitory properties. Syndecans have also been shown to interact with various signaling molecules and to modulate cell proliferation, migration, adhesion and differentiation (Essner et al., 2006). The function of TMTC1 has not been well characterized. We correlated the expression of SDC2 and TMTC1 with survival using the REMBRANDT database (http://rembrandt.nci. nih.gov). Kaplan–Meier survival analysis revealed that decreased TMTC1 expression is associated with reduced patient survival. No correlation was found between SDC2 expression and patient survival. Of note, suppression of DNMT1 and DNMT3b did not result in growth inhibition of GBM cells using a colony focus assay (Supplementary Figure S5). Alterations in the wideOncogene

spread epigenetic regulation of multiple gene regulatory pathways may have offset the specific actions of SDC2 and TMTC1. Further investigations of the functional implications of these and other epigenetically regulated genes identified in this study will enhance our understanding of key regulatory pathways responsible for malignant transformation of normal brain cells to malignant derivatives. Although much has been written about the association of promoter hypermethylation and transcriptional silencing (Esteller, 2007, 2008; Karpf, 2007; Miranda and Jones, 2007; Calvanese et al., 2008), our findings support an emerging view that histone modifications and chromatin conformation play a dominant role in transcriptional activation. Despite chronic suppression of DNMT, we did not see any change in promoter methylation with bisulfite sequencing. Yet transcriptional activation did occur, with increased and sustained expression after even short-term DNMT suppression. We found specific histone tail modifications in response to DNMT suppression, which are known to lead to a more relaxed or open chromatin which is permissive for transcription (Shilatifard, 2006). We confirm the modulation of chromatin structure by demonstrating increased DNA accessibility in response to DNMT


2673

specific chromatin modifying enzymes, such as SAHA (Merck, Whitehouse Station, NJ, USA), an HDAC inhibitor currently in clinical trial for GBM. As we identify an increasing role for histone modification as an important regulator of gene silencing in GBM, the pharmacological reversal of this epigenetic mechanism remains an attractive therapeutic approach.

DUSP5

Fold Change

1.5

1.0

0.5

0.0 siRNA1

siRNA2

siRNA1

DNMT1

siRNA2

DNMT3b SDC2

Fold Change

1.5

1.0

0.5

0.0 siRNA1

siRNA2

siRNA1

DNMT1

siRNA2

DNMT3b TMTC1

Fold Change

1.5

1.0

0.5

0.0 siRNA1

siRNA2 DNMT1

siRNA1

siRNA2

DNMT3b

H3-K9 monomethyl

H3-K27 monomethyl

H3-K9 dimethyl

H3-K27 dimethyl

H3-K9 trimethyl

H3-K27 trimethyl

Figure 6 Chromatin immunoprecipitation (ChIP) assays on DNA harvested from T98 cells following DNA methyltransferases (DNMT1 or DNMT3b) inhibition. The changes in level of histone H3-K9 mono-, di- and trimethylation, and H3-K27 mono-, di- and trimethylation after DNMT1 or DNMT3b inhibition were evaluated using antibodies specific for these modifications. Normal rabbit immunoglobulin G (IgG) or no antibodies were used as negative controls. PCR products were analysed by agarose gel electrophoresis and quantitated by densitometry. The changes in histone tail modifications were measured relative to input DNA. The assays were done in triplicate.

suppression. This strengthens our earlier published observations that histone modifications appear to be a major regulatory determinant of epigenetic silencing in GBM cells. This is of particular relevance as newer pharmacological agents are introduced, which target

