Oncogenic HRAS Suppresses Clusterin Expression Through Promoter Hypermethylation

Oncogene (2006) 25, 4890–4903 & 2006 Nature Publishing Group All rights reserved 0950-9232/06 $30.00 www.nature.com/onc ORIGINAL ARTICLE Oncogenic HRAS suppresses clusterin expression through promoter hypermethylation

P Lund1, K Weihaupt1, T Mikeska2, D Jammas2, X Chen1, R-J Kuban3, U Ungethu¨ m3, U Krapfenbauer1, H-P Herzel4, R Scha¨ fer1,3, J Walter2 and C Sers1

1Laboratory of Molecular Tumor Pathology, Institute of Pathology, Charite´, Berlin, Germany; 2Department of Natural Sciences – Technical Faculty III FR 8.3, Biological Sciences, Institute of Genetics/Epigenetics, University of Saarland, Saarbru¨cken, Germany; 3Laboratory of Functional Genome Research, Charite´, Berlin, Germany and 4Institute for Theoretical Biology, Humboldt University Berlin, Berlin, Germany

Silencing of gene expression by methylation of CpG Introduction islands in regulatory elements is frequently observed in cancer. However, an influence of the most common Activation of RAS proteins by point mutations is oncogenic signalling pathways onto DNA methylation frequently observed in cancers of the colon, pancreas, has not yet been investigated thoroughly. To address this thyroid, lung and the myeloid system (Bos, 1989). RAS issue, we identified genes suppressed in HRAS-trans- genes encode small GTP-binding proteins that act as formed rat fibroblasts but upregulated after treatment major molecular switches in signal transduction pro- with the demethylating agent 5-Aza-2-deoxycytidine and cesses, transmitting extracellular signals to cytoplasmic with the MEK1,2 inhibitor U0126. Analysis of gene signalling cascades (Campbell et al., 1998). The onco- expression by microarray and Northern blot analysis genic forms of RAS proteins are locked in their active revealed the MEK/ERK target genes clusterin, matrix state and contribute to transformation, angiogenesis, metalloproteinase 2 (Mmp2), peptidylpropyl isomerase invasion and metastasis by constitutive activation of C-associated protein, syndecan 4, Timp2 and Thbs1 to be downstream pathways. Among these are the RAF/ repressed in the HRAS-transformed FE-8 cells in a MEK/ MEK/ERK (MAPK) cascade of cytoplasmic kinases ERK- and methylation-dependent manner. Hypermethy- (Mansour et al., 1994), the phosphoinositide 3-phos- lation of putative regulatory elements in HRAS-trans- phate lipid kinases (Downward, 1998), the small GTP- formed cells as compared to immortalized fibroblasts binding proteins RAC and RHO (Zohn et al., 1998) was detected within a CpG island 14.5 kb upstream of and the family of guanine nucleotide exchange factors clusterin, within the clusterin promoter and within a CpG (RalGDS, RGL and Rlf/RGL2) that serve as activators island of the Mmp2 promoter by bisulphite sequencing. of the small GTPases RalA and RalB (Wolthuis and Furthermore, hypermethylation of the clusterin promoter Bos, 1999). was observed 10 days after induction of HRAS in RAS signalling is linked to the activation of immortalized rat fibroblasts and a clear correlation transcription factors of the AP-1 and Ets families, and between reduced clusterin expression and hypermethlya- also to NF-kB, SRF and c-Myc, all of which induce tion could also be observed in distinct rat tissues. These a complex set of transcriptional targets. Meanwhile, a results suggest that silencing of individual genes by DNA number of RAS-dependent gene expression profiles have methylation is controlled by oncogenic signalling path- confirmed the quantitative and qualitative impact of ways, yet the mechanisms responsible for initial target RAS signalling on gene expression control (Zuber et al., gene suppression are variable. 2000; Brem et al., 2001; Gadal et al., 2003; Fensterer Oncogene (2006) 25, 4890–4903. doi:10.1038/sj.onc.1209502; et al., 2004; Tchernitsa et al., 2004). These studies published online 27 March 2006 revealed that similar numbers of genes are activated and repressed by RAS oncogenes and suggested that RAS Keywords: HRAS(G12V); MEK/ERK signalling; gene can influence the expression of defined groups of genes expression; CpG methylation; clusterin; Mmp2 in a signalling pathway-dependent manner (Tchernitsa et al., 2004). Transcriptional activation of critical RAS target genes involved in cell cycle control, mitogenic signalling or metastasis has been studied in much detail. In contrast, the mechanisms of RAS-dependent gene repression have been unfolded only for a few indivi- Correspondence: Dr C Sers, Laboratory of Molecular Tumorpathol- dual genes. A recent example of RAS-dependent gene ogy, Institute of Pathology, Charite´ , Schumannstr. 20/21, D-10117 repression is the observed transcriptional block onto the Berlin, Germany. E-mail: [email protected] thrombospondin 1 gene (THBS1), which is repressed Received 20 May 2005; revised 8 February 2006; accepted 10 February in a RAS/PI-3K/RHO- and ROCK-dependent manner 2006; published online 27 March 2006 through a phosphorylated c-myc protein (Watnick et al., RAS induces hypermethylation of MAPK target genes P Lund et al 4891 2003). The Ets-2 protein is a member of the ETS family cation and proliferation, excluding cytotoxic effects. of transcription factors directly activated by mito- Western blot analysis demonstrated histone H3 hyper- gen-activated protein kinases and functionally involved acetylation after TSA treatment, whereas the U0126- in the activation of RAS target genes. However, induced inhibition of MEK1,2 reduced the level of p44/ Ets-2 can inhibit BRCA-1 transcription through recruit- 42 ERK phosphorylation. Total DNMT1 protein levels ment of components of the SWI/SNF repressive were identical in the two cell lines and were also not chromatin-remodelling complex (Baker et al., 2003; affected by treatment with U0126 (Figure 1c). Wei et al., 2003). Furthermore, HDAC4, which is a member of the histone deacetylase family, was shown to translocate into the nucleus in cells transfected A distinct set of genes downregulated in FE-8 cells is with oncogenic RAS (Zhou et al., 2000). These obser- concordantly upregulated after treatment with 5-Aza-CdR vations indicate that RAS-dependent signalling path- as well as with U0126 ways alter DNA–protein interactions at regulatory Silencing of genes by methylation, particularly with elements by stimulating the recruitment of components regard to tumour suppressor genes, is frequently des- determining DNA structure. Also, increased expression cribed in cancer cells and tumour samples. However, the of DNMT1 has been detected in tumour cells (Kautiai- mechanisms and signalling pathways involved in gene- nen and Jones, 1986; Issa et al., 1993). Stable expression specific methylation are largely unknown. Therefore, we of HRAS(G12V) induces transcription of DNMT1 via tested the hypothesis that methylation is involved in the an AP-1 site in the promoter region (Rouleau et al., downregulation of genes suppressed in a MEK/ERK- 1995), and elevated DNMT1 levels were correlated with dependent manner in HRAS(G12V)-transformed FE-8 the 5-methylcytosine content and demonstrated to be a cells. prerequisite for fos-mediated transformation (Bakin and First we investigated the expression of 73 RAS target Curran, 1999). Recently, a set of genes was identified genes previously shown to be suppressed in FE-8 cells that was commonly downregulated in rat fibroblasts by Northern blot analysis (Zuber et al., 2000). Eighteen transformed by c-fos,v-fos, Hras(G12V) or Dnmt1 of these genes are regulated in a MEK/ERK-dependent (Ordway et al., 2004) suggesting a close connection and 55 genes are regulated in a MEK/ERK-independent between RAS transformation and epigenetic gene manner. Treatment with 0.1 mM 5-Aza-CdR revealed regulation. a 5-Aza-CdR-dependent upregulation of 10 of the In the current manuscript, we investigated the impact 18 MEK/ERK target genes (Table 1). In addition, 31 of constitutive RAS activation on DNA methylation of 55 MEK/ERK-independent genes downregulated of putative regulatory elements by correlating gene in FE-8 cells are upregulated after treatment with expression data with bisulphite sequencing data. We 5-Aza-CdR, suggesting that alternative RAS-dependent identified a set of MEK/ERK target genes down- pathways influence the methylation of genes (data not regulated in FE-8 cells and re-expressed after treatment shown). with 5-Aza-2-deoxycytidine (5-Aza-CdR) and after To gain further access to candidate genes silenced treatment with the MEK1,2 inhibitor U0126. The by methylation in response to oncogenic signalling clusterin gene revealed suppression due to hypermethy- pathways, we hybridized cDNA probes derived from lation of regulatory regions in cells transformed stably untreated and 5-Aza-CdR-treated 208F and FE-8 cells or inducibly by activated HRAS, suggesting that to an Affymetrix rat U34A oligonucleotide microarray. oncogenic signalling can directly influence DNA After treatment of 208F cells with 0.3 mM 5-Aza-CdR methylation. for 96 h, 108 genes (9%) were significantly upregulated and 169 genes (13%) were significantly downregulated. In FE-8 cells, 115 genes (8%) were significantly Results upregulated after treatment with 5-Aza-CdR and 216 genes (15%) were significantly downregulated. Thus, the 5-Aza-CdR and Trichostatin A cause growth inhibition overall number of genes deregulated after treatment and cell death in immortalized and HRAS-transformed with 5-Aza-CdR is not related to the RAS status of the fibroblasts cells. Further analysis identified 16 genes significantly We investigated the effects of the demethylating agent lower expressed in untreated FE-8 cells compared to 5-Aza-CdR and the HDAC1 inhibitor trichostatin A untreated 208F cells but upregulated after treatment (TSA) on growth and viability of immortalized 208F rat with 5-Aza-CdR, exclusively in FE-8 cells (Table 2). fibroblasts and the HRAS-transformed derivative FE-8. Within this set of genes silenced by a methylation- A dose-dependent inhibition of cell growth was ob- dependent mechanism, syndecan 4, clusterin and Timp2 served for both cell lines after treatment with 0.1–10 mM had earlier been demonstrated to be suppressed in 5-Aza-CdR for 96 h and with 5–50 ng/ml TSA for 72 h a MEK/ERK-dependent manner in FE-8 cells (Zuber (Figure 1a). Concentrations above 1 mM 5-Aza-CdR et al., 2000) and revealed a similar 5-Aza-CdR-medi- and 25 ng/ml TSA caused growth inhibition owing to ated re-expression when analysed on Northern blots cytotoxic effects as indicated by an increase in cell death (Table 1). (Figure 1b). For further experiments, we selected We verified the expression of six genes that were concentrations between 0.1 and 0.3 mM 5-Aza-CdR and strongly induced after treatment with 5-Aza-CdR only between 5 and 25 ng/ml TSA allowing for DNA repli- in FE-8 cells. Clusterin, matrix metalloproteinase 2

