Increased Expression of Unmethylated CDKN2D by 5-Aza-2'-Deoxycytidine in Human Lung Cancer Cells

Increased expression of unmethylated CDKN2D by 5-aza-2'-deoxycytidine in human lung cancer cells

Wei-Guo Zhu1, Zunyan Dai2,3, Haiming Ding1, Kanur Srinivasan1, Julia Hall3, Wenrui Duan1, Miguel A Villalona-Calero1, Christoph Plass3 and Gregory A Otterson*,1

1Division of Hematology/Oncology, Department of Internal Medicine, The Ohio State University-Comprehensive Cancer Center, Columbus, Ohio, OH 43210, USA; 2Department of Pathology, The Ohio State University-Comprehensive Cancer Center, Columbus, Ohio, OH 43210, USA; 3Division of Human Cancer Genetics, Department of Molecular Virology, Immunology and Medical Genetics, The Ohio State University-Comprehensive Cancer Center, Columbus, Ohio, OH 43210, USA

DNA hypermethylation of CpG islands in the promoter Introduction region of genes is associated with transcriptional silencing. Treatment with hypo-methylating agents can Methylation of cytosine residues in CpG sequences is a lead to expression of these silenced genes. However, DNA modi®cation that plays a role in normal whether inhibition of DNA methylation in¯uences the mammalian development (Costello and Plass, 2001; expression of unmethylated genes has not been exten- Li et al., 1992), imprinting (Li et al., 1993) and X sively studied. We analysed the methylation status of chromosome inactivation (Pfeifer et al., 1990). To date, CDKN2A and CDKN2D in human lung cancer cell lines four mammalian DNA methyltransferases (DNMT) and demonstrated that the CDKN2A CpG island is have been identi®ed (Bird and Wole, 1999). Disrup- methylated, whereas CDKN2D is unmethylated. Treat- tion of the balance in methylated DNA is a common ment of cells with 5-aza-2'-deoxycytidine (5-Aza-CdR), alteration in cancer (Costello et al., 2000; Costello and an inhibitor of DNA methyltransferase 1, induced a dose Plass, 2001; Issa et al., 1993; Robertson et al., 1999). A and duration dependent increased expression of both proposed reason for this disruption seen in cancer is p16INK4a and p19INK4d, the products of CDKN2A and that genes that are methylated in the promoter regions CDKN2D, respectively. These data indicate that global (such as critical tumor suppressor genes) may be DNA demethylation not only in¯uences the expression of transcriptionally silenced in the tumors, thus providing methylated genes but also of unmethylated genes. a growth advantage for these cells (Jones and Laird, Histone acetylation is linked to methylation induced 1999). Usually, hypermethylation mediated gene silen- transcriptional silencing. Depsipeptide, an inhibitor of cing is limited to CpG islands in the promoter region histone deacetylase, acts synergistically with 5-Aza-CdR of genes (Baylin et al., 1998; Chan et al., 2000). In in inducing expression of p16INK4a and p19INK4d. However, contrast, gene exons are more frequently hypermethy- when cells were treated with higher concentrations of 5- lated at isolated CpGs (outside of CpG islands) and Aza-CdR and depsipeptide, p16INK4a expression was these methylated exons are not related to gene silencing decreased together with signi®cant suppression of cell (Herman and Baylin, 2000). There are some reports growth. Interestingly, p19INK4d expression was enhanced that the degree of methylation of CpG islands outside even more by the higher concentrations of 5-Aza-CdR of the promoter is positively correlated with an and depsipeptide. Our data suggest that p19INK4d plays a increase in gene expression (Chan et al., 2000; Glenn distinct role from other INK4 family members in et al., 1996; Gonzalgo et al., 1998). response to the cytotoxicity induced by inhibition of Recently, evidence has been accumulating that DNA methylation and histone deacetylation. Oncogene methylation inhibits gene transcription either directly (2001) 20, 7787 ± 7796. by interrupting the binding of transcription factors to promoters (Bird and Wole, 1999; Staiger et al., 1989), Keywords: DNA methylation; histone acetylation; or indirectly through methyl DNA binding proteins CDKN2D; p16INK4a; p19INK4d (MBPs) that bind preferentially to methylated DNA and recruit histone deacetylase enzymes (Nan et al., 1998; Ng and Bird, 1999). The binding sites of some transcription factors such as AP-2, E2F, and c-myc contain CpG dinucleotides (Tate and Bird, 1993). It *Correspondence: GA Otterson, Division of Hematology and has been postulated that methylation occurring at these Oncology, Department of Internal Medicine, The Ohio State binding sites may repress gene expression by blocking University, Room B415, Starling Loving Hall, 320 West 10th the access of transcription factors to the methylated Avenue, Columbus, OH 43210-1240, USA; DNA sequence (Kumari and Usdin, 2000; Staiger et E-mail: [email protected] Received 22 May 2001; revised 5 September 2001; accepted 13 al., 1989). Alternatively, or in addition to this September 2001 mechanism, MBPs have been shown to bind speci®cally 5-Aza-CdR increases expression of unmethylated CDKN2D W-G Zhu et al 7788 to methylated DNA sequences and may compete with tions of 5-Aza-CdR and depsipeptide induce decreased transcription factors for their binding sites (Robertson p16INK4a expression but even higher p19INK4d expres- and Jones, 2000). MBPs have also been demonstrated sion. to recruit histone deacetylases to hypermethylated areas of the promoter, thereby increasing localized histone deacetylation (Bird and Wole, 1999; Nan et Results al., 1998), resulting in gene