Materials and methods Oligonucleotide promoter array analysis All GBM and nontumor brain tissue samples were collected using protocols approved by the Institutional Review Boards at University of Iowa and Swedish Neuroscience Institute. For whole-genome promoter experiments, genomic DNA was prepared from histologically confirmed GBM tissues and nontumor brain tissue samples using Charge Switch gDNA kit (Invitrogen, Carlsbad, CA, USA), per manufacturer’s protocol. Genomic DNA (30 mg) was sonicated for 50 s with continuous output using a Branson sonifier (Branson Ultrasonics Corporation, Danbury, CT, USA) to produce random fragments ranging in size from 300 to 1000 bp. The lysate was centrifuged for 15 min at 13 200 r.p.m. at 4 1C, after which the supernatant was incubated with protein G-agarose beads (Millipore, Billerica, MA, USA) for 2 h. The slurry was removed by centrifugation at 1000 r.p.m. for 1 min, and the supernatant was divided into three parts. The first part was used as input control and the other two parts were incubated with 10 ml of 5-methylcytosine antibody (Abcam, Cambridge, MA, USA) or no antibody (negative control) at 4 1C overnight. The immunoprecipitated complexes were collected by incubation with protein G-Sepharose beads (Millipore) for 1 h at 4 1C. After washing the beads with buffers (low salt, high salt, LiCl and TE; Millipore), DNA was eluted from the beads using 200 ml of elution buffer (10 ml 20% SDS, 20 ml 1 M NaHCO3, 170 ml water). Eluted DNA was treated with proteinase K and further purified using spin columns (Millipore). The GeneChip Human Promoter 1.0 R Arrays consisting of 4.6 million probes tiled through over 25 500 human promoter regions were used to interrogate methylation content of immunoprecipitated DNA samples as described by Hayashi et al. (2007). Briefly, DNA samples were amplified, fragmented and labeled according to Affymetrix GeneChip Human Promoter Arrays Technical Mannual (Affymetrix). Array Hybridization and scanning were performed at the University of Iowa DNA Core Facility according to manufacturer’s instructions. Data were analysed using Affymetrix Tiling Analysis Software as described earlier (Hayashi et al., 2007) Cell cultures and transfection of small interfering RNA molecules The immortalized T98 GBM cell line (American Type Culture Collection, Manassas, VA, USA) was cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum as described previously (Foltz et al., 2006). Specific siRNA molecules for DNMT1, DNMT3b and nonspecific controls were obtained from Applied Biosystems, and T98 cells (5 104) were transfected with different siRNA molecules using siPORT NeoFX Transfection Kit (Applied Biosystems), per manufacturer’s protocol. Three independent siRNAs targeting different regions of DNMT1 and/or DNMT3b were used to investigate the effects of their knockdown. This ensured off-target effects were not responsible for observed Oncogene


2674 Control DNMT1-siRNA1 DNMT1-siRNA2 DNM3b-siRNA2 DNMT3b-iRNA3

Relative accessibility

1.2

0.8

0.4

0 DUSP5

SDC2

TMTC1

Figure 7 T98 glioma cells display a more open chromatin structure after DNA methyltransferase (DNMT) inhibition. Nuclei isolated from siRNA (DNMT1, DNMT3b or nonsilencing)-treated cells were incubated with DNaseI. Real-time PCR was performed using primers specific for DUSP5, SDC2 and TMTC1 promoter regions to determine the number of target sequences after DNaseI digestion. The accessibility index that defines the degree of chromatin accessibility was determined by the following formula (AI ¼ 2((Ct Dnase treated) – (Ct Untreated))). Relative accessibility of promoter regions in cells treated with DNMT-specific siRNA relative to nonsilencing siRNA molecules were determined by taking the ratio of their accessibility indices.

changes due to silencing of off-target transcripts. At various time points after transfection, the cells were harvested for RNA, DNA and protein preparation as described below. Real-time PCR analysis Total RNA was extracted using Trizol (Invitrogen) from T98 cells, which had been treated every 24 h for 30 days with siRNA molecules. It was further purified using RNeasy (Qiagen, Valencia, CA, USA) and its quality was assessed with the Agilent Bioanalyzer (Palo Alto, CA, USA). Purified RNA (1 mg) was reverse transcribed using random primers, per manufacturer’s protocol (High Capacity cDNA Archive Kit; Applied Biosystems). The resulting cDNA was diluted 20-fold and used as template for real-time PCR. Real-time PCR analysis was performed using Assay-on-Demand gene expression reagents (Assay ID: Hs00945899_m1 for DNMT1 and Hs01003405_m1 for DNMT3b; Applied Biosystems) on the ABI PRISM 7900 HT Sequence Detection System under default conditions: 95 1C for 10 min followed by 40 cycles of 95 1C for 15 s and 60 1C for 1 min. The expression of housekeeping genes human glutathione synthetase (hGUS) was used as the endogenous control and comparative Ct method was used for quantification of the transcripts, per manufacturer’s protocol. Measurement of DCt was performed in triplicate.