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4892

Figure 1 (a) Growth properties of immortalized 208F ﬁbroblasts and HRAS(G12V)-transformed FE-8 cells either untreated or after treatment with 5-Aza-CdR or TSA. 5 Â 104 cells were seeded per 60 mm dish and treated the next day (arrow) for the following 4 days. Cell numbers were determined by counting with a haemacytometer. (b) In parallel, necrotic cells were determined by Trypan blue staining. (c) Western blot analysis showing TSA (25 ng/ml)-induced Histone H3(Lys9) hyperacetylation after treatment for 24 h and decrease of phospho-p44/42 ERK levels after treatment with 10 mM U0126. DNMT1 levels in 208F and FE-8 cells after treatment with 10 mM U0126 for 48 h were similar.

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4893 (Mmp2), peptidylpropyl isomerase C-associated protein TSA, 5 or 25 ng/ml TSA and 10 mM U0126 (Figure 2). (Ppicap), syndecan 4, Thbs1 and Timp2 expression was Comparison of gene expression after treatment with investigated after treatment of 208F and FE-8 cells 5-Aza-CdR or U0126 revealed particular differences, with 0.1 mM 5-Aza-CdR, 0.1 mM 5-Aza-CdR with 5 ng/ml depending on the RAS status. The low mRNA levels of clusterin, Mmp2, Ppicap, syndecan 4, Thbs1 and Timp2 in FE-8 cells were differentially increased after treatment Table 1 MEK/ERK target genes downregulated in FE-8 cells and with 5-Aza-CdR and U0126. In contrast, there were upregulated after treatment with 5-Aza-CdR only slight, if any, effects in 208F cells. Notably, Gene name Accession no. Species Method to verify clusterin, Mmp2 and Thbs1 were strongly suppressed in the effect of FE-8 cells, but were clearly induced after treatment with 5-Aza-CdR 5-Aza-CdR as well as U0126. The U0126-dependent upregulation of Mmp2 in FE-8 cells appeared hardly Signalling molecules syndecan 4 S61868 r N detectable on the Northern blot performed together with the other genes. Therefore, we performed an indepen- Nuclear proteins dent experiment with a longer exposure time showing Tead2 D50563 m N re-expression of Mmp2 48 h after the addition of the MEK1,2 inhibitor (Figure 2, small picture to the right). Protein processing Hsp90 X15183 h N Ppicap, syndecan 4 and Timp2 exhibited incomplete silencing in the FE-8 cells. Trichostatin A increased the Cytoskeletal components expression of clusterin, syndecan 4, Thbs1 and to a lesser P-cadherin (placental) X06340 m N extent Timp2 in FE-8 cells, whereas Ppicap was adver- Extracellular proteins sely downregulated and Mmp2 expression was not Timp2 NM_021989 r N affected (Figure 2). Combined treatment with 5-Aza- Thbs1 NM_011580 m N CdR and TSA induced no further changes in gene Mmp2 U65656 r N expression compared to treatment with 5-Aza-CdR Lamb1-1 M15525 m N alone. It is also remarkable that Ppicap and Timp2, Others both located on chromosome 10q32.3 within a distance Clusterin BC061543 r N, RT of 38 kb between the transcription start sites and no Ppicap BC081724 r N, RT other coding regions in between, were induced by 5-Aza- CdR and U0126, suggesting that the expression of genes Increased expression of individual genes was analysed after treatment within a certain chromosomal region may be subject to a with 0.1 mM 5-Aza-CdR for 96 h by Northern blot (N) or Reverse transcription PCR analysis (RT). The gene name is selected according common regulation. to its annotation in Entrez Gene http://www.ncbi.nlm.nih.gov/entrez/ query.fcgi?db ¼ gene. The accession number refers to the best match in blast analysis for the cDNA used as probe in Northern blot Clusterin and matrix metalloproteinase 2 regulatory hybridization with the species indicated: rat (r), mouse (m) and human regions are methylated in HRAS-transformed cells (h). RT–PCR was only performed in cases where the sequence of the cDNA clone used for Northern blot was unambiguously identiﬁed in We assumed that at least a subset of genes, upregulated the rat. in the HRAS(G12V)-transformed FE-8 cells after

Table 2 Relative gene expression alterations in FE-8 cells following 5-Aza-CdR treatment Affymetrix ID Gene name 5-Aza-CdR/control MEK/ERK target

S61868_g_at syndecan 4 3.4 Yes S61865_s_at syndecan 3.0 No J02722cds_at heme oxygenase 1 2.0 No rc_AI230354_at phosphatidate phosphohydrolase type 2 1.9 ND rc_H31535_at EST105653 Rattus norvegicus cDNA, 30 end 1.9 ND M64733mRNA_s_at clusterin 1.8 Yes rc_AI177503_at EST221135 Rattus norvegicus cDNA, 30 end 1.8 ND J04791_s_at ornithine decarboxylase 1.8 No S72594_s_at Timp2 1.6 Yes D13417_g_at transcription factor HES-1 1.5 ND AJ012603UTR#1_at TNF-alpha converting enzyme 1.5 ND X57405_g_at rat notch protein 1.5 ND rc_AA891441_s_at EST195244 Rattus norvegicus cDNA, 30 end 1.5 ND rc_AI008131_s_at EST202582 Rattus norvegicus cDNA 1.5 ND rc_AI639518_g_at Rattus norvegicus cDNA clone rx00570 3 1.5 ND AF022081_at ring ﬁnger protein 4 1.5 ND

5-Aza-CdR/control indicates the fold change induced in gene expression after treatment with 0.3 mM 5-Aza-CdR for 96 h with regard to normalized data from the Affymetrix Rat U34A oligonucleotide microarray analysis. To determine whether these genes are MEK/ERK targets, their expression was validated by RT–PCR or Northern blot after treatment with 10 mM U0126 for 48 h as indicated: MEK/ERK target gene (Yes), non- MEK/ERK target gene (No) or not tested (ND).