repression. However, some genes appear to be repressed by The CpG island in the promoter of CDKN2D is mechanisms distinct from methylation. For example, unmethylated in human lung cancer cells Loeb et al. (2001) analysed the promoter region of the Wilms' tumor suppressor 1 gene (WT1) in 20 samples Although evidence has shown that the promoter region of normal breast epithelium and 19 primary breast of CDKN2A is hypermethylated in many human carcinomas. They found that the promoter region was cancers and cancer derived cell lines (Herman et al., hypermethylated in six out of 19 tumors and in one out 1995; Otterson et al., 1995), there has been no report of 20 normal breast epithelial cells. However, WT1 studying the methylation status of CDKN2D.To mRNA or protein was expressed in the six tumors that investigate the CpG island in the promoter region of were hypermethylated, whereas WT1 protein was not CDKN2D, we used the WebGene computer program expressed in the 19 normal breast epithelial cells that (http://www.itba.mi.cnr.it/webgene/) to select CpG were unmethylated (Loeb et al., 2001). In addition, islands around the predicted promoter of CDKN2D. Evron et al. (2001) have reported that although most Brie¯y, as shown in Figure 1a, a non-repetitive region breast tumors showed a correlation between hyper- B (from 71100 to +1) was identi®ed as a CpG island. methylation of the cyclin D2 promoter and gene In this region the GC content is 67% and the observed/ silencing, some did not. By analysing methylation in expected presence of CpG is 0.84, consistent with this the promoter of cyclin D2 in 13 breast tumors by being a CpG island (Gardiner-Garden and Frommer, methylation speci®c PCR (MS-PCR), the investigators 1987). In order to further narrow the putative found that ®ve of 13 tumors were unmethylated in the promoter region of CDKN2D, three computer algo- promoter sequences. Among these samples, two rithms were used to predict the transcriptional start showed cyclin D2 mRNA and protein expression site. Using Promoter 2.0 (Knudsen, 1999), McPromoter whereas three did not. These observations suggest (Ohler et al., 2000) and Promoter Scan II (Prestridge, alternative transcriptional silencing pathways. 1995), the transcriptional start sites fall within a 350 bp Loss of p16INK4a and p15INK4b protein expression in region C that is between (7) 400 to (7)750 bp from cancers has been frequently related to DNA methyla- the translational start site (+1). Our estimate regarding tion (Herman et al., 1995, 1996b; Otterson et al., 1995). the promoter region of human CDKN2D is consistent The methylation status of the related INK4 family with that suggested by another group (Yue Xiong, members, p18INK4c and p19INK4d has been less well personal communication). studied. Among the four members of the INK4 family, The methylation status of the CpG island in the methylation induced inactivation of p16INK4a has been region B (71000 to +1) was investigated with extensively reported in almost all human cancer cells Southern blot hybridization using a methylation (Herman et al., 1995), whereas methylation mediated sensitive restriction enzyme NotI in 12 human lung inactivation of p15INK4b has been described principally cancer cell lines and normal human lung tissue. DNA in hematologic malignancies (Herman et al., 1996b). As from human normal lung tissue demonstrated bands of a member of the family of cyclin D kinase inhibitors, 24 and 5.8 kb following digestion with EcoRV or p19INK4d is reported to play a similar function to EcoRV/NotI, respectively (Figure 1b, lanes 1 and 2). p16INK4a (Guan et al., 1996; Okuda et al., 1995). This The 24 kb band after incubation with both enzymes functional similarity to p16INK4a implies that p19INK4d would be consistent with lack of NotI digestion, may functionally replace p16INK4a when there is suggesting methylation of the NotI site. DNA from deletion, mutation or methylation of CDKN2A. all of the human lung cancer cell lines tested was Here we planned to investigate the methylation completely digested with NotI (showing the 5.8 kb status of CDKN2D and determine if expression of band), indicating an unmethylated NotI site in all p19INK4d is in¯uenced by treatment of cells with a human lung cancer cell lines tested (Figure 1b). As a DNA demethylating agent (5-aza-2'-deoxycytidine (5- comparison, we also performed Southern hybridization Aza-CdR)) or histone deacetylase inhibitor in human using SacII as a methylation-sensitive restriction lung cancer cells. We also analysed methylation status enzyme to detect methylation status of CDKN2A in of CDKN2A and expression of p16INK4a in these cells. these human lung cancer lines as previously described Our data show that the CpG islands of both genes (Herman et al., 1995; Otterson et al., 1995) (Figure 1c). have a distinct methylation status, but that there is a This analysis shows that H23, H211, H719 and H841 similar eect on protein expression after treatment with (lanes 4, 10, 12 and 14 respectively) exhibit a pattern 5-Aza-CdR. In addition, low concentrations of 5-Aza- consistent with methylation of the SacII site. H1155 CdR combined with a histone deacetylase inhibitor (lane 6) exhibits a partial methylation pattern such that (depsipeptide) lead to a synergistic re-expression of there are two bands of 4.3 kb and 3.4 kb seen after both protein products. In contrast, higher concentra- digestion with EcoRI/SacII. Other cell lines exhibit

Oncogene 5-Aza-CdR increases expression of unmethylated CDKN2D W-G Zhu et al 7789