Western blot analysis Protein extracts were prepared from siRNA-treated T98 cells for various times. Western blot analysis was done as previously described (Foltz et al., 2006) using primary antibodies directed to DNMT1 or DNMT3b and actin (Abcam; catalogue numbers ab13537, ab13604 and ab6276) as a control. Secondary horseradish peroxidase–conjugated anti-rabbit antibodies (Santa Cruz Biotechnology, Santa Cruz, CA, USA) were used at a dilution of 1:5000. Oncogene

RNA microarray analysis Total RNA was extracted from cells transfected with DNMT1, DNMT3b or control siRNAs as described above. Total RNA (2 mg) was reverse transcribed with the Chemiluminescent RT-IVT Labeling Kit (Applied Biosystems) and hybridized to the ABI 1700 Human Genome Expression Microarray containing 29 098 gene-specific 60-mer oligonucleotide probes, per the manufacturer’s protocol. Data were analysed as described previously (Foltz et al., 2006) and genes that showed expression changes with a threshold of twofold threshold with a false discovery rate of 5% was classified as differentially expressed genes. CpG island bisulfite analysis The methylation status of the promoter-associated CpG islands associated with DUSP5, SDC2 and TMTC1 was determined in tumor samples using methylation-specific PCR. Primers specific for methylated and unmethylated DNA were designed using Methprimer (http://www.urogene.org/ methprimer/index1.html). The sequences of these primers are provided in Supplementary Table S2. Bisulfite modification of genomic DNA (4 mg) was carried out as previously described (Foltz et al., 2006). The reaction mixtures consisted of 20 ng of bisulfite-modified template 1 PCR buffer, 2.5 mmol/l MgCl2 final concentration 400 nmol/l of forward and reverse primers, 200 mmol/l of each dNTP, 1.25 units of AmpliTaq Gold (Applied Biosystems) in a final volume of 50 ml. The cycling protocol started with one cycle of 95 1C for 10 min for enzyme activation, followed 40 cycles of 95 1C for 15 s, 49 1C for 15 s, 72 1C for 15 s and 1 cycle of 72 1C for 8 min. A portion of 5 ml of the PCR products were resolved on 2.5% agarose gels and stained with ethidium bromide. Each plate contained wells with water only (no template control). We also used MALDITOF MS-based method (Ehrich et al., 2005, 2008) to determine the methylation status of selected candidate genes in T98 cells following DNMT knockdown. Briefly, methylation-sensitive PCR was used to amplify 200–600 bp target regions using one of the primers tagged with T7 RNA polymerase promoter. The PCR products were transcribed into single-stranded RNA transcripts that were cleaved in a base-specific manner. The cleaved products were analysed by MALDI-TOF MS. The sequences of all the primers used to amplify DNA for MALDI-TOF are provided in Supplementary Table S3. As an alternate approach, the methylation status of the entire CpG island was evaluated using bisulfite sequencing as previously described (Foltz et al, 2006). Briefly, quantitatively methylated PCR products (400–600 bp) were cloned into TOPO TA (Invitrogen). Plasmid DNA from 10 independent clones were prepared using Sprint prep (Agencourt Biosciences, Beverly, MA, USA) on a Biomek robot and sequenced from both ends with BIG Dye terminators v3.1. Sequences were resolved on Applied Biosystems 3730 XL capillary sequencer and data were analysed using Phred-phrap and consed (Gordon et al, 1998). The extent of methylation at each CpG site was quantitated using bioinformatics tools developed in house. Chromatin immunoprecipitation assays Formaldehyde cross-linking and ChIP assays in T98 cells were performed as described previously (Foltz et al., 2006). Under the various conditions of our experiments, T98 glioma cells were fixed at room temperature with 1% formaldehyde for 10 min. After isolation of nuclei, sonication and lysis the soluble chromatin was precleared with protein G-agarose beads (Millipore). Different antibodies for immunoprecipitation procedure included histone H3 (K9) mono-, di- and