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4894 treatment with 5-Aza-CdR, are silenced owing to summarized in Table 3. A detailed presentation of the hypermethylation. To test this hypothesis, we compared methylation pattern of each clone analysed for clusterin, the methylation patterns within presumable regulatory Mmp2, Timp2 and syndecan 4 is depicted in the supple- regions of clusterin, Mmp2, Timp2 and syndecan 4 in mentary figures, which are labelled according to the 208F and FE-8 cells. Thbs1 and Ppicap could not be figures within this manuscript (Bisulphite_supplementary identified unambiguously within the rat genome at the data). time of analysis and were not further investigated. The A recent publication suggested that Timp2 is silenced distribution and properties of the CpG islands within by promoter hypermethylation in invasive cervical the promoter regions for all genes further analysed are carcinomas and in cervical cancer cell lines (Ivanova et al., 2004). However, the overall level of methylation within the CpG island of the Timp2 promoter was very low (Figure 3a). Also, no significant differences in the methylation pattern within the CpG island of the syndecan 4 promoter were observed between 208F and FE-8 cells (Figure 3b). In both genes, individual methylated CpGs were distributed randomly throughout the clones analysed (Supplementary Figure 3a and b). It remains an open question whether differences in methylation occur in other areas of the Timp2 and syndecan 4 CpG islands. Within the clusterin promoter, no CpG island matching the criteria of Takai and Jones (2002) was found. Nevertheless, we analysed the clusterin promoter as hypomethylation of the clusterin promoter was described to be associated with high gene expression in rat testis and epididymis (Rosemblit and Chen, 1994). Within the region comprising the entire exon 1, the overall level of methylation was low and no differences in the methylation frequency were observed between

Table 3 Distribution of CpG islands in individual genes regulated via 5-Aza-CdR Gene Name CpG Length CpGs CpG(o)/ Localisa- Figure 2 The MEK/ERK target genes clusterin, peptidylpropyl islands (bp) (%) CpG(e) tion isomerase C-associated protein (Ppicap), syndecan 4, matrix metalloproteinase 2 (Mmp2), Thbs1 and Timp2 are upregulated Clusterin — after treatment with 5-Aza-CdR or U0126 in FE-8 cells. Northern Mmp2 1 591 61.4 0.652 À344- blot analysis for each gene was performed after the treatment +247 indicated. The RNA probes were arranged as follows: 1: 208F, untreated, 96 h; 2: 208F, 0.1 mM 5-Aza-CdR, 96 h; 3: FE-8, Timp2 1 1198 68.2 0.681 À450- untreated, 96 h; 4: FE-8, 0.1 mM 5-Aza-CdR, 96 h; 5: 208F, +530 untreated, 120 h; 6: 208F, 0.1 mM 5-Aza-CdR, 96 h þ 5 ng/ml Ppicap — TSA, 24 h; 7: FE-8, untreated, 120 h; 8: FE-8, 0.1 mM 5-Aza-CdR, Syndecan 4 1 965 59.3 0.671 À650- 96 h þ 5 ng/ml TSA, 24 h; 9: 208F, untreated, 24 h; 10: 208F, 5 ng/ +310 ml TSA, 24 h; 11: 208F, 25 ng/ml TSA, 24 h; 12: FE-8, untreated, Thbs1 ND 24 h; 13: FE-8, 5 ng/ml TSA, 24 h; 14: FE-8, 25 ng/ml TSA, 24 h; 15: 208F, 0.1% dimethylsulphoxide (DMSO), 48 h; 16: 208F, 10 mM CpG islands were searched within the region of –2000 to +2000 at U0126, 48 h; 17: FE-8, 0.1% DMSO, 48 h; 18: FE-8, 10 mM U0126, NCBI Locus Link (http://www.ncbi.nlm.nih.gov/LocusLink/list.cgi) 48 h. The columns on top indicate the agents used for treatment and analysed with the CpG Island searcher (http://www.uscnorris. and whether RNA was derived from treated ( þ ) or from untreated com/cpgislands/cpg.cgi) according to the criteria of Takai and Jones (À) 208F and FE-8 cells. GAPDH hybridization served as a loading (2002): length >500 bp; CpG content >55%, observed (CpG)/ control. Picture to the right: Northern blot demonstrating the expected (CpG) >0.65. No CpG island found (À), no data (ND). induction of Mmp2 expression in FE-8 cells by U0126. Localisation is indicated relative to the transcriptional start site.

Figure 3 Comparative analysis of the methylation status of putative regulatory regions of (a) Timp2,(b) syndecan 4,(c, d) clusterin exon 1 and promoter, (e) clusterin CpG island and (f) matrix metalloproteinase 2 (Mmp2) in 208F and FE-8 cells by bisulphite sequencing reveals hypermethylation of a CpG island 15 kb upstream of clusterin and distinct CpG dinucleotides within the clusterin and Mmp2 promoter in FE-8 cells. Comparison of per cent methylation for each individual CpG (numbered in 50 to 30 direction) in untreated 208F (grey columns) and FE-8 (black columns) cells is shown. The number of clones analysed for each cell line is indicated (n). Schematical representation of all regions analysed by bisulphite sequencing depicts the CpG islands (grey boxes), exons (open boxes) and the transcription start site (arrow). The localization of the sequences amplified is shown with enlargement depicting individual CpGs (bubbles) with primers used (arrows). Nucleotides not reliably identified by sequencing are marked (X). The first CpG analysed of each amplified sequence is numbered.

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4895 208F and FE-8 cells (Figure 3c). The picture changed methylation frequency of four out of six CpGs in FE-8 upon investigation of the region ranging from À560 cells was signiﬁcantly higher as compared to 208F cells to À314 relative to the transcriptional start site. The (Figure 3d). Overall, the methylation frequency for all