Figure 1 Predicted promoter of CDKN2D and Southern blot analysis of the CDKN2A and CDKN2D CpG islands. (a) A repetitive region (region A, from 710 000 to 71100) and a nonrepetitive region (region B, from 71100 to +1) upstream of the open reading frame of CDKN2D were identi®ed by a computer program (WebGene). Region B (71100 to +1) was found to be a CpG island. An Alu repetitive element was immediately upstream of the CpG island. The predicted transcriptional start site was estimated to lie between position 7750 and 7400 bp (region C). Bisul®te DNA sequencing was performed in a region including 86 CGs and covering 792 bp (region D, from 7774 to +18 bp). The position of NotI and EcoRV sites is indicated. Figure not drawn to scale. (b) Human lung cancer cell lines were digested with EcoRV/NotI for CDKN2D,or(c) EcoRI/SacII for CDKN2A. Normal human lung tissue was utilized as a control (NL). The 24 kb and the 5.8 kb bands in b represent a methylated NotI and an unmethylated NotI site, respectively. The 4.3 kb and the 3.4 kb bands in c represent a methylated SacII site and an unmethylated SacII site, respectively

deleted (H290, H125 and H792) or unmethylated (H69, three of the transcriptional start sites predicted by the H82, H209 and N417) CDKN2A (Figure 1c). computer algorithms noted above as well as part of One drawback to Southern blot analysis to study exon 1. DNA from human normal lung tissue was methylation is that it detects methylation of one or at utilized as a control. H23 and H719 cells with most two CpGs. To more thoroughly study the hypermethylated CDKN2A were selected for the methylation status of the CpG island in CDKN2D, bisul®te sequencing. Methylated CpGs are rare in this we designed three sets of ampli®cation primers in the region in normal human lung tissue. When we putative promotor region (Figure 1a, Region D) that examined ®ve separate clones from these cell lines includes 86 CGs from (7)774 to (+)18 from the and from normal human lung tissue, we found, on translational start site to perform bisul®te sequencing. average, less than one methylated cytosine per clone This region covers most of the CpG island and all out of 86 CGs in this region (data not shown). In

Oncogene 5-Aza-CdR increases expression of unmethylated CDKN2D W-G Zhu et al 7790 contrast, the CpG island in CDKN2A is heavily immunoblotting, no detectable expression of p16INK4a methylated. One study reported methylation of 31 ± in H23 or H719 cell line was observed at baseline 75% of cytosines from dierent clones of human (Figure 2a ± d). When these cells were treated with 5- bladder cancer cells were treated with bisul®te and then Aza-CdR however, a dose and duration dependent sequenced (Gonzalgo et al., 1998). This data together expression of p16INK4a was observed in both cell lines with ours shows that methylation of the entire (Figure 2a ± d). Interestingly, upon 5-Aza-CdR treat- promoter is mirrored by the methylation status of the ment, similar results were observed in p19INK4d enzymes selected for Southern blot analysis (SacII for expression in both cell lines (Figure 2a ± d). As shown CDKN2A and Not1 for CDKN2D). in Figure 2, there is low (H23) or undetectable (H719) expressions of p19INK4d in untreated control samples. Treatment with 5-Aza-CdR at 1 mM for various times Inhibition of DNA methylation induces expression of (from 6 to 72 h) also increased the expression of silenced p16INK4a and increases expression of p19INK4d p19INK4d in both cell lines (Figure 2a,b). In addition, Since H23 and H719 cell lines have a hypermethylated low concentrations of 5-Aza-CdR were sucient to CpG island in CDKN2A (Figure 1c), we selected these induce an increase in p19INK4d. The expression of two cell lines as target cells to detect the eects of p19INK4d, for instance, was increased when cells were treatment with a DNA demethylating agent on treated with 5-Aza-CdR at 0.1 mM (Figure 2c,d). The 5- p16INK4a and p19INK4d expression. Using Western Aza-CdR induced increase in p19INK4d reached a

Figure 2 Western blot analysis of p16INK4a and p19INK4d expression. H23 (a) or H719 cells (b) were treated with 5-Aza-CdR (1 mM) for 6 ± 72 h and harvested at 96 h after initial treatment. Equal amounts of protein were size fractionated on 15% SDS ± PAGE for detection of protein expression. H23 (c) or H719 cells (d) were treated with 5-Aza-CdR at varying concentrations (0.1 to 10 mM) for 72 h and then harvested at 96 h after initial treatment. p16INK4a, p19INK4d and a-tubulin expression were analysed by Western immunoblot. a-tubulin was chosen as a loading control in all blots. (e) H23 or H719 cells were treated with 5-Aza-CdR (1 mM)as indicated and DNA was extracted at 96 h after initial treatment for MS-PCR as described. DNA from human normal lung tissue was treated with SssI methylase as a positive methylated control (Positive M lane) or left untreated as a positive unmethylated control (Positive U lane). CTR lanes represent either H719 or H23 DNA from cells untreated with 5-Aza-CdR (demonstrating 100% methylation as previously reported)