2675 trimethyl (Millipore; catalogue numbers 07–450, 07–441 and 07–442), histone H3 (K27) mono-, di- and trimethyl (Millipore; catalogue numbers 07–448, 07–452 and 07–449) and histone H3 (K4) mono-, di- and trimethyl (Millipore; catalogue numbers 07–436, 07–030 and 07–473) For control experiments a rabbit IgG control antibody (Santa Cruz Biotechnology; catalogue number SC-2027) was used. Immunoprecipitated DNA samples were analysed by PCR using primers specific for candidate genes. Their sequences and positions can be found in Supplementary Table S4. Chromatin accessibility assays Chromatin accessibility assays were performed by the method developed by Rao et al., 2001. Nuclei isolated from DNMTs or control siRNA T98-treated cells at various time periods were suspended in 50 ml of digestion buffer (50 mM Tris pH 7.5, 100 mM NaCl, 10 mM MgCl2 and 1 mM DTT) containing 20 mg of RNAse I (Qiagen). Nuclease digestions were carried out by adding 20 U of RQ1 DNase (Promega, Madison, WI, USA) and incubated at 37 1C for 10 min. DNA was then isolated from nuclei using the DNAEasy kit (Qiagen) according to the manufacturer’s protocol. Real-time PCR was carried out with primers designed to amplify promoter regions of selected candidate genes. Real-time PCR reaction mixtures contained 5 pmol of each primer, 20 ng of DNA and 2 SYBR Green PCR Master Mix (Applied Biosystems). The PCR conditions used were as follows: 10-min denaturation at 95 1C, followed by 40 cycles of 94 1C for 30 s and 60 1C for 30 s. Amplification of the target amplicon was monitored as a function of increased SYBR green fluorescence. An analysis threshold was set, and the cycle threshold (Ct) was computed for each sample. The accessibility index was determined by the following formula (AI ¼ 2((Ct Dnase treated) – (Ct Untreated))). Relative accessibility of promoter regions for various genes in DNMTs

siRNA-treated cells were calculated relative to control siRNAtreated cells by taking the ratio of their accessibility indices. Transfection and colony formation assays Full-length open reading frames for DUSP5, SDC2 and TMTC1 were obtained from Origene (Origene Technologies, Rockville, MD, USA). Clones were sequence verified and colony formation assays were performed in monolayer culture as described earlier (Foltz et al., 2006). Briefly, cells were plated at 1.5 105 per well using six-well plates and transfected with either pc (GOIs) or no insert (mock control) using Trans It-Neural transfection reagents (Mirus, Madison, WI, USA). The cells were selected in G418 (1 mg/ml) supplemented media at 24 h after transfection. Some cells were simultaneously harvested to confirm increased expression of the transfected genes by real-time PCR. G418 resistant cells were maintained for 2 weeks in culture. Cells were detached, resuspended in media containing 0.3% agarose and overlaid on 0.6% agarose. A portion of 0.5 ml media was added to the plates every 4 days and colony formation was quantitated after fixation and staining with methylene blue after 3 weeks. Conflict of interest Dr Ehrich is a shareholder and employee of Sequenom Inc. Acknowledgements We thank DNA core sequencing facility at University of Iowa for its helpful discussions and Dr Gi-Yung Ryu for his help in analysing microarray data. We are thankful to Matt Howard for his support and help in tissue collection.