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4896 six CpGs increased from 41% in 208F cells to 61% in cells, we investigated clusterin expression and clusterin FE-8 cells. Whereas only three of 20 clones (15%) promoter methylation in another HRAS-transformed analysed in 208F cells were at least methylated at five cell line (FYJ10; Schwarte-Waldhoff et al., 1994) and in of six CpGs, this was true for 11 of 22 clones (50%) two phenotypically revertant cell lines (F4 and F9; analysed in FE-8 cells. Strikingly, three CpGs hyper- Scha¨ fer et al., 1998). The revertants maintain a normal methylated in FE-8 cells are located in close proximity phenotype similar to their non-transformed precursors within a range of 11 nucleotides (region À385 to À376) and exhibit low levels of ERK activation despite the in the clusterin promoter (Figure 3d). continuous expression of the oncogene (Raudies et al., The only CpG island localized 50 of the clusterin 2005). As in FE-8 cells, clusterin is downregulated in promoter is found 14.5 kb upstream, with no coding the FYJ10 cells and upregulated after treatment with sequences in between. This CpG island was hypermethyl- 5-Aza-CdR, U0126 and TSA, whereas the higher ated in FE-8 cells compared to 208F cells within 28 expression levels of clusterin in the revertants F4 and of 30 CpGs investigated (Figure 3e). Overall, 66% of all F9 cells are only slightly upregulated after TSA treat- CpGs analysed in this specific location were methylated ment (Figure 5a). The reduction of clusterin expression in FE-8 cells, whereas only 22% were methylated in correlates with methylation throughout the complete 208F cells and at least 12 of 30 CpGs analysed were clusterin promoter in FYJ10 cells (overall methylation methylated in each clone of the FE-8 cells. rate of 40%), whereas F4 and F9 cells show no Within the Mmp2 gene, we analysed the methylation methylation at all, except at the first CpG (Figure 5b). pattern of a fragment covering the sequence from À327 to À89 relative to the transcriptional start site harbouring a bona fide CpG island. The total methylation rate Long-term activation of HRAS establishes methylation- increased from 31% in 208F cells to 80% in FE-8 cells mediated silencing, whereas short-term activation (Figure 3f), where each clone was at least methylated at suppresses clusterin by histone deacetylation five of 11 positions. To analyse the impact of RAS signalling on Mmp2 Next, we investigated to which extent the hyper- and clusterin expression in a time-resolved manner, methylation observed at the presumed regulatory we analysed the regulation of these genes in IR-4 regions for clusterin and Mmp2 expression was altered cells, which harbour an isopropyl-b-D-thiogalactoside after treatment with 0.1 mM 5-Aza-CdR in FE-8 cells. (IPTG)-inducible HRAS (Liu et al., 1992; Sers et al., The clusterin promoter was completely demethylated 2002). IR-4 cells were derived from 208F immortalized after treatment with 5-Aza-CdR; only at the first 50 rat fibroblast and display a similar immortalized CpG, the methylation frequency was not reduced phenotype. Upon induction of the HRAS oncogene, (Figure 4a). In contrast, treatment with 0.1 mM 5-Aza- IR-4 cells transform and within 4 days adopt a pheno- CdR induced unequally distributed changes with regard type similar to FE-8 cells (Sers et al., 2002). Northern to the methylation pattern of the clusterin CpG island blot analysis revealed that 6–12 h after induction of (Figure 4b). This was a surprising finding, which cannot HRAS in IR-4 cells, clusterin and Mmp2 mRNA levels be explained at the moment. Recently, a differential begin to decline and are further downregulated until regulation of DNMT1, 3a and 3b in response to 5-Aza- 72 h post induction (Figure 6a, upper panel). Western CdR treatment was reported (Schneider-Stock et al., blot analysis demonstrated that RAS protein levels 2005), whether this also induces regional differences in in IR-4 cells increase slowly up to 6 h and are strongly DNA methylation has to be investigated in more detail. enhanced from 12 to 72 h (Figure 6a, lower panel). At the Mmp2 promoter region, treatment with 5-Aza- Then, we treated IR-4 cells simultaneously with IPTG CdR induced an equally distributed, but incomplete to induce RAS and with either TSA (25 ng/ml) or 5-Aza- demethylation in FE-8 cells (Figure 4c). Interestingly, CdR (0.1 mM; Figure 6b). Demethylation or histone within the region harbouring the first five CpGs, the hyperacetylation disturbed the previously described overall methylation frequency decreased from 89% in morphological transformation of IR-4 cells within 72 h the untreated cells to 63% in the 5-Aza-CdR-treated after HRAS induction (Sers et al., 2002). Further- FE8 cells. In nine of 32 5-Aza-CdR-treated clones, less more, TSA-induced histone hyperacetylation increased than three CpGs were methylated, whereas methylation clusterin expression already 30 min after treatment and in the others was similiar to the untreated. Probably, a overrode the HRAS-induced repression of clusterin subpopulation of 5-Aza-CdR-treated cells is responsible observed between 24 and 72 h (Figure 6b, right panel), for the increase of Mmp2 expression. In addition, we whereas the HRAS-induced repression of Mmp2 was find that silencing of clusterin and Mmp2 is DNMT1 not affected by treatment with TSA. These observations dependent, as inhibition of DNMT1 for 72 h by siRNA are in agreement with the effects observed in FE-8 cells resulted in upregulation of clusterin and Mmp2 expres- and point to a clear difference in the mechanisms of sion (Figure 4d). initial Mmp2 and clusterin repression in response to As the clusterin promoter is largely demethylated HRAS. Treatment with 5-Aza-CdR neither affected after treatment with 5-Aza-CdR, we supposed that the downregulation of clusterin nor Mmp2 in IR-4 cells methylation pattern of the clusterin promoter is crucial within 72 h (Figure 6b, left panel). for clusterin expression. To rule out that the highlighted To determine whether the HRAS-mediated hyper- correlation between HRAS-mediated ERK activation methylation of the clusterin promoter that we observed and DNA hypermethylation occurred randomly in FE-8 in FE-8 and FYJ10 cells is an immediate or delayed

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4897

Figure 5 Expression and methylatin of clusterin in the HRAS- transformed FYJ10 cells and in the revertant cell lines F4 and F9. (a) Northern blot analysis for clusterin was performed after the treatment indicated. The RNA probes were arranged as follows (upper panel) 1: FYJ10, untreated, 96 h; 2: FYJ10, 0.1 mM 5-Aza- CdR, 96 h; 3: F4, untreated, 96 h; 4: F4, 0.1 mM 5-Aza-CdR, 96 h; 5: F9, untreated, 96 h; 6: F9, 0.1 mM 5-Aza-CdR, 96 h; 7: FYJ10, 0.1% DMSO, 48 h; 8: FYJ10, 10 mM U0126, 48 h; 9: F4, 0.1% DMSO, 48 h; 10: F4, 10 mM U0126, 48 h; 11: F9, 0.1% DMSO, 48 h; 12: F9, 10 mM U0126, 48 h (lower panel), 1: FYJ10, untreated, 24 h; 2: FYJ10, 5 ng/ml TSA, 24 h; 3: FYJ10, 25 ng/ml TSA, 24 h; 4: F4, untreated, 24 h; 5: F4, 5 ng/ml TSA, 24 h; 6: F4, 25 ng/ml TSA, 24 h; 7: F9, untreated, 24 h; 8: F9, 5 ng/ml TSA, 24 h; 9: F9, 25 ng/ml TSA, 24 h. The columns on top indicate the agents used for treatment and whether RNA was derived from treated ( þ )or from the untreated cells (À). GAPDH hybridization served as a loading control. Comparative analysis of the methylation status of the (b) clusterin promoter in FYJ10, F4 and F9 cells by bisulphite sequencing, is shown. Comparison of per cent methylation for each individual CpG (numbered in 50 to 30 direction) in untreated FYJ10 (black columns), F4 (grey columns) and F9 (white columns) cells is indicated. The number of clones analysed for each cell line is indicated (n).

Figure 4 Comparative analysis of the methylation status of the (a) clusterin promoter, (b) clusterin CpG island and (c) matrix metalloproteinase 2 (Mmp2) promoter by bisulphite sequencing in untreated cells were observed, demonstrating that initial FE-8 cells treated with 5-Aza-CdR reveals signiﬁcant alterations in repression of clusterin expression is directly mediated by the methylation pattern. Comparison of percent methylation for each individual CpG (numbered in 50 to 30 direction) in untreated RAS activation, yet independent of promoter hyper- FE-8 (black columns) and 5-Aza-CdR-treated FE-8 (grey columns) methylation. However, 240 h after induction, the clus- cells is shown. The number of clones analyzed for each cell line is terin promoter was methylated at an overall rate of 35% indicated (n). (d) Inhibition of DNMT1 transcription in FE-8 cells (Figure 6c). Among the 16 clones sequenced 10 days by siRNA results in upregulation of clusterin and Mmp2 after RAS induction, 50% showed methylation only transcription. 1: FE-8, mock treated; 2: FE-8, siRNA-EGFP 0 (10 nM); 3: FE-8, siRNA-DNMT1 (10 nM). GAPDH hybridization at the most 5 CpG; all other CpGs were unmethylated. served as a loading control. The remaining clones exhibited a random distribution of CpG methylation (Supplementary Figure 6c). consequence of RAS activation, we treated IR-4 cells Whether this indicates an allele-speciﬁc methylation of with IPTG up to 240 h. After treatment for 48 h, no the clusterin promoter by oncogenic RAS remains to be differences in the methylation pattern compared to the investigated.