Oncogene 5-Aza-CdR increases expression of unmethylated CDKN2D W-G Zhu et al 7791 maximum level between 0.5 and 1 mM in both cell lines et al., 2001), mRNA level was quanti®ed and normal- (Figure 2c,d). ized by a loading control, GAPDH (Figure 3b). The The 5-Aza-CdR induced p16INK4a expression is due relative mRNA level was then plotted as a ratio of the to a partial demethylation of CDKN2A (Figure 2e). By treated samples to untreated control (Figure 3c). A MS-PCR, the completely methylated CDKN2A gene 2.2 ± 2.4-fold increase in p19INK4d mRNA was observed was partially changed to unmethylated status when when H719 cells were treated with 5-Aza-CdR for 24 ± both cell lines were treated with 5-Aza-CdR at 1 mM 72 h (Figure 3c). Compared to 5-Aza-CdR treated for 24 to 72 h (Figure 2e). Since CDKN2D is cells, the p19INK4d mRNA level was increased 3.2-fold unmethylated, we did not observe any changes in (24 h), ®vefold (48 h) and 2.9-fold (72 h) in cells methylation status of this gene when cells were treated treated with 5-Aza-CdR and depsipeptide (Figure 3c). with 5-Aza-CdR by using MS-PCR (data not shown). However, when the same membrane was stripped and then hybridized with p16INK4a probe, the increase in p16INK4a mRNA level was not observed in the Synergistic effect of DNA methyltransferase and histone treatment with a combination of both drugs (data deacetylase inhibition on p19INK4d and p16INK4a expression not shown). We used depsipeptide (FR901228) as a histone These changes in p19INK4d and p16INK4a mRNA deacetylase inhibitor to test the hypothesis that levels are also consistent with results from Western blot inhibition of histone deacetylase might enhance 5- (Figure 4). Compared to 5-Aza-CdR (1 mM) treated Aza-CdR induced expression of p16INK4a and/or H719 cells, a signi®cant increase in p19INK4d was p19INK4d. H719 cells were treated with 5-Aza-CdR observed in the cells treated with 5-Aza-CdR (1 mM) (1 mM) for 24, 48 or 72 h, respectively, and then together with depsipeptide (0.05 mM) (Figure 4a, top depsipeptide (0.05 mM) was added into the 5-Aza-CdR panel). In contrast to the expression of p19INK4d, pretreated cells for the ®nal 6 h (18 ± 24, 42 ± 48 and p16INK4a expression was decreased after treatment with 66 ± 72 h). Northern blot RNA analysis shows that 5- 5-Aza-CdR and depsipeptide in the same protein Aza-CdR (1 mM) induced a duration dependent in- lysates (Figure 4a, middle panel). Similar to p16INK4a crease in p19INK4d mRNA and this increase was expression, p18INK4c expression was decreased when enhanced in the presence of depsipeptide (Figure 3a). cells were treated with the same concentrations of 5- Since treatment with 5-Aza-CdR and depsipeptide Aza-CdR and depsipeptide (Figure 4c). p15INK4b induced a signi®cant enhancement of cell death (Zhu protein was not detected in control H719 cells or in

Figure 3 CDKN2D RNA blot analysis. (a) H719 cells were treated with 5-Aza-CdR (1 mM) for 24 to 72 h, or depsipeptide (0.05 mM) was added to the 5-Aza-CdR pretreated cells for the ®nal 6 h (at 18 ± 24, 42 ± 48 or 66 ± 72 h, respectively). Cells were also treated with depsipeptide alone at 0.05 mM for 6 h. The treated cells were then harvested at 96 h from initial treatment. mRNA was extracted and loaded equally on agarose-formaldehyde gel for Northern blot. (b) GAPDH was chosen as a loading control. (c) The mRNA level of the CDKN2D expression was calculated. The density for each band of CDKN2D mRNA was compared to the density of the control band. The relative density of each band was then compared to the GAPDH band

Oncogene 5-Aza-CdR increases expression of unmethylated CDKN2D W-G Zhu et al 7792

Figure 4 Western blot of p16INK4a, p18INK4c, p19INK4d and MTT assay in treated cells. H719 cells were treated as indicated. At 96 h after initial treatment, cells were harvested and equal amounts of protein were size fractionated on SDS ± PAGE. The changes in p16INK4a and p19INK4d are shown in (a) and the changes in p18INK4c are shown in (c). (b) H719 cells were treated with low concentrations of 5-Aza-CdR and depsipeptide and analysed as described above. (d) MTT assay showing cell growth suppression when cells were treated with 5-Aza-CdR and depsipeptide. Lane 1: untreated control. Lanes 2 ± 4: cells were treated with 5-Aza-CdR (1 mM) for 24, 48 and 72 h, respectively. Lanes 5 ± 7: cells were treated with 5-Aza-CdR (1 mM) for 24, 48 and 72 h, respectively with depsipeptide (0.05 mM) added to the 5-Aza-CdR pretreated cells for the ®nal 6 h. Lane 8: cells were treated with depsipeptide (0.05 mM) alone for 6 h. Lane 9: cells were treated with low concentration of 5-Aza-CdR (0.25 mM) for 72 h. Lane 10: cells were treated with 5-Aza-CdR (0.25 mM) for 72 h and depsipeptide (0.025 mM) for 6 h at 66 ± 72 h. Lane 11: cells were treated with depsipeptide (0.025 mM) for 6 h

5-Aza-CdR treated H719 cells with Western blot (data Discussion not shown). To investigate the in¯uence that 5-Aza-CdR and Our data regarding CDKN2A in this study provides depsipeptide treatment induced apoptotic cell death direct evidence that methylation in CpG islands of may have on the expression of p16INK4a and p19INK4d, CDKN2A is linked to gene silencing in human lung low concentrations of 5-Aza-CdR and depsipeptide cancer cell lines as demonstrated by Southern blot, were used in these same cells. As shown in Figure 4b, MS-PCR and Western blot (Figures 1 and 2). both p16INK4a and p19INK4d were increased in cells Treatment with the DNA methyltransferase inhibitor treated with lower concentrations of 5-Aza-CdR (5-Aza-CdR) leads to re-expression of p16INK4a, (0.25 mM) and depsipeptide (0.025 mM) compared to indicating that hypermethylation of CDKN2A is the 5-Aza-CdR (0.25 mM) alone treated cells. Results from primary inactivating event in these cells (Figure 2). MTT assay further con®rmed that higher concentra- Interestingly, treatment with 5-Aza-CdR induced an tions of 5-Aza-CdR and depsipeptide induce a increased expression of p19INK4d (Figure 2). However, signi®cant suppression of cell growth (Figure 4d). we found no consistent CpG methylation of Compared to 5-Aza-CdR alone treated cells, for CDKN2D by Southern blot (Figure 1b), or bisul®te example, cell growth rate decreases approximately sequencing. ®vefold when H719 cells were treated with higher A lack of association between gene repression and concentrations of the combination of 5-Aza-CdR and promoter methylation status has been reported in depsipeptide (Figure 4d, lanes 2 ± 4 compared with several breast tumors and cancer cell lines by Evron et lanes 5 ± 7). In contrast, a lower concentration of 5- al. (2001) and Loeb et al. (2001). These investigators Aza-CdR and depsipeptide compared with 5-Aza-CdR did not test whether treatment with DNA methyl- alone had only a marginal eect on cell growth rate transferase inhibitors (such as 5-Aza-CdR) could such that the cell growth was only 35% less in the induce expression of unmethylated silenced genes. combination treated cells (Figure 4d, compare lanes 9 Recently, it was reported that inhibition of DNA and 10). methylation was able to induce expression of a