References Abdullah JM, Ahmad F, Ahmad KA, Ghazali MM, Jaafar H, Ideris A et al. (2007). Molecular genetic analysis of BAX and cyclin D1 genes in patients with malignant glioma. Neurol Res 29: 239–242. Ahluwalia A, Hurteau JA, Bigsby RM, Nephew KP. (2001). DNA methylation in ovarian cancer. II. Expression of DNA methyltransferases in ovarian cancer cell lines and normal ovarian epithelial cells. Gynecol Oncol 82: 299–304. Arai E, Kanai Y, Ushijima S, Fujimoto H, Mukai K, Hirohashi S. (2006). Regional DNA hypermethylation and DNA methyltransferase (DNMT) 1 protein overexpression in both renal tumors and corresponding nontumorous renal tissues. Int J Cancer 119: 288–296. Baylin SB. (2005). DNA methylation and gene silencing in cancer. Nat Clin Pract Oncol 2(Suppl 1): S4–11. Behin A, Hoang-Xuan K, Carpentier AF, Delattre JY. (2003). Primary brain tumours in adults. Lancet 361: 323–331. Beier D, Hau P, Proescholdt M, Lohmeier A, Wischhusen J, Oefner PJ et al. (2007). CD133(+) and CD133() glioblastoma-derived cancer stem cells show differential growth characteristics and molecular profiles. Cancer Res 67: 4010–4015. Calvanese V, Horrillo A, Hmadcha A, Suarez-Alvarez B, Fernandez AF, Lara E et al. (2008). Cancer genes hypermethylated in human embryonic stem cells. PLoS ONE 3: e3294. Chen WY, Baylin SB. (2005). Inactivation of tumor suppressor genes: choice between genetic and epigenetic routes. Cell Cycle 4: 10–12. Choe G, Horvath S, Cloughesy TF, Crosby K, Seligson D, Palotie A et al. (2003). Analysis of the phosphatidylinositol 30 -kinase signaling pathway in glioblastoma patients in vivo. Cancer Res 63: 2742–2746.

Costello JF, Berger MS, Huang HS, Cavenee WK. (1996). Silencing of p16/CDKN2 expression in human gliomas by methylation and chromatin condensation. Cancer Res 56: 2405–2410. Curtin JA, Fridlyand J, Kageshita T, Patel HN, Busam KJ, Kutzner H et al. (2005). Distinct sets of genetic alterations in melanoma. N Engl J Med 353: 2135–2147. Dahl C, Guldberg P. (2007). The genome and epigenome of malignant melanoma. APMIS 115: 1161–1176. Deng Y, Yao L, Chau L, Ng SS, Peng Y, Liu X et al. (2003). N-Myc downstream-regulated gene 2 (NDRG2) inhibits glioblastoma cell proliferation. Int J Cancer 106: 342–347. Dulaimi E, Hillinck J, Ibanez de Caceres I, Al-Saleem T, Cairns P. (2004). Tumor suppressor gene promoter hypermethylation in serum of breast cancer patients. Clin Cancer Res 10: 6189–6193. Ehrich M, Nelson MR, Stanssens P, Zabeau M, Liloglou T, Xinarianos G et al. (2005). Quantitative high-throughput analysis of DNA methylation patterns by base-specific cleavage and mass spectrometry. Proc Natl Acad Sci USA 102: 15785–15790. Ehrich M, Turner J, Gibbs P, Lipton L, Giovanneti M, Cantor C et al. (2008). Cytosine methylation profiling of cancer cell lines. Proc Natl Acad Sci USA 105: 4844–4849. Essner JJ, Chen E, Ekker SC. (2006). Syndecan-2. Int J Biochem Cell Biol 38: 152–156. Esteller M. (2007). Cancer epigenomics: DNA methylomes and histone-modification maps. Nat Rev Genet 8: 286–298. Esteller M. (2008). Epigenetics in cancer. N Engl J Med 358: 1148–1159. Esteller M, Hamilton SR, Burger PC, Baylin SB, Herman JG. (1999). Inactivation of the DNA repair gene O6-methylguanine-DNA Oncogene