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4898

Figure 7 High expression of clusterin in testis correlates with low Figure 6 Induction of HRAS in IR-4 cells reveals hypermethyla- clusterin promoter methylation and clusterin CpG island methylation of the clusterin promoter after 240 h, whereas early transcription. (a) Northern blot analysis for clusterin in several normal rat tional repression is methylation independent, but histone tissues. The RNA probes were arranged as follows: 1: lung; 2: acetylation dependent. Kinetics of HRAS-mediated downregula- brain; 3: brain (cerebellum); 4: heart; 5: breast; 6: small intestine (I); tion of clusterin and matrix metalloproteinase 2 (Mmp2) expression 7: skin; 8: stomach; 9: spleen; 10: liver; 11: kidney; 12: spleen; 13: after HRAS induction is shown (a, upper panel). Time points after muscle; 14: colon (C); 15: placenta; 16: liver; 17: testis (T). 18S addition of isopropyl-b-D-thiogalactoside (IPTG) relative to rRNA hybridization served as a loading control. Comparative untreated control cells (À) are indicated. Western blot analysis of analysis of the methylation status within the (b) clusterin promoter RAS protein levels in IR-4 cells up to 72 h postinduction. Actin was and (c) clusterin CpG island in rat tissue from colon, small intestine used as a loading control (a, lower panel). (b) Kinetics of clusterin and testis is shown. Comparison of per cent methylation for each and Mmp2 expression after simultaneous addition of IPTG and individual CpG (numbered in 50 to 30direction) in tissue of colon 0.1 mM 5-Aza-CdR (left panel) or IPTG and 25 ng/ml TSA (right (black columns), small intestine (grey columns) and testis (white panel). Comparative analysis of the methylation status of the (c) columns). The number of clones analysed for each cell line is clusterin promoter 48 and 240 h after HRAS induction in IR-4 cells indicated (n). by bisulphite sequencing. Comparison of per cent methylation for each individual CpG (numbered in 50 to 30 direction) in untreated IR-4 cells after 48 h (white, shaded columns) and 240 h (white columns) and IPTG-treated IR-4 cells after 48 h (grey columns) and methylation of the clusterin promoter was 20% in colon 240 h (black columns). The number of clones analysed for each cell and increased to a rate of 29% in small intestine. The line is indicated (n). overall methylation frequencies at the clusterin CpG island were at least twofold in colon and small intestine as compared to testis. These observations strengthen the clusterin Expression of clusterin in normal rat tissues is inversely hypothesis that silencing of is accompanied by clusterin correlated with the methylation rate of the clusterin hypermethylation within the promoter at an promoter and the clusterin CpG island overall rate of 30%, similar to the rates observed in FYJ10 and IR-4 cells transformed by HRAS. In order to determine whether methylation of the clusterin promoter and of the clusterin CpG island also correlates with clusterin expression in normal rat tissues, we analysed clusterin expression in 17 tissue Discussion probes (Figure 7a). Then, we selected three tissues exhibiting high (testis), intermediate (colon) and low DNA methylation is essential for the control of gene (small intestine) clusterin expression for methylation expression. In mammals, methylation serves several studies. The clusterin promoter and the clusterin CpG important biological purposes such as imprinting, island were nearly unmethylated in testis as expected control of gene expression during development and from the high level of clusterin expression. The overall X-chromosome inactivation (Avner and Heard, 2001;

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4899 Reik and Walter, 2001; Reik et al., 2001). In addition, prominent in FE-8 cells than within the promoter gene silencing by methylation is of particular impor- region. Furthermore, analysis of clusterin expression tance in tumours, where DNA is often globally in rat tissues revealed very low methylation of this hypomethylated, but locally hypermethylated in CpG- CpG island in testis, supporting a role for this region rich promoter regions of tumour suppressor genes such in the control of clusterin expression. A splice variant of as p16INK4a, p15INK4b, p14ARF, p73, APC,orBRCA1 rat clusterin (Ensembl transcript ID: ENSRNOT000000 (Esteller et al., 2001; Esteller, 2003). 29806) with an alternative exon 1 is annotated approxi- The sequence of events resulting in gene-specific mately 1.5 kb downstream of the amplified region. suppression owing to DNA methylation is still contro- Methylation of the CpG island might act as a switch versial (Esteller, 2002; Villar-Garea and Esteller, 2004). for the induction of the alternative transcript. However, A key epigenetic modification is methylation at histone neither Northern blot analysis nor Reverse transcrip- H3 Lys9 (Lachner and Jenuwein, 2002) and in mam- tion–PCR using primers specific for the predicted exon 1 malian cells, DNMT3a was described in a complex with gave any hint for the existence of the alternative the methyltransferase SUV39H1. This demonstrates a transcript (data not shown). Studies on the imprinted connection between the enzymes responsible for DNA Igf2-H19-region in the mouse have already shown that methylation and histone methylation (Fuks et al., long-range chromatin interactions contribute to the 2003a). Methyl-DNA-binding proteins, for example control of gene expression (Lopes et al., 2003). Thus, MeCP2, also associate with histone methyltransferases we assume that this CpG island might regulate trans- in vivo and recruit histone deacetylases to reinforce a cription of clusterin over a range of 14.5 kb in conjunc- repressive chromatin state (Fuks et al., 2003b). Whether tion with the regulatory regions directly upstream of the DNA methylation is a prerequisite for histone methyla- transcriptional start. tion or vice versa is currently unknown. In addition, a In general, the overall methylation rates for clusterin direct recruitment of DNMTs by transcription factors increased by a rate of 20–40% in HRAS-transformed such as Myc (Brenner et al., 2005) or PML-RAR (Di cell lines as compared to immortalized cell lines result- Croce et al., 2002) can establish local changes in ing in a strong downregulation. Similar effects were chromatin structure. Whereas the aforementioned observed by Di Croce et al. (2002) for the RARb2 mechanisms rather direct DNA methylating or histone promoter, for example showing that RARb2 expression methylating enzymes to specific regulatory regions, a was silenced in NB4 cells at an overall methylation rate third possibility to repress gene expression is the of 47%. After treatment with retinoid acid, the overall transcription factor-associated recruitment of histone methylation rate decreased to 23%, which resulted in a deacetylases (Laduron et al., 2004). This latter mechan- re-expression of the RARb2 gene. This indicates that ism might govern the initial suppression of the clusterin even moderate alterations in methylation might mediate gene in response to HRAS. Analysis of clusterin profound changes in gene expression, an idea that was inhibition upon HRAS induction in IR-4 cells indicated first suggested several years ago by Hsieh (1994). that histone hypoacetylation is the primary event Furthermore, alterations induced by continuous cultur- mediating inhibition of clusterin within the first 72 h. ing might induce subtle changes in expression and This primary suppression is likely to be mediated via the methylation of individual genes, as obvious from the MEK/ERK signalling pathway, followed by promoter comparison of 208F, F9 and IR-4 cells. Therefore, the methylation 240 h after RAS induction. Thus, a relevance of a certain methylation rate for the expres- potential interaction between HDAC1 and DNMT1 sion of an individual gene has to be investigated in (Fuks et al., 2000) could then result in a further several different sources. repression of gene expression by subsequent DNA In contrast to clusterin, Mmp2 could be subject to a methylation. A region from À426 to À311 is necessary direct, methylation- and histone deacetylation-indepen- for maximal clusterin promoter activity and deletion of dent but MEK/ERK-dependent inhibition first, fol- this fragment decreased transcription of a reporter gene lowed by methylation of the CpG island. Whereas by up to 70% in rat Sertoli cells (Lymar et al., 2000). clusterin was suppressed in response to activated HRAS Three CpGs detected as hypermethylated by bisulphite in different cell lines, Mmp2 exhibited a heterogeneous sequencing of the clusterin promoter are located at regulation via both HRAS-dependent and HRAS- positions À385, À379 and À376 in close proximity to independent effects (data not shown). Therefore, it is each other. The cytosines hypermethylated (underlined) currently unclear whether the observations made in reside within the sequence 50-AACGTGGGCGTCGGG FE-8 and IR-4 cells are of general importance for Mmp2 CGGTG-30 harbouring one Egr1 and two Sp1-binding regulation. sites. Most interestingly, induction of clusterin expres- Silencing of tumour suppressor genes by methylation sion via MEK/ERK following ionizing radiation was clearly confers a selective advantage to the transformed recently shown to be dependent on the binding of Egr-1 cells, for example owing to the abrogation of cell cycle to the promoter (Criswell et al., 2005). These findings control; so what might be the role of clusterin and Mmp2 suggest that binding or activation of Egr1 and/or Sp1 silencing in tumorigenesis? Clusterin is constitutively are affected upon hypermethylation within the above- expressed in almost all mammalian tissues and is mentioned area of the clusterin promoter. implicated in diverse physiological processes, including Hypermethylation within the CpG island about cell adhesion, tissue remodelling, membrane recycling, 14.5 kb upstream of the clusterin gene was even more cell–cell interaction and apoptosis (reviewed by Jones