Oncogene 5-Aza-CdR increases expression of unmethylated CDKN2D W-G Zhu et al 7793 silenced gene that was unmethylated in its promoter in cells (data not shown). No methylation changes in the melanoma cells (Soengas et al., 2001). The Apaf-1 Alu element were found in H23 cells before or after 5- gene, which is involved in p53-dependent apoptosis Aza-CdR treatment (data not shown). We therefore (Soengas et al., 1999), had no mutations or deletions conclude that methylation status of the Alu repetitive in eight melanoma cell lines tested (Soengas et al., is not relevant to 5-Aza-CdR induced increase in 2001). By analysing the CpG island of Apaf-1 gene in p19INK4d expression. the putative promoter region, no CpG methylation A synergistic eect of inhibition of histone was found. However, a signi®cant increase in Apaf-1 deacetylase and DNA demethylation on gene expres- expression was demonstrated after 5-Aza-CdR treat- sion has been reported previously (Cameron et al., ment, indicating that an alternative mechanism of 1999; Eden et al., 1998). Consistent with these reports, epigenetic silencing for the inactivation of Apaf-1. The our study shows that treatment with a low concentra- authors suggest the possibility of methylation related tion of 5-Aza-CdR and depsipeptide enhanced the alterations of an enhancer element of Apaf-1 or of a expression of p16INK4a and p19INK4d (Figure 4b). methylated transcription factor being expressed lead- Although the mechanism of this synergistic activity ing to Apaf-1 expression. on gene expression is unclear, a model demonstrating Although the mechanism of enhanced p19INK4d a connection between histone acetylation and DNA expression remains to be elucidated, DNA demethyla- methylation is attractive. A group of proteins with tion may reactivate an unknown gene or transcription methyl DNA binding activity have been identi®ed and factors that controls expression of CDKN2D. Alter- demonstrate a link between DNA methylation and natively, methylation of a gene enhancer or some histone acetylation (Hendrich and Bird, 2000). One of other regulatory element outside of the promoter these identi®ed methyl DNA binding proteins, region might interfere with CDKN2D expression. The MeCP2, is capable of binding to methylated CpG insulin-like growth factor 2 (IGF2) and H19 genes are within DNA by a methyl-CpG-binding domain (Cross imprinted, resulting in silencing of the maternal and et al., 1994) and interacting with Sin3A by a paternal alleles, respectively, and share an enhancer transcriptional repression domain (Nan et al., 1998). (Yoo-Warren et al., 1988). A region of paternal- Since Sin3A interacts with histone deacetylases known speci®c methylation upstream of H19 contains an to be associated with gene repression (Alland et al., element that blocks enhancer activity (Elson and 1997), the methylated DNA is connected with gene Bartolomei, 1997). The activity of this element is repression by MeCP2. Furthermore, recent reports regulated by an enhancer blocking protein CTCF and also demonstrated that DNA methyltransferase 1 is the binding of CTCF was abolished if its binding site directly associated with histone deacetylases (Fuks et is methylated (Bell and Felsenfeld, 2000; Hark et al., al., 2000; Robertson et al., 2000; Rountree et al., 2000). These reports reveal that DNA methylation can 2000). modulate enhancer function to control gene expres- However, in contrast to prior reports, when we sion. Another possibility for the increased expression treated cells with higher concentrations of 5-Aza-CdR of p19INK4d by 5-Aza-CdR is that 5-Aza-CdR itself and depsipeptide, the expression of p16INK4a is may activate CDKN2D by trapping the DNA decreased (Figure 4a). This decrease in 16INK4a methyltransferase enzyme in a covalent complex with expression may result from apoptotic cell death (Zhu DNA during replication (Juttermann et al., 1994). In et al., 2001). Similarly, in this study, by analysing cell addition, as 5-Aza-CdR is incorporated into DNA, it growth with the MTT assay, we also observed that is feasible that part of the eects we (and others) have growth suppression is signi®cantly higher with higher observed may be related to an anti-metabolite concentrations of 5-Aza-CdR and depsipeptide than at mechanism of action through non-speci®c eects. We lower concentrations (Figure 4d). Therefore, we do not believe this to be the principal mechanism of conclude that the decrease in p16INK4a expression after action since the gene expression and synergistic higher concentrations of 5-Aza-CdR and depsipeptide cytotoxic eects we have noted are not seen when may be due to cell death. Supporting this, 5-Aza-CdR we utilize other cytosine analogs such as cytosine and depsipeptide-mediated cell death induced a arabinoside in combination with HDAC inhibitors decrease in p18INK4c expression as well (Figure 4c). (unpublished observations). In addition, evidence from Interestingly, however, p19INK4d expression is still several groups indicates that repetitive DNA, such as increased regardless of cell death induced by the Alu elements, may be a target for epigenetic silencing higher concentrations of 5-Aza-CdR and depsipeptide (Garrick et al., 1998; Tycko, 2000). Usually, Alu treatment (Figure 4a). The mechanism of p19INK4d elements are hypermethylated, which may induce loss expression upon inhibition of DNA methylation and of promoter occupancy or alter promoter con®gura- histone deacetylase is unclear. Future studies sug- tion, thus leading to promoter inactivation (Gra et gested by this work include an analysis of the al., 1997). We analysed an Alu repetitive region methylation status of transcription factors predicted (Figure 1a, 71380 to 71164 bp from the transla- to bind in the CDKN2D promoter. An additional tional start site) that is a boundary of the CpG islands question for further investigation raised by this work of CDKN2D using Southern blot with a methylation is whether the increase in p19INK4d expression plays a sensitive enzyme, SmaI. We found very low level role in apoptotic cell death observed in cells treated methylation in H23 cells and no methylation in H719 with 5-Aza-CdR and depsipeptide.