2676 methyltransferase by promoter hypermethylation is a common event in primary human neoplasia. Cancer Res 59: 793–797. Fang JY, Cheng ZH, Chen YX, Lu R, Yang L, Zhu HY et al. (2004). Expression of Dnmt1, demethylase, MeCP2 and methylation of tumor-related genes in human gastric cancer. World J Gastroenterol 10: 3394–3398. Fears CY, Gladson CL, Woods A. (2006). Syndecan-2 is expressed in the microvasculature of gliomas and regulates angiogenic processes in microvascular endothelial cells. J Biol Chem 281: 14533–14536. Foltz G, Ryu GY, Yoon JG, Nelson T, Fahey J, Frakes A et al. (2006). Genome-wide analysis of epigenetic silencing identifies BEX1 and BEX2 as candidate tumor suppressor genes in malignant glioma. Cancer Res 66: 6665–6674. Fontijn D, Adema AD, Bhakat KK, Pinedo HM, Peters GJ, Boven E. (2007). O6-methylguanine-DNA-methyltransferase promoter demethylation is involved in basic fibroblast growth factor induced resistance against temozolomide in human melanoma cells. Mol Cancer Ther 6: 2807–2815. Girault I, Tozlu S, Lidereau R, Bieche I. (2003). Expression analysis of DNA methyltransferases 1, 3A, and 3B in sporadic breast carcinomas. Clin Cancer Res 9: 4415–4422. Gonzalez-Gomez P, Bello MJ, Arjona D, Lomas J, Alonso ME, De Campos JM et al. (2003). Promoter hypermethylation of multiple genes in astrocytic gliomas. Int J Oncol 22: 601–608. Gordon D, Abajian C, Green P. (1998). Consed: a graphical tool for sequence finishing. Genome Res 8: 195–202. Gronbaek K, Hother C, Jones PA. (2007). Epigenetic changes in cancer. APMIS 115: 1039–1059. Hambardzumyan D, Squatrito M, Holland EC. (2006). Radiation resistance and stem-like cells in brain tumors. Cancer Cell 10: 454–456. Han I, Park H, Oh ES. (2004). New insights into syndecan-2 expression and tumourigenic activity in colon carcinoma cells. J Mol Histol 35: 319–326. Hayashi H, Nagae G, Tsutsumi S, Kaneshiro K, Kozaki T, Kaneda A et al. (2007). High-resolution mapping of DNA methylation in human genome using oligonucleotide tiling array. Hum Genet 120: 701–711. Hegi ME, Diserens AC, Gorlia T, Hamou MF, de Tribolet N, Weller M et al. (2005). MGMT gene silencing and benefit from temozolomide in glioblastoma. N Engl J Med 352: 997–1003. Ibanez de Caceres I, Battagli C, Esteller M, Herman JG, Dulaimi E, Edelson MI et al. (2004). Tumor cell-specific BRCA1 and RASSF1A hypermethylation in serum, plasma, and peritoneal fluid from ovarian cancer patients. Cancer Res 64: 6476–6481. James SR, Link PA, Karpf AR. (2006). Epigenetic regulation of X-linked cancer/germline antigen genes by DNMT1 and DNMT3b. Oncogene 25: 6975–6985. Kanai Y, Ushijima S, Kondo Y, Nakanishi Y, Hirohashi S. (2001). DNA methyltransferase expression and DNA methylation of CPG islands and peri-centromeric satellite regions in human colorectal and stomach cancers. Int J Cancer 91: 205–212. Karpf AR. (2007). Epigenomic reactivation screening to identify genes silenced by DNA hypermethylation in human cancer. Curr Opin Mol Ther 9: 231–241. Kim H, Kwon YM, Kim JS, Han J, Shim YM, Park J et al. (2006a). Elevated mRNA levels of DNA methyltransferase-1 as an independent prognostic factor in primary nonsmall cell lung cancer. Cancer 107: 1042–1049. Kim TY, Zhong S, Fields CR, Kim JH, Robertson KD. (2006b). Epigenomic profiling reveals novel and frequent targets of aberrant DNA methylation-mediated silencing in malignant glioma. Cancer Res 66: 7490–7501. Kim Y, Park H, Lim Y, Han I, Kwon HJ, Woods A et al. (2003). Decreased syndecan-2 expression correlates with trichostatin-A induced-morphological changes and reduced tumorigenic activity in colon carcinoma cells. Oncogene 22: 826–830. Kusano Y, Yoshitomi Y, Munesue S, Okayama M, Oguri K. (2004). Cooperation of syndecan-2 and syndecan-4 among cell surface Oncogene