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4900 and Jomary, 2002). Opposing biological functions In summary, we investigated the impact of RAS of clusterin have been described in distinct tumour oncogene activation on the methylation of candidate types, which seem to be determined by different iso- genes. We find that activation of an oncogenic signalling forms with cell type-specific modifications such as pathway can induce epigenetic modification to suppress glycosylation and by alternative splicing of the human gene expression directly as shown for clusterin and clusterin gene. Recent work suggested that only the indirectly as suggested for Timp2 and syndecan 4. Future secreted form of clusterin correlates with tumour studies will aim to investigate the impact of distinct progression, whereas the proapoptotic nuclear form, RAS-dependent signalling pathways on methylation detected in normal tissue, decreases in expression of individual regulatory regions and to unravel the in tumour cells (Leskov et al., 2003; Pucci et al., 2004). sequence of molecular events resulting in a persistent Several findings support a role of clusterin as a potential gene suppression following RAS activation. tumour suppressor gene: clusterin induces p21-dependent apoptosis of colon and prostate carcinoma cells (Chen et al., 2004; Scaltriti et al., 2004b). It is Materials and methods downregulated in low- and high-grade human prostate cancer (Scaltriti et al., 2004a) and in testicular germ cell Cell culture conditions tumours (Behrens et al., 2001). Upregulation of clusterin Immortal rat 208F fibroblasts, the HRAS(G12V)-transformed leads to suppression of tumorigenesis in concert with derivatives FE-8 (Griegel et al., 1986) and FYJ10 (Schwarte- the downregulation of HRAS (Kyprianou et al., 1991). Waldhoff et al., 1994), and the revertant cell clones F4 and F9 In RAS-transformed cell lines, clusterin expression (Scha¨ fer et al., 1988) were cultured in Dulbecco’s modified is consistently downregulated (Klock et al., 1998; Eagle’s medium supplemented with 10% fetal calf serum, 2% Tchernitsa et al., 2004). Nevertheless, further experi- penicillin–streptomycin, and 2 mML-glutamine. IR-cells (clone IR-4) harbour an IPTG-inducible HRAS oncogene and were ments are required to investigate the contribution of described earlier (Sers et al., 2002). Expression of HRAS was clusterin suppression to RAS transformation and to induced by the addition of 20 mM isopropyl-1-thio-b-D- investigate whether clusterin silencing is caused by galactoside. Stock solutions of 5-Aza-CdR (Sigma, St Louis, promoter hypermethylation in tumour samples with MO, USA) were prepared by dissolution at 10 mM in mutated RAS. phosphate-buffered saline (pH 6.5). Trichostatin A (Sigma) Mmp2 belongs to the group of endopeptidases usually was dissolved in ethanol (1 mg/ml) and U0126 (Promega, upregulated in cancer and their expression is associated Madison, WI, USA) in dimethylsulphoxide (DMSO) (10 mM). with poor survival. Mice lacking Mmp2 develop fewer Immediately before use, stock solutions were further diluted in medium. For demethylation experiments, cells were plated at cancers than wild-type mice (Itoh et al., 1998), but there 5 is no linear relationship between Mmp2 expression level a density of 1 Â 10 cells/100 mm dish and treated the next day. Medium and 5-Aza-CdR were changed every 24 h. For and the ability of cancer cell lines to invade matrigel treatment with the HDAC inhibitor TSA and the MEK1, (Deryugina et al., 1997). Mmp2 is also involved 2 inhibitor U0126, cells were plated at a density of 3 Â 105 cells/ in tumour angiogenesis; however, a selective role of 100 mm dish and treated the next day. Concentrations for different Mmps was demonstrated in the RIP-Tag2 treatments and time intervals were as indicated. Solvents were insulinoma model (Bergers et al., 2000). This suggests added to the controls in the appropriate concentration. that the Mmp pattern of a given tumour cell is tightly Cell growth was determined by seeding 5 Â 104 cells into a controlled to provide the cell with angiogenic and 60 mm dish followed by treatment the next day. At the time invasive potential. Remarkably, Mmp2, Thbs1 and points indicated, cells were trypsinized, cell numbers deter- Timp2, all belonging to the same transcriptional mined using a haemacytometer and the fraction of living cells module, functionally interact in tumorigenesis. Mmp2 was calculated by Trypan blue exclusion (0.2% v/v). is activated at the cell surface through a unique multistep pathway that involves Timp2 (Wang et al., Western blot analyses 2000; Deryugina et al., 2001), and also THBS1, shown Cells were solubilized in lysis buffer (20 mM Tris-HCl pH 8.0, to be silenced by methylation in human cancers 100 mM NaCl, 1% DOC, 1% NP-40, 0.1% SDS, complete protease inhibitor mix (Roche Diagnostics, Mannheim, (Li et al., 1999), has been reported to increase Germany)) and 20 mg of the whole cell extracts was separated MMP2 activation (Taraboletti et al., 2000). In contrast by SDS–polyacrylamide gel electrophoresis. After semi-dry to Mmp2, two related genes Mmp3 and Mmp10 are blotting (BioRad, Hercules, CA, USA) to PVDF membranes upregulated in FE-8 and IR-4 cells and downregulated (Hybond P, Amersham Pharmaceuticals, Buckinghamshire, after inhibition of MEK1 by PD98059 (Zuber et al., UK), the membranes were blocked for 1 h in TBST (10 mM 2000 and data not shown). It is therefore likely that the Tris pH 8.0, 150 mM NaCl, 0.05% Tween 20) with 5% non-fat MEK/ERK signalling pathway is responsible for a dry milk and incubated with primary antibodies against RAS coordinated transcriptional regulation of Mmps and this (Transduction Laboratories, Los Angeles, CA, USA), acetyl- might vary between individual cells such as FE-8 and histone H3 (Lys9) (Cell Signaling), phospho-p44/42 MAPK FYJ10. The fact that no differences in the methylation (Thr202/Tyr204) (New England Biolabs, Ipswich, MA, USA), anti-DNMT1 (Alexis Biochemicals, San Diego, CA, USA), pattern of the Timp2 region were observed is in line with anti-a-tubulin (Chemicon, Temecula, CA, USA) and anti-b- the observations showing that Timp2 re-expression may actin (Chemicon). Membranes were washed and incubated be induced after treatment with the 5-Aza-CdR via an with peroxidase-conjugated secondary antibodies. Signals were indirect mechanism involving the activation of NF-Y detected by chemiluminescence reaction (ECL; Amersham (Cappabianca et al., 2003). Pharmaceuticals) according to the manufacturer’s instructions.

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4901 Oligonucleotide microarray hybridizations about a microarray experiment (MIAME) and microarray Total RNA was prepared from cultured cells using the TRIzol gene expression database (MGED) group standards (Brazma Reagent (Invitrogen) according to the manufacturer’s instruc- et al., 2001). The gene expression Omnibus accession number is tions and dissolved in nuclease-free water. After isolation, GSE4427. RNA used for microarray hybridizations was purified by the RNeasy Mini Kit (Qiagen, Hilgen, Germany) and the RNA Northern blot analysis quality was determined on a LabChip Bioanalyzer (Agilent Ten micrograms of RNA were denatured for 5 min at 951C Technologies, Palo Alto, CA, USA). cDNA was synthesized and run on a 0.8% agarose/0.6 M formaldehyde gel for 4 h. from 5 mg of total RNA and synthesis of biotin-labelled cRNA The RNA was transferred to a nylon membrane (Nytran N, was performed using the BioArray High Yield RNA tran- Schleicher & Schuell) and crosslinked by UV light. Membranes scription kit (Enzo Diagnostics, Farimigdale, NY, USA). were pre-hybridized for 1 h at 661C in hybridization buffer cRNA was hybridized for 16 h at 451C to the Rat Genome (ExpressHyb, Clontech) with 100 mg/ml yeast tRNA. Twenty- oligonucleotide microarray U34A (Affymetrix, Santa Clara, five nanograms of the cDNA probe were radiolabelled with CA, USA). Subsequent washing and staining of the arrays dCTP[a-32P] (ICN) by random priming, and between 0.5 and was performed using the GeneChip fluidics station protocol 1 Â 106 cpm of the labelled probe was added per ml hybridiza- EukGE-WS2. Following washing and staining, the arrays were tion buffer and hybridized overnight at 661C. Membranes were scanned using an Agilent G2500A GeneArray Scanner washed to a stringency of 2 Â SSC, 0.1% SDS at 421C, exposed controlled by Microarray Suite 5.0 software (Affymetrix). to X-ray films and stored at –801C until detection. To verify The Data Mining Tool 3.0 (Affymetrix) and GeneSpring equal loading and integrity of RNA, all gels were stained with software package 6.1 (Silicon Genetics, Redwood City, CA, ethidium bromide. mRNA levels were normalized with USA) were used to average results from different replicates glyceraldehyde 3-phosphate dehydrogenase (GAPDH)or18S and to perform statistical analysis. Experiments were per- rRNA. formed in duplicate for each treatment and cell line, and only genes with a maximal deviation of fold change of o1.5 between the replicates and with a statistically significant t-test siRNA transfection were further analysed. Genes with signal intensities below the The siRNA duplexes for targeted silencing of DNMT1 background threshold in either control or 5-Aza-CdR-treated were prepared by using the Silencer siRNA Construction Kit cells were regarded as background and excluded from analysis. (Ambion). The target DNMT1-mRNA is encoded from exon All raw fluorescence intensity data and microarray image 37 of the DNMT1 gene. The sequence of the siRNA is files are deposited within the public repository for micro- 50-UGUGCUCGCUCCCAAGGCUUU-30 (sense strand) and array-based gene expression data ArrayExpress (http://www. 30-UUACACGAGCGAGGGUUCCGA-50 (antisense strand). ncbi.nlm.nih.gov/geo/), complying with minimum information A Blast search of the rat genome database was carried out to

Table 4 Primer pairs used for PCR ampliﬁcation before bisulphite sequencing

Sequence TA (1C) PCR product Region ampliﬁed size (bp) clusterin CpG island 60 335 (À15 138- À14 804)a Forward 50-GGTTTTGTGTTTTTGTTTAGATAG-30 Reverse 50-AAAATCCAAAATCCTAATCCC-30 clusterin promoter/exon 1 53 221 À59-+162 Forward 50-TTTTAAAGGGTGTATTTGAGTAGAG-30 Reverse 50- AAACTCTAACCACCTCCCACTTAC -30 clusterin promoter 53 247 À560-À314 Forward 50-TTTTTTTATTTGTTTGTTAAGGGAT-30 Reverse 50-AAAAATCCACACTCTACTCTAAACC-30

Mmp2 promoter CpG island 53 238 À327-À90 Forward 50-TGAGATTTTTTTATTTAATTGTTTTAGTTT-30 Reverse 50-ATACTACCAACAATCTTCCAACCC-30

Timp2 promoter CpG island 53 258 À193-+66 Forward 50-GGAGAAGAAAGAGGTTAGGGGG-30 Reverse 50-CAACCTATACCAAACTAAACACAC-30 syndecan 4 promoter CpG island 58 319 À318-+1 Forward 50-TAATTCCAACAACACAAACCAATAA-30 Reverse 50-TTAGTTATAAAGTGGGGTTGGGT-30 syndecan 4 exon 1 CpG island 55 269 À23-+246 Forward 50-CCAAAACAAAATTTCCCCTTAAAA-30 Reverse 50-TATTGGTTTGTGTTGTTGGAATTA-30

Sequences of the primers used for amplification of bisulphite-modified DNA are listed with the annealing temperature (TA) used. In addition, the length of the PCR product and the location of the region amplified relative to the transcription start site (+1) are indicated. aThe region amplified is located about 15 kb upstream of the transcription start site.

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4902 ensure that the sequence would not target other gene tran- and cooled to 41C. Primers (Metabion, Martinsried, Germany) scripts. Briefly, 1 Â 105 cells were plated and grown in six-well for amplification of bisulphate-modified DNA were designed plates 24 h before transfection, washed with 1 Â PBS, and with the assistance of the Oligo3 software and MethPrimer 800 ml OptiMEM (Gibco) was added per well. For each well, (Li and Dahiya, 2002). PCR products were resolved on a 10 nM siRNA was mixed with 3 ml of Oligofectamine (Invitro- 1–2% agarose gel, the specific band was excised and the PCR gen, Carlsbad, CA, USA) in 40 ml of OptiMEM. The mixtures product was purified using the QIAquick Gel Extraction were incubated for 20 min at room temperature and then system (Qiagen) or Jetsorb (Genomed). The purified PCR added to cells. Serum was added 4 h later to a final products were cloned by the TOPO TA Cloning Kit (pCR2.1- concentration of 10%. Transfection was repeated 24 h later, TOPO Vector; Invitrogen, CA, USA) or the pGem-T Vector and 72 h after the second transfection, total RNA was System I (Promega), and at least 12 clones were sequenced collected for Northern blot analysis. using the Sequiterm sequencing kit (Epicentre, Madison, WI, USA) by sequencing on a CEQ8000 (Beckman-Coulter) in Bisulphite sequencing each experiment or by using a commercial sequencing service DNA was isolated by phenol/chloroform extraction. Follow- (AGOWA, Berlin, Germany). ing digestion of 500 ng DNA with EcoRV overnight, probes were denatured for 10 min at 951C, cooled on ice, and finally Abbreviations 50 ml of 2% LMP agarose (Sigma) prewarmed to 501C was added to keep the DNA in a denatured state. Then 750 mlof 5-Aza-CdR, 5-Aza-2-deoxycytidine; Thbs1, thrombospondin a 2.7 M sodium bisulphite solution (pH 5; Merck, Darmstadt, 1; Mmp2, matrix metalloproteinase 2; Ppicap, peptidylpropyl Germany) and 5 mM hydroquinone (Sigma) were overlaid with isomerase C-associated protein; Thbs1, thrombospondin 1; 750 ml of heavy mineral oil (Aldrich) and cooled on ice. A Timp2, tissue inhibitor of metalloproteinase 2. measure of 10 ml of the sample/agarose mixture was pipetted into the cold mineral oil layer to form agarose beads (Olek et al., 1996). The solution was incubated for 3.5 h at 501C. Acknowledgements After washing three times (1 ml) for 15 min with 1 Â TE Buffer (pH 8.0), the conversion reaction was completed by de-sulpho- We thank Conny Giesseler, Katrin Kremp, Kerstin Lehmann, nating twice for 15 min in 0.2 M NaOH. After repeated Kristin Lucht and Christina Daniel for technical assistance washing, 2–5 ml of the bisulphate-modified DNA was used and Karsten Juerchott for helpful discussion. We are grateful as a template in each PCR reaction performed with the to Pablo Steinberg (German Institute of Human Nutrition, HotStartTaq Master Mix Kit (Qiagen, Hilden, Germany) Potsdam-Rehbru¨ cke), who provided us with samples from rat using the following conditions: 951C for 15 min, 40 cycles of colon and small intestine. This work was supported by grants 1 min denaturation at 951C, 1 min annealing at primer-specific from the Deutsche Forschungsgemeinschaft (SFB618 grant to temperatures (Table 4) and 1 min of extension at 721C. The RS and HpH) and the Stifterverband fu¨ r die Deutsche PCR reaction was terminated with 10 min of extension at 721C Wissenschaft (grant to CS, RS, JW and HpH).

References

Avner P, Heard E. (2001). Nat Rev Genet 2: 59–67. Di Croce L, Raker VA, Corsaro M, Fazi F, Fanelli M, Faretta Baker KM, Wei G, Schaffner AE, Ostrowski MC. (2003). M et al. (2002). Science 295: 1079–1082. J Biol Chem 278: 17876–17884. Downward J. (1998). Curr Op Genet Dev 8: 49–54. Bakin AV, Curran T. (1999). Science 283: 387–390. Esteller M. (2002). Oncogene 21: 5427–5440. Behrens P, Jeske W, Wernert N, Wellmann A. (2001). Esteller M. (2003). Lancet Oncol 4: 351–358. Pathobiology 69: 19–23. Esteller M, Corn PG, Baylin SB, Herman JG. (2001). Cancer Bergers G, Brekken R, McMahon G, Vu TH, Itoh T, Tamaki Res 61: 3225–3229. K et al. (2000). Nat Cell Biol 2: 737–744. Fensterer H, Giehl K, Buchholz M, Ellenrieder V, Buck A, Bos JL. (1989). Cancer Res 49: 4682–4689. Kestler HA et al. (2004). Genes Chromosomes Cancer 39: Brazma A, Hingamp P, Quackenbush J, Sherlock G, Spellman 224–235. P, Stoeckert C et al. (2001). Nat Genet 29: 365–371. Fuks F, Burgers WA, Brehm A, Hughes-Davies L, Kouzarides Brem R, Certa U, Neeb M, Nair AP, Moroni C. (2001). T. (2000). Nat Genet 24: 88–91. Oncogene 20: 2854–2858. Fuks F, Hurd PJ, Deplus R, Kouzarides T. (2003a). Nucleic Brenner C, Deplus R, Didelot C, Loriot A, Vire E, De Smet C Acids Res 31: 2305–2312. et al. (2005). EMBO J 24: 336–346. Fuks F, Hurd PJ, Wolf D, Nan X, Bird AP, Kouzarides T. Campbell SL, Khosravi-Far R, Rossman KL, Clark GJ, Der (2003b). J Biol Chem 278: 4035–4040. CJ. (1998). Oncogene 17: 1395–1413. Gadal F, Bozic C, Pillot-Brochet C, Malinge S, Wagner S, Cappabianca L, Farina AR, Tacconelli A, Mantovani R, Le Cam A et al. (2003). Nucleic Acids Res 31: 5789–5804. Gulino A, Mackay AR. (2003). Exp Cell Res 286: Griegel S, Traub O, Willecke K, Scha¨ fer R. (1986). Int J 209–218. Cancer 38: 697–705. Chen T, Turner J, McCarthy S, Scaltriti M, Bettuzzi S, Hsieh CL. (1994). Mol Cell Biol 14: 5487–5494. Yeatman TJ. (2004). Cancer Res 64: 7412–7419. Issa JP, Vertino PM, Wu J, Sazawal S, Celano P, Nelkin BD Criswell TL, Beman M, Araki S, Leskov K, Cataldo E, Mayo et al. (1993). J Natl Cancer Inst 85: 1235–1240. LD et al. (2005). J Biol Chem 280: 14212–14221. Itoh T, Tanioka M, Yoshida H, Yoshioka T, Nishimoto H, Deryugina EI, Luo GX, Reisfeld RA, Bourdon MA, Strongin Itohara S. (1998). Cancer Res 58: 1048–1051. A. (1997). Anticancer Res 17: 3201–3210. Ivanova T, Vinokurova S, Petrenko A, Eshilev E, Solovyova Deryugina EI, Ratnikov B, Monosov E, Postnova TI, N, Kisseljov F et al. (2004). Int J Cancer 108: 882–886. DiScipio R, Smith JW et al. (2001). Exp Cell Res 263: Jones SE, Jomary C. (2002). Int J Biochem Cell Biol 34: 209–223. 427–431.

Oncogene RAS induces hypermethylation of MAPK target genes P Lund et al 4903 Kautiainen TL, Jones PA. (1986). J Biol Chem 261: 1594–1598. Scaltriti M, Brausi M, Amorosi A, Caporali A, D’Arca D, Klock G, Storch S, Rickert J, Gutacker C, Koch-Brandt C. Astancolle S et al. (2004a). Int J Cancer 108: 23–30. (1998). J Cell Physiol 177: 593–605. Scaltriti M, Santamaria A, Paciucci R, Bettuzzi S. (2004b). Kyprianou N, English HF, Davidson NE, Isaacs JT. (1991). Cancer Res 64: 6174–6182. Cancer Res 51: 162–166. Scha¨ fer R, Iyer J, Iten E, Nirkko AC. (1988). Proc Natl Acad Lachner M, Jenuwein T. (2002). Curr Opin Cell Biol 14: 286–298. Soc USA 85: 1590–1594. Laduron S, Deplus R, Zhou S, Kholmanskikh O, Godelaine Schneider-Stock R, Dias-Assef M, Rohrbeck A, Foltzer- D, De Smet C et al. (2004). Nucleic Acids Res 32: 4340–4350. Jourdainne C, Boltze C, Hartig R et al. (2005). J Pharmacol Leskov KS, Klokov DY, Li J, Kinsella TJ, Boothman DA. Exp Ther 312: 525–536. (2003). J Biol Chem 278: 11590–11600. Schwarte-Waldhoff I, Martin W, Willecke K, Scha¨ fer R. Li LC, Dahiya R. (2002). Bioinformatics 18: 1427–1431. (1994). Oncogene 9: 899–909. Li Q, Ahuja N, Burger PC, Issa JP. (1999). Oncogene 18: Sers C, Tchernitsa OI, Zuber J, Diatchenko L, Zhumabayeva 3284–3289. B, Desai S et al. (2002). Adv Enzyme Regul 42: 63–82. Liu HS, Scrable H, Villaret DB, Lieberman MA, Stambrook Takai D, Jones PA. (2002). Proc Natl Acad Sci USA 99: PJ. (1992). Cancer Res 52: 983–989. 3740–3745. Lopes S, Lewis A, Hajkova P, Dean W, Oswald J, Forne T Taraboletti G, Morbidelli L, Donnini S, Parenti A, Granger et al. (2003). Hum Mol Genet 12: 295–305. HJ, Giavazzi R et al. (2000). FASEB J 14: 1674–1676. Lymar ES, Clark AM, Reeves R, Griswold MD. (2000). Biol Tchernitsa OI, Sers C, Zuber J, Hinzmann B, Grips M, Reprod 63: 1341–1351. Schwendel A et al. (2004). Oncogene 23: 4536–4555. Mansour SJ, Matten WT, Hermann AS, Candia JM, Rong S, Villar-Garea A, Esteller M. (2004). Int J Cancer 112: 171–178. Fukasawa K et al. (1994). Science 265: 966–970. Wang Z, Juttermann R, Soloway PD. (2000). J Biol Chem 275: Olek A, Oswald J, Walter J. (1996). Nucleic Acids Res 24: 26411–26415. 5064–5066. Watnick RS, Cheng YN, Rangarajan A, Ince TA, Weinberg Ordway JM, Williams K, Curran T. (2004). Oncogene 23: RA. (2003). Cancer Cell 3: 219–231. 3737–3748. Wei G, Schaffner AE, Baker KM, Mansky KC, Ostrowski Pucci S, Bonanno E, Pichiorri F, Angeloni C, Spagnoli LG. MC. (2003). Anticancer Res 23: 2173–2178. (2004). Oncogene 23: 2298–2304. Wolthuis RM, Bos JL. (1999). Curr Opin Genet Dev 9: 112–117. Raudies O, Kuban RJ, Hamacher F, Klein-Hitpass L, Tcher- Zhou X, Richon VM, Wang AH, Yang XJ, Rifkind RA, nitsa OI, Sers C et al. (2005). Adv Enzyme Regul 45: 63–84. Marks PA. (2000). Proc Natl Acad Sci USA 97: Reik W, Walter J. (2001). Nat Rev Genet 2: 21–32. 14329–14333. Reik W, Dean W, Walter J. (2001). Science 293: 1089–1093. Zohn IM, Campbell SL, Khosravi-Far R, Rossman KL, Der Rosemblit N, Chen CL. (1994). J Mol Endocrinol 13: 69–76. CJ. (1998). Oncogene 17: 1415–1438. Rouleau J, Macleod AR, Szyf M. (1995). J Biol Chem 270: Zuber J, Tchernitsa OI, Hinzmann B, Schmitz AC, Grips M, 1595–1601. Hellriegel M et al. (2000). Nat Genet 24: 144–152.

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

Oncogene