Oncogene 5-Aza-CdR increases expression of unmethylated CDKN2D W-G Zhu et al 7794 Materials and methods Northern blot analysis Total RNA was extracted with TRIzol Reagent (Life Cell culture and drug treatment Technologies, Gaithersburg, MD, USA). Messenger RNA Human lung cancer cell lines H23, H69, H82, H125, H209, was isolated from total RNA by an Oligotex mRNA Mini H211, H290, H719, H792, H841, H1155 and N417 were Kit (Qiagen, Valencia, CA, USA). CDKN2A exon 1a grown in RPMI medium 1640 with 10% fetal bovine serum fragment (340 bp) was generated by PCR from normal and penicillin/streptomycin at 378C (5% CO2). 5-Aza-CdR human DNA. The exon 1a primers and PCR conditions (Sigma, St. Louis, MO, USA) was dissolved in 50% acetic are as follows: 5'-GAA GAA AGA GGA GGG GCT G acid as a 10 mM stock solution. Aliquots were prepared and (forward), 5'-GCG CTA CCT GAT TCA AAT TC (reverse), frozen at 7808C. Stock solutions of depsipeptide (kind gift 958C for 10 min, 968C30s,628C40s,728C for 30 s, and of Michael Grever) were prepared in DMSO. Exponentially 728C for 7 min, 35 cycles. The CDKN2D probe was growing cells were treated with 5-Aza-CdR at dierent generated from CDKN2D cDNA by RT ± PCR. The primers doses (from 0.1 to 10 mM) for 72 h, or treated with of CDKN2D and RT ± PCR conditions: 5'-CAA CCG CTT 5-Aza-CdR (1 mM) for dierent times (6 to 72 h). Fresh CGG CAA GAC (forward), 5'-CAG GGT GTC CAG GAA 5-Aza-CdR was replaced every 24 h and cells were harvested TCC A (reverse), 958C for 10 min, 968C for 39 s, 648C for at 96 h after initial treatment. Depsipeptide (0.025 ± 0.05 mM) 1 min, 728C for 66 s and 728C for 7 min, 35 cycles. was added into 5-Aza-CdR pretreated cells for the ®nal 6 h Northern blot analysis was performed as follows: 1 mgof of drug treatment at 18 ± 24, 42 ± 48 and 66 ± 72 h, respec- mRNA was size-fractionated by agarose-formaldehyde gel tively, to observe the changes of p16INK4a and p19INK4d electrophoresis (1%) and transferred to Nylon membrane expression. (Hybond-N, Amersham, Amersham, UK). Hybridization with random-primed 32P-labeled probes was performed at 428C overnight. Filters were washed (26SSC/0.1% SDS, Southern hybridization 5 min twice at RT; 0.2 SSC/0.1% SDS, 5 min twice at RT Southern hybridization was performed as described pre- and 15 min twice at 428C; 0.1 SSC/0.1% SDS, 15 min twice viously (Otterson et al., 1995). Brie¯y, DNA from normal at 688C). The washed membrane was exposed to X-ray ®lm human lung tissue or lung cancer cell lines was utilized. For at 7808C for 4 ± 7 days and subsequently scanned with a CDKN2A, SacII was chosen as a methylation sensitive phosphor imager. enzyme, whereas NotI was chosen as a methylation sensitive enzyme for CDKN2D. DNA was doubly digested with NotI/ Western immunoblotting EcoRV to detect methylation in CDKN2D and with SacII/ EcoRI to detect methylation in CDKN2A. The probes were Cells were lysed in lysis buer (50 mM Tris-HCl, 250 mM prepared by PCR ampli®cation from human genomic DNA. NaCl, 5 mM EDTA, 50 mM NaF, 0.15% Igepal CA-630 and The primers used were: CDKN2A,5'-TTG GAA ACA AAG 1.5 mM PMSF). Equal amount of proteins (100 ± 150 mg) CCA TTT CC (forward) 5'-CCA AGA CGA AGG ACT were size fractionated on 12.5 ± 15% SDS ± PAGE. Proteins TCA TTG (reverse); CDKN2D,5'-CCG GAT GCA GAT were then transferred onto a nitrocellulose membrane. The TTT AGA GG (forward), 5'-CCC ATT GCA GCA CTT membrane was blocked with blocking buer (5% nonfat CAG TA (reverse). Appropriately sized PCR fragments were milk, 200 mM NaCl, 50 mM Tris and 0.05% Tween-20). The puri®ed from agarose gels, and subsequently labeled with blocked membrane was then incubated with primary a-32P-CTP using the Prime it II kit (Stratagene, La Jolla, antibodies at 48C overnight. After washing the membrane CA, USA). with TBS-T (20 mM Tris, 500 mM NaCl and 0.1% Tween-20) for 665 min, the membrane was incubated with secondary antibody at 48C for 1 h. The detection of speci®c protein MS-PCR binding was performed with a chemiluminescence kit The methylation speci®c PCR (MS-PCR) was performed as (Amersham Pharmacia Biotech, Uppsala, Sweden). The described previously (Herman et al., 1996a). Brie¯y, antibodies and concentrations used are anti-p16INK4a (Phar- genomic DNA (1 mg) in a volume of 50 ml was denatured Mingen, San Diego, CA, USA, 1 mg/ml); anti- p18INK4c (Ab- by NaOH (®nal concentration, 0.275 M) for 10 min at 1) and anti-p19INK4d (Ab-1) (Oncogene Research Products, 378C. The denatured DNA was then treated with 10 ml of Darmstadt, Germany, 2 mg/ml); a-tubulin (Oncogene Re- 10 mM hydroquinone, 520 ml of 3 M sodium bisul®te at search Products, 0.3 mg/ml). 508C overnight. The bisul®te-modi®ed DNA was puri®ed with Qiaquick gel extraction kit (Qiagen, Germany) DNA bisulfite sequencing according to the manufacturer's instruction. The DNA was then precipitated with sodium acetate (®nal concentra- DNA from human normal lung tissue (as control), H23 and tion, 0.45 M) and isopropanol. DNA was eluted with dH2O H719 cell lines was bisul®te treated and puri®ed for PCR. and used for PCR. The primers and PCR conditions are as The PCR products were gel extracted (Qiagen) and ligated follows: into a plasmid vector, pCR2.1-TOPO using the TA cloning system (Invitrogen, Carlsbad, CA, USA). Plasmid-trans- Methylated p16, 5'-TTA TTA GAG GGT GGG GCG formed bacteria TOP10F' was cultured overnight and the GAT CGC (forward), 5'-GAC CCC GAA CCG GGA plasmid was isolated (Qiagen). According to the sequence of CCG TAA (reverse), 958C 10 min, 968C39s,678C40s, CpG islands around the putative CDKN2D promoter region 728C40s,728C 10 min, 34 cycles; (see Figure 1a), three pairs of primers were designed for PCR. Pair A: 5'-TTA TTT GGT TGT TTG AAT GTG Unmethylated p16, 5'-TTA TAA GAG GGT GGG CTG (forward), 5'-CCC ACA AAA CCT ACA TAA CC (reverse). GAT TGT (forward), 5'-CAA CCC CAA ACC ACA ACC Pair B: 5'-TTA TTG GTT ATG TAG GTT TTG TGG ATA A (reverse), 958C 10 min, 968C39s,588C30s,728C (forward), 5'-CCT AAC TCA CCC TCC CTC CT (reverse). 30 s, 728C 10 min, 35 cycles. Pair C: 5'-AGG AGGGAG GGT GAG TTA G (forward),

Oncogene 5-Aza-CdR increases expression of unmethylated CDKN2D W-G Zhu et al 7795 5'-AAC CTC CTC CAA CAA CAT ATC (reverse). For each Abbreviations cell line or control sample, at least ®ve clones were picked for 5-Aza-CdR, 5-aza-2'-deoxycytidine; DNMT, DNA Methyl- sequence analysis. transferase; HDAC, Histone deacetylase; MBP, Methyl- CpG binding proteins Cell viability assay An equal number of cells (approximately 5000) were seeded into a 96-well plate (Becton Dickinson and Company, Franklin Lakes, NJ, USA) 24 h prior to experiments. Cells were treated with 5-Aza-CdR, depsipeptide alone or combination of both drugs. At 96 h after various treatments, dimethylthiazolyl-2-5-diphenyltetrazolium bromide (MTT) dye solution (Sigma, St. Louis, MO, USA) was added into Acknowledgments the 96-well plate for 4 h at 378C and then Stop Solution We thank Drs Yue Xiong and Xin-Hai Pei for kindly (isopropanol with 0.04 N HCl) was added into the plate. The providing us information about the putative promoter absorbance at 570 nm wavelength was recorded using an region of human CDKN2D. We also thank Drs Lawrence ELISA plate reader (Dynatech MR7000). The average of the Druhan and Melissa Hunter for their technical assistance, absorbance values in column 1 (with no cells) was used as a and Dr Michael Grever for the gift of depsipeptide. This blank value and subtracted from all absorbance values to work is supported in part by grants from the American yield corrected absorbance values. The corrected absorbance Cancer Society, Ohio Division, Ohio Cancer Research of each sample was calculated by comparing with untreated Associates and by grant P30 CA16058, National Cancer control. Institute, Bethesda, MD, USA.

References

Alland L, Muhle R, Hou Jr H, Potes J, Chin L, Schreiber- GuanKL,JenkinsCW,LiY,O'KeefeCL,NohS,WuX, Agus N and DePinho RA. (1997). Nature, 387, 49 ± 55. Zariwala M, Matera AG and Xiong Y. (1996). Mol. Biol. Baylin SB, Herman JG, Gra JR, Vertino PM and Issa JP. Cell, 7, 57 ± 70. (1998). Adv. Cancer Res., 72, 141 ± 196. Hark AT, Schoenherr CJ, Katz DJ, Ingram RS, Levorse JM Bell AC and Felsenfeld G. (2000). Nature, 405, 482 ± 485. and Tilghman SM. (2000). Nature, 405, 486 ± 489. Bird AP and Wole AP. (1999). Cell, 99, 451 ± 454. Hendrich B and Bird A. (2000). Curr. Top. Microbiol. Cameron EE, Bachman KE, Myohanen S, Herman JG and Immunol., 249, 55 ± 74. Baylin SB. (1999). Nat. Genet., 21, 103 ± 107. Herman JG and Baylin SB. (2000). Curr.Top.Microbiol. Chan MF, Liang G and Jones PA. (2000). Curr. Top. Immunol., 249, 35 ± 54. Microbiol. Immunol., 249, 75 ± 86. Herman JG, Gra JR, Myohanen S, Nelkin BD and Baylin Costello JF, Fruhwald MC, Smiraglia DJ, Rush LJ, SB. (1996a). Proc.Natl.Acad.Sci.USA,93, 9821 ± 9826. Robertson GP, Gao X, Wright FA, Feramisco JD, Herman JG, Jen J, Merlo A and Baylin SB. (1996b). Cancer Peltomaki P, Lang JC, Schuller DE, Yu L, Bloom®eld Res., 56, 722 ± 727. CD, Caligiuri MA, Yates A, Nishikawa R, Su Huang H, Herman JG, Merlo A, Mao L, Lapidus RG, Issa JP, Petrelli NJ, Zhang X, O'Dorisio MS, Held WA, Cavenee Davidson NE, Sidransky D and Baylin SB. (1995). Cancer WK and Plass C. (2000). Nat. Genet., 24, 132 ± 138. Res., 55, 4525 ± 4530. Costello JF and Plass C. (2001). J. Med. Genet., 38, 285 ± 303. IssaJP,VertinoPM,WuJ,SazawalS,CelanoP,NelkinBD, Cross SH, Charlton JA, Nan X and Bird AP. (1994). Nat. Hamilton SR and Baylin SB. (1993). J. Natl. Cancer Inst., Genet., 6, 236 ± 244. 85, 1235 ± 1240. Eden S, Hashimshony T, Keshet I, Cedar H and Thorne AW. Juttermann R, Li E and Jaenisch R. (1994). Proc. Natl. Acad. (1998). Nature, 394, 842. Sci. USA, 91, 11797 ± 11801. Elson DA and Bartolomei MS. (1997). Mol. Cell. Biol., 17, Jones PA and Laird PW. (1999). Nat. Genet., 21, 163 ± 167. 309 ± 317. Knudsen S. (1999). Bioinformatics, 15, 356 ± 361. Evron E, Umbricht CB, Korz D, Raman V, Loeb DM, Kumari D and Usdin K. (2000). J. Biol. Chem., 31, 31. Niranjan B, Buluwela L, Weitzman SA, Marks J and Li E, Beard C and Jaenisch R. (1993). Nature, 366, 362 ± 365. Sukumar S. (2001). Cancer Res., 61, 2782 ± 2787. Li E, Bestor TH and Jaenisch R. (1992). Cell, 69, 915 ± 926. Fuks F, Burgers WA, Brehm A, Hughes-Davies L and Loeb DM, Evron E, Patel CB, Sharma PM, Niranjan B, Kouzarides T. (2000). Nat. Genet., 24, 88 ± 91. Buluwela L, Weitzman SA, Korz D and Sukumar S. Gardiner-Garden M and Frommer M. (1987). J. Mol. Biol., (2001). Cancer Res., 61, 921 ± 925. 196, 261 ± 282. Nan X, Ng HH, Johnson CA, Laherty CD, Turner BM, Garrick D, Fiering S, Martin DI and Whitelaw E. (1998). Eisenman RN and Bird A. (1998). Nature, 393, 386 ± 389. Nat. Genet., 18, 56 ± 59. Ng HH and Bird A. (1999). Curr.Opin.Genet.Dev.,9, 158 ± Glenn CC, Saitoh S, Jong MT, Filbrandt MM, Surti U, 163. Driscoll DJ and Nicholls RD. (1996). Am.J.Hum.Genet., Ohler U, Stemmer G, Harbeck S and Niemann H. (2000). 58, 335 ± 346. Pac. Symp. Biocomput., 380 ± 391. Gonzalgo ML, Hayashida T, Bender CM, Pao MM, Tsai Okuda T, Hirai H, Valentine VA, Shurtle SA, Kidd VJ, YC, Gonzales FA, Nguyen HD, Nguyen TT and Jones Lahti JM, Sherr CJ and Downing JR. (1995). Genomics, PA. (1998). Cancer Res., 58, 1245 ± 1252. 29, 623 ± 630. Gra JR, Herman JG, Myohanen S, Baylin SB and Vertino Otterson GA, Khleif SN, Chen W, Coxon AB and Kaye FJ. PM. (1997). J. Biol. Chem., 272, 22322 ± 22329. (1995). Oncogene, 11, 1211 ± 1216.

Oncogene 5-Aza-CdR increases expression of unmethylated CDKN2D W-G Zhu et al 7796 Pfeifer GP, Tanguay RL, Steigerwald SD and Riggs AD. Soengas MS, Capodieci P, Polsky D, Mora J, Esteller M, (1990). Genes Dev., 4, 1277 ± 1287. Opitz-Araya X, McCombie R, Herman JG, Gerald WL, Prestridge DS. (1995). J. Mol. Biol., 249, 923 ± 932. Lazebnik YA, Cordon-Cardo C and Lowe SW. (2001). Robertson KD, Ait-Si-Ali S, Yokochi T, Wade PA, Jones PL Nature, 409, 207 ± 211. and Wole AP. (2000). Nat. Genet., 25, 338 ± 342. Staiger D, Kaulen H and Schell J. (1989). Proc. Natl. Acad. Robertson KD and Jones PA. (2000). Carcinogenesis, 21, Sci. USA, 86, 6930 ± 6934. 461 ± 467. Tate PH and Bird AP. (1993). Curr. Opin. Genet. Dev., 3, Robertson KD, Uzvolgyi E, Liang G, Talmadge C, Sumegi J, 226 ± 231. Gonzales FA and Jones PA. (1999). Nucleic Acids Res., 27, Tycko B. (2000). J. Clin. Invest., 105, 401 ± 407. 2291 ± 2298. Yoo-Warren H, Pachnis V, Ingram RS and Tilghman SM. Rountree MR, Bachman KE and Baylin SB. (2000). Nat. (1988). Mol. Cell. Biol., 8, 4707 ± 4715. Genet., 25, 269 ± 277. Zhu WG, Lakshmanan RR, Beal MD and Otterson GA. Soengas MS, Alarcon RM, Yoshida H, Giaccia AJ, Hakem (2001). Cancer Res., 61, 1327 ± 1333. R, Mak TW and Lowe SW. (1999). Science, 284, 156 ± 159.

Oncogene