heparan sulfate proteoglycans in the actin cytoskeletal organization of Lewis lung carcinoma cells. J Biochem (Tokyo) 135: 129–137. Leu YW, Rahmatpanah F, Shi H, Wei SH, Liu JC, Yan PS et al. (2003). Double RNA interference of DNMT3b and DNMT1 enhances DNA demethylation and gene reactivation. Cancer Res 63: 6110–6115. Lorente A, Mueller W, Urdangarin E, Lazcoz P, Lass U, von Deimling A et al. (2008). RASSF1A, BLU, NORE1A, PTEN and MGMT expression and promoter methylation in gliomas and glioma cell lines and evidence of deregulated expression of de novo DNMTs. Brain Pathol. Martinez R, Setien F, Voelter C, Casado S, Quesada MP, Schackert G et al. (2007). CpG island promoter hypermethylation of the proapoptotic gene caspase-8 is a common hallmark of relapsed glioblastoma multiforme. Carcinogenesis 28: 1264–1268. McGarvey KM, Fahrner JA, Greene E, Martens J, Jenuwein T, Baylin SB. (2006). Silenced tumor suppressor genes reactivated by DNA demethylation do not return to a fully euchromatic chromatin state. Cancer Res 66: 3541–3549. Mellinghoff IK, Wang MY, Vivanco I, Haas-Kogan DA, Zhu S, Dia EQ et al. (2005). Molecular determinants of the response of glioblastomas to EGFR kinase inhibitors. N Engl J Med 353: 2012–2024. Milutinovic S, Brown SE, Zhuang Q, Szyf M. (2004). DNA methyltransferase 1 knock down induces gene expression by a mechanism independent of DNA methylation and histone deacetylation. J Biol Chem 279: 27915–27927. Miranda TB, Jones PA. (2007). DNA methylation: the nuts and bolts of repression. J Cell Physiol 213: 384–390. Mischel PS, Shai R, Shi T, Horvath S, Lu KV, Choe G et al. (2003). Identification of molecular subtypes of glioblastoma by gene expression profiling. Oncogene 22: 2361–2373. Mizuno S, Chijiwa T, Okamura T, Akashi K, Fukumaki Y, Niho Y et al. (2001). Expression of DNA methyltransferases DNMT1, 3A, and 3B in normal hematopoiesis and in acute and chronic myelogenous leukemia. Blood 97: 1172–1179. Modrowski D, Orosco A, Thevenard J, Fromigue O, Marie PJ. (2005). Syndecan-2 overexpression induces osteosarcoma cell apoptosis: implication of syndecan-2 cytoplasmic domain and JNK signaling. Bone 37: 180–189. Mueller W, Nutt CL, Ehrich M, Riemenschneider MJ, von Deimling A, van den Boom D et al. (2007). Downregulation of RUNX3 and TES by hypermethylation in glioblastoma. Oncogene 26: 583–593. Park H, Kim Y, Lim Y, Han I, Oh ES. (2002). Syndecan-2 mediates adhesion and proliferation of colon carcinoma cells. J Biol Chem 277: 29730–29736. Park HJ, Yu E, Shim YH. (2006). DNA methyltransferase expression and DNA hypermethylation in human hepatocellular carcinoma. Cancer Lett 233: 271–278. Paz MF, Wei S, Cigudosa JC, Rodriguez-Perales S, Peinado MA, Huang TH et al. (2003). Genetic unmasking of epigenetically silenced tumor suppressor genes in colon cancer cells deficient in DNA methyltransferases. Hum Mol Genet 12: 2209–2219. Piccirillo SG, Reynolds BA, Zanetti N, Lamorte G, Binda E, Broggi G et al. (2006). Bone morphogenetic proteins inhibit the tumorigenic potential of human brain tumour-initiating cells. Nature 444: 761–765. Rao S, Procko E, Shannon MF.. (2001). Chromatin remodeling, measured by a novel real-time polymerase chain reaction assay, across the proximal promoter region of the IL-2 gene. J Immunol 167: 4494–4503. Rhee I, Bachman KE, Park BH, Jair KW, Yen RW, Schuebel KE et al. (2002). DNMT1 and DNMT3b cooperate to silence genes in human cancer cells. Nature 416: 552–556. Robert MF, Morin S, Beaulieu N, Gauthier F, Chute IC, Barsalou A et al. (2003). DNMT1 is required to maintain CpG methylation and aberrant gene silencing in human cancer cells. Nat Genet 33: 61–65. Sato M, Horio Y, Sekido Y, Minna JD, Shimokata K, Hasegawa Y. (2002). The expression of DNA methyltransferases and methylCpG-binding proteins is not associated with the methylation status


2677 of p14(ARF), p16(INK4a) and RASSF1A in human lung cancer cell lines. Oncogene 21: 4822–4829. Schiappacassi M, Lovat F, Canzonieri V, Belletti B, Berton S, Di Stefano D et al. (2008). p27Kip1 expression inhibits glioblastoma growth, invasion, and tumor-induced neoangiogenesis. Mol Cancer Ther 7: 1164–1175. Schuebel KE, Chen W, Cope L, Glockner SC, Suzuki H, Yi JM et al. (2007). Comparing the DNA hypermethylome with gene mutations in human colorectal cancer. PLoS Genet 3: 1709–1723. Shilatifard A. (2006). Chromatin modifications by methylation and ubiquitination: implications in the regulation of gene expression. Annu Rev Biochem 75: 243–269. Shostak KO, Dmitrenko VV, Garifulin OM, Rozumenko VD, Khomenko OV, Zozulya YA et al. (2003). Downregulation of putative tumor suppressor gene TSC-22 in human brain tumors. J Surg Oncol 82: 57–64. Singh SK, Hawkins C, Clarke ID, Squire JA, Bayani J, Hide T et al. (2004). Identification of human brain tumour initiating cells. Nature 432: 396–401. Stone AR, Bobo W, Brat DJ, Devi NS, Van Meir EG, Vertino PM. (2004). Aberrant methylation and down-regulation of TMS1/ASC in human glioblastoma. Am J Pathol 165: 1151–1161.

Suzuki M, Sunaga N, Shames DS, Toyooka S, Gazdar AF, Minna JD. (2004). RNA interference-mediated knockdown of DNA methyltransferase 1 leads to promoter demethylation and gene reexpression in human lung and breast cancer cells. Cancer Res 64: 3137–3143. Tan SH, Ida H, Lau QC, Goh BC, Chieng WS, Loh M et al. (2007). Detection of promoter hypermethylation in serum samples of cancer patients by methylation-specific polymerase chain reaction for tumour suppressor genes including RUNX3. Oncol Rep 18: 1225–1230. Wang Z, Hao Y, Lowe AW. (2008). The adenocarcinoma-associated antigen, AGR2, promotes tumor growth, cell migration, and cellular transformation. Cancer Res 68: 492–497. Yan L, Nass SJ, Smith D, Nelson WG, Herman JG, Davidson NE. (2003). Specific inhibition of DNMT1 by antisense oligonucleotides induces re-expression of estrogen receptor-alpha (ER) in ER-negative human breast cancer cell lines. Cancer Biol Ther 2: 552–556. Zhang W, Bauer M, Croner RS, Pelz JO, Lodygin D, Hermeking H et al. (2007). DNA stool test for colorectal cancer: hypermethylation of the secreted frizzled-related protein-1 gene. Dis Colon Rectum 50: 1618–1626; discussion 1626–1627.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene