Growth Delay of Human Pancreatic Cancer Cells by Methylase Inhibitor 5-Aza-2

Growth delay of human pancreatic cancer cells by methylase inhibitor 5-aza-20-deoxycytidine treatment is associated with activation of the interferon signalling pathway

Edoardo Missiaglia1,2, Massimo Donadelli3, Marta Palmieri3, Tatjana Crnogorac-Jurcevic1, Aldo Scarpa*,2 and Nicholas R Lemoine*,1

1Cancer Research UK, Molecular Oncology Unit, Imperial College School of Medicine at Hammersmith Campus, London, UK; 2Dipartimento di Patologia, Sezione di Anatomia Patologica, Universita` di Verona, Strada Le Grazie, I-37134, Verona, Italy; 3Dipartimento di Scienze Neurologiche e della Visione, Sezione di Biochimica, Universita` di Verona, Strada Le Grazie, I-37134, Verona, Italy

Alteration of methylation status has been recognized as a profile, with a correlation between methylation of CpG possible epigenetic mechanism of selection during tumor- dinucleotides and the progression of tumours to igenesis in pancreatic cancer. This type of cancer is malignancy (Costello et al., 2000; Adorjan et al., characterized by poor prognosis partly due to resistance to 2002). DNA methylation normally occurs at the conventional drug treatments. We have used microarray cytosine residues within CpG dinucleotides as a result technology to investigate the changes in global gene of DNA methyltransferase (DNA MTase) activity. expression observed after treatment of different pancrea- DNA MTase recognizes methylated CpG sites in the tic cancer cell lines with the methylase inhibitor 5-aza-20- parent strand and catalyses addition of a methyl group deoxycytidine (5-aza-CdR). We have observed that this to the daughter strand (Leonhardt et al., 1992; Yoder agent is able to inhibit to various degrees the growth of and Bestor, 1996). The presence in the eukaryotic three pancreatic cancer cell lines. In particular, this genome of these CpG sites has been progressively inhibition was associated with induction of interferon reduced during evolution, with the exception of CpG (IFN)-related genes, as observed in other tumour types. islands (Antequera and Bird, 1993; Bird, 1993). These Thus, expression of STAT1 seems to play a key role in the are frequently contained in the promoter regions or in cellular response to treatment with the cytosine analogue. the exons of different genes and are usually unmethy- Moreover, we found increased p21WAF1 and gadd45A lated in normal cells (Bird et al., 1985; Cross and Bird, expression to be associated with the efficacy of the 1995; De Smet et al., 1999). Hypermethylation of CpG treatment; this induction may correlate with activation of islands is associated with delayed DNA replication, the IFN signalling pathway. Expression of the p16INK condensed chromatin and inhibition of transcription protein was also linked to the ability of cells to respond to initiation (Antequera et al., 1990; Tazi and Bird, 1990; 5-aza-CdR. Finally, genome-wide demethylation induced Baylin et al., 1998; Jones and Wolffe, 1999), and it has sensitization that significantly increased response to been observed in oncogenic transformation (Antequera further treatment with various chemotherapy agents. et al., 1990), in cancer-derived cell lines (Rideout et al., Oncogene (2005) 24, 199–211. doi:10.1038/sj.onc.1208018 1994) and uncultured tumours (Baylin et al., 1998; Costello et al., 2000). Keywords: pancreatic cancer cell lines; microarrays; There are several growth-regulatory genes that have 5-aza-20-deoxycytidine been found to be silenced by hypermethylation in different tumour types, including, for example, RB1 (Sakai et al., 1991), VHL (Herman et al., 1994), cyclin kinase inhibitors (p15, p16) (Herman et al., 1995, 1996; Introduction Bender et al., 1998; Lubomierski et al., 2001), estrogen ONCOGENOMICS receptors (Ottaviano et al., 1994) and MLH1 (Kane Abnormal patterns of DNA methylation have been et al., 1997). recognized as a frequent molecular change involved in Pancreatic ductal adenocarcinoma is a lethal disease the initiation and progression of human neoplasia and represents the fifth most common cause of cancer (Jones and Buckley, 1990; Baylin et al., 1991; Costello death in the United States and Europe (Pisani et al., et al., 2000). The importance of these epigenetic 1999). This is mainly because it is inoperable at the time alterations has been highlighted in recent studies where of diagnosis and has a stinking resistance to conven- human cancers could be classified by their methylation tional treatments. The molecular basis of this aggressive clinical behaviour is as yet incompletely elucidated. *Correspondence: NR Lemoine; E-mail: [email protected] Alterations in methylation status have been recog- and A Scarpa; E-mail: [email protected] nized as an epigenetic mechanism of selection during Received 8 March 2004; revised 9 June 2004; accepted 6 July 2004 tumorigenesis in this type of cancer, and a correlation Growth delay of human pancreatic cancer cells E Missiaglia et al 200 between methylation patterns of pancreatic adenocarci- methyltransferase and inhibits its activity, leading to nomas and their clinico-pathological features has been genome-wide demethylation (Creusot et al., 1982; Santi suggested (Ueki et al., 2001). In addition, the hyper- et al., 1984; Michalowsky and Jones, 1987). Numerous methylation of multiple specific genes appears a possible studies have shown the ability of 5-aza-CdR to inhibit or way for the tumour to silence some key pathways (Ueki suppress the growth of human tumours in vitro and et al., 2000; Venkatasubbarao et al., 2001). The in vivo (Yamada et al., 1996; Bender et al., 1998; Eggert reactivation of these key genes using a demethylating et al., 2000; Suh et al., 2000; Karpf et al., 2001), while agent might sensitize tumours to the action of clinical trials are currently assessing the effectiveness of chemotherapy. 5-aza-CdR in the treatment of both haematological and The methylase inhibitor 5-Aza-20-deoxycytidine solid tumours (for reviews, see Santini et al., 2001; (5-aza-CdR) is a cytosine analogue that is incorporated Aparicio and Weber, 2002). into DNA during replication. It covalently binds DNA In the present study, we analysed the gene expression profile of three pancreatic cancer cell lines showing differential growth inhibition to treatment with 5-aza- CdR. The growth inhibition was associated with the induction of interferon (IFN)-related genes. Impor- tantly, the genome-wide demethylation significantly increased the response of PaCa44 and CFPAC1 cells to treatment with IFN-g, gemcitabine and cisplatin. This suggests a possible application of 5-aza-CdR in pancreatic cancer therapy in combination with other drugs.

Results

Growth inhibition, cell cycle proﬁle and apoptosis induced by 5-aza-CdR treatment Treatment with 5-aza-CdR induced growth inhibition that was more marked and lasted longer in PaCa44 and CFPAC1 than in MiaPaCa2 cells (Figure 1). At the sixth day, when the total RNAs for the gene expression proﬁle analysis were extracted, the percentage of cells for PaCa44 and CFPAC1 compared to the control were, respectively, 18 and 28%, while MiaPaCa2 was 57%. The cell cycle analysis by propidium iodide staining showed a delay in the progression with an increased proportion of cells in S and G2-M phases particularly for PaCa44 and CFPAC1, while the effect on MiaPaCa2 was less marked (Table 1). The level of apoptotic cells measured at the sixth day after treatment, by detection of histone-associated DNA in the cytoplasmic fraction, revealed that the treatment

Table 1 Cell cycle distribution: percentage of cells involved in the three phases of cell cycle as measured by propidium iodide staining in treated and control samples Treated cells (%) Controls (%)

Cell lines Days after G1/G0 S G2/M G1/G0 S G2/M treatment

PaCa44 3 9 56 35 59 33 8 613444372235 CFPAC1 3 35 20 45 73 19 8 649262585105 Figure 1 Growth curves of cell lines (a) MiaPaCa2, (b) PaCa44 MiaPaCa2 3 60 27 13 76 14 10 and (c) CFPAC1 treated with 5-aza-CdR (J) and their controls 663261077158 (’). The ratio on Y-axis was obtained by dividing the absorbance of treated or untreated cell lines by the mean absorbance of each The assay was carried out in triplicate for each cell line and time point cell line measured just before the treatment (time 0). Each point and performed the third and sixth days after treatment with 5-Aza- represents the mean ratio of eight replicates CdR

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 201 with 5-aza-CdR induced only modest levels of apoptosis. Application of Annexin V staining conﬁrmed the low levels of apoptosis induced (data not shown). This suggests that the low number of cells observed in treated samples was mostly an effect of growth inhibition.

Methylation status and DNMTs expression The content of 5-methylcytosine (5mC) in the genomic DNA of control and treated cells was semiquantitatively measured using the McrBC enzyme. Figure 2 shows that treatment with 5-aza-CdR was able to induce general- ized demethylation in the genomic DNA of all the three Figure 3 RT–PCR analysis of gene expression in control and cell lines. treated cell lines 6 days after treatment with 5-aza-CdR. (a) RT– The expression status of the four main methyltrans- PCR of methyltransferases genes. (b) RT–PCR of genes involved in S-G2 transition. b-Actin was used as the endogenous control ferase genes assayed by RT–PCR assay is shown in Figure 3a. DNMT1, which is the principal target of the cytosine analogues, showed a lower level of expression after treatment in PaCa44 and CFPAC1, while a small increase occurred in MiaPaCa2, conﬁrming the data obtained by microarray analysis (Supplementary Table 1). The observed expression levels of DNMT1 may not be due to a direct effect of 5-aza-CdR, but could be explained by its tight regulation during the cell cycle, with lower levels in G1/G0 and peak levels in S phase (Robertson et al., 2000). In fact, 5-aza-CdR affects the enzyme activity rather than its expression, which is also in agreement with our data when we observed a minor upregulation of DNMT1 (Figure 3a) associated with genomic DNA demethylation in the MiaPaCa2 cell line (Figure 4). Another interesting aspect is the presence of different splicing variants of DNMT3b in the cell lines Figure 4 Dendrogram obtained by hierarchical cluster analysis using a wide gene set (5306 features) of control and treated samples analysed. In particular, MiaPaCa2 shows a higher level at the 6th day after treatment. The analysis was performed using of expression of splice variant 3, which is either not the ‘mva’ package applying an agglomerative method, average expressed or expressed at only a very low level in the linkage and Euclidian distance. The dendrogram is cut by a other two cell lines. horizontal line to demonstrate that all the treated CFPAC1 and PaCa44 samples fall in the same cluster, while treated MiaPaCa2 samples are closest to the cluster of control samples

The global gene expression profile shows similarities between PaCa44 and CFPAC1 and a different pattern for MiaPaCa2 In order to identify similarities in the effect of 5-aza- CdR treatment on gene expression profile among the cell lines, we first compared the profile of control and treated cells using the Spearman correlation method. The coefficients were calculated comparing the log ratios obtained from each control and treated cell line against their appropriate reference (see the experimental design in methods). The results revealed a high correlation between PaCa44 and CFPAC1 (r-value ¼ 0.66; Po0.0001), while neither of these showed such a high Figure 2 Effect of 5-aza-CdR treatment on genomic DNA correlation with MiaPaca2 (r-value ¼ 0.32 and 0.36, methylation for each cell line. Genomic DNA was digested with respectively; Po0.0001)). In addition, the differences the McrBC enzyme at different time points after treatment and between these correlations were statistically significant compared to the controls. The enzyme cleaves DNA-containing (both Po0.001). The same result was obtained using methylcytosine on one or both strands. The increased amounts of fragments of larger size observed 3 and 6 days after the treatment Euclidian distance based on wide gene sets (5306 prove the ability of 5-aza-CdR to reduce the content of features), as visualized by the hierarchical clustering methylcytosine in genomic DNA reported in Figure 4. Cutting the dendrogram as

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 202 illustrated shows that the 5-aza-CdR-treated PaCa44 MAPPFinder, although there is its parent term ‘oxygen and CFPAC1 samples fall in a common cluster, while and reactive oxygen species metabolism’). Among the 5-aza-CdR-treated MiaPaCa2 replicates clearly fall downregulated genes, ‘cell cycle’ was the most repre- in a different cluster closer to the untreated samples of sented term (at 52.4%) and this had a highly significant all lines. These observations are in agreement with the adjusted P-value (Po0.0001), while ‘DNA metabolism’ effect of 5-aza-CdR treatment on the growth character- and ‘cytokinesis’ showed the same level of significance istics of these cell lines, which was similar between but lower percentage representation (32 and 11%, PaCa44 and CFPAC1 and less pronounced in MiaPa- respectively). Other terms significant only at the Ca2. This suggests the possibility that the treatment may unadjusted P-value were: ‘nucleobase biosynthesis’, have activated or silenced pathways common to the ‘nucleobase metabolism’, ‘defence response’, ‘RNA former two cell lines and different from those affected in metabolism’, ‘amino-acid metabolism’, ‘nucleotide me- MiaPaca2. tabolism’, ‘cell surface receptor-linked signal transport’, ‘ion transport’ and ‘carboxylic acid metabolism’. As Genes induced or repressed by 5-aza-CdR treatment observed for the upregulated set, these terms were in the cell lines among those found significantly over-represented also with MAPPFinder (see Supplementary Table 4). To select differentially expressed genes, we used the Furthermore, several terms were specifically related to statistical package SAM, with stringency and conditions DNA replication, mitosis and cell cycle regulation, described in Materials and methods. This highlighted suggesting an inhibition in the cell cycle progression. 555 up- and 590 downregulated features in PaCa44, 395 In Supplementary Table 1, we list the genes known to up- and 357 downregulated features in CFPAC1 and be associated with the activation of the IFN signalling 118 up- and four downregulated features in MiaPaCa2. pathway, whose expression was found upregulated after Since both PaCa44 and CFPAC1 showed similar treatment in the cell lines (43 out of 127 genes, which growth and gene expression patterns after treatment represents 34% of all the upregulated genes). Among with 5-aza-CdR, it is possible that a common set of these, the interferon gamma (IFNg) receptor 1 genes is involved in their more marked growth inhibi- (IFNGR1) and the Signal Transducer and Activator of tion. In order to identify such a common set, the lists of Transcription 1 (STAT1) were upregulated only in differentially expressed features of PaCa44 and PaCa44- and CFPAC1-treated cells. CFPAC1 were superimposed. This analysis identified 127 unique genes (175 clones) upregulated and 123 Western blot analysis of STAT1, p21WAF1 and p16INK4 unique genes (162 clones) downregulated in both PaCa44 and CFPAC1 (for the full lists see Supplemen- To investigate further the involvement of the IFN tary Tables 1 and 2). pathway in the response to 5-aza-CdR treatment, we We then analysed the biological process terms focused our attention on STAT1 and p21WAF1 (Karpf associated with the differentially expressed genes using et al., 1999; Liang et al., 2002). Western blot analysis the FatiGO Data mining and MAPPFinder programs showed that the expression of STAT1 protein and its (see Gene ontology analysis in Materials and methods). phosphorylated form (Figure 5) paralleled the increased The total number of upregulated genes that show expression of STAT1 mRNA observed by microarray biological process terms in GO database at the fifth analysis after treatment with 5-aza-CdR in PaCa44 and level using FatiGO was 77 (out of 117 genes identified CFPAC1 cell lines, while in MiaPaca2 STAT1 was with Ensembl-ID), while the number of genes down- expressed at a very low level under all conditions. regulated was 79 (out of 113 genes identified with We also analysed the expression of p21WAF1 by Ensembl-ID). Extending the analysis with the MAPP- Western blot, as its promoter region contain the Finder program to higher-level nodes brought the figure palindromic sequences recognized by STAT1 (Chin to 112 and 106 genes, respectively, up- or downregulated et al., 1996). In addition, it has been shown that there and linked to a GO term. is a correlation between hypermethylation of the STAT- In the upregulated data set, the only gene ontology binding site (sis-inducible element (SIE)-1) in the term statistically significant at the adjusted level with p21WAF1 promoter region and decreased protein expres- FatiGO was ‘defence response’ (adjusted Po0.05), sion (Chen et al., 2000). The results shown in Figures 5 which was shared by 22% of all genes, while terms such and 6 are concordant with this hypothesis, confirming as ‘response to oxidative stress’ (6.25%, P ¼ 0.001), the hypermethylation of SIE-1 in all three cell lines, ‘response to pest/pathogen/parasite’ (12.5%, P ¼ 0.006) which is reverted by 5-aza-CdR. However, we observed and ‘superoxide metabolism’ (3.1%, P ¼ 0.02) were an increased expression of p21WAF1 only in PaCa44 and significant only at the unadjusted level. These terms CFPAC1 cells, where STAT1 was expressed and were usually confirmed to be over-represented also by upregulated after treatment. On the other hand, the the analysis with MAPPFinder, even if extended to a lack of STAT1 expression in MiaPaCa2 is associated wider set of gene ontology terms (see Supplementary with no change in p21WAF1 mRNA and protein level. Table 3). Indeed, several terms are related through the Inactivation of the p16INK4 gene is a frequent event in hierarchical relationship among them (for example pancreatic cancer, due to either mutation and/or ‘superoxide metabolism’, that was found over-repre- deletion of the gene locus, or methylation of its sented by FatiGO, is not in the list selected by promoter region responsible for expression silencing

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 203

Figure 7 MSP analysis of 14-3-3s and RARRES1 genes. The treated samples (T) were analysed after 6 days from the 5-aza-CdR exposure and compared to the untreated samples (C). The lanes U and M represent the ampliﬁcation products from unmethylated and methylated templates, respectively

Figure 5 Western blot analysis of three gene products (Stat1, p21 and p16) found differentially expressed 6 days after treatment with 5-aza-CdR reduced by the treatment in both PaCa44 and CFPAC1 cells, while 14-3-3s and PCNA were downregulated only in CFPAC1. Expression of the 14-3-3s gene in MiaPaCa2 cells is known to be silenced by methylation and its re-expression after 5-aza-CdR treatment has already been described (Iacobuzio-Donahue et al., 2003).

Methylation status of genes found re-expressed after Figure 6 Methylation status of the SIE-1 element in p21WAF1 5-aza-CdR promoter region by genomic DNA digestion with methyl-sensitive restriction enzyme HpaII, followed by a PCR that spans the SIE-1 Among the genes induced after 5-aza-CdR treatment, element. Amplification of the PCR product indicates the presence we selected two genes known to be frequently methy- of methylcytosine in the SIE-1 CpG site, which block the enzyme lated, 14-3-3s and RARRES1. The former was only cleavage. 5-aza-CdR treatment induces the methyl group removal upregulated in MiaPaCa2 cells, while the latter was so that HpaII disrupts the target template for the PCR reaction. c- Kit was used as positive control upregulated in both PaCa44 and CFPAC1 cells. The results are shown in Figure 7. The MSP analysis revealed that only MiaPaCa2 was originally fully methylated in the CpG island within the first 14-3-3s (Herman et al., 1995). As p16INK4 was not represented on exon, and 5-aza-CdR treatment partially reverted that our cDNA microarray and its reactivation has been status. This is in agreement with both our RT–PCR associated with the inhibition of tumour growth induced results and reported data (Iacobuzio-Donahue et al., in different cell lines by 5-aza-CdR treatment (Bender 2003). RARRES1 showed a partial methylation in its et al., 1998; Suh et al., 2000), we investigated its promoter region in all three cell lines, although to a expression using Western blot analysis. The status of different degree. However, the treatment affected mostly p16INK4 in different pancreatic cancer cell lines has MiaPaCa2 and PaCa44, as evident by the decreased already been assayed showing promoter hypermethyla- intensity of the signal for the methylated fragment in tion in PaCa44 and CFPAC1, while MiaPaCa2 has these two cell lines. p16INK4 homozygous deletion (Moore et al., 2001). As expected, we found upregulated protein expression only IFN pathway-related genes in PaCa44 and CFPAC1, supporting the successful demethylation of the p16INK4 promoter (Figure 5) as well The results obtained using quantitative RT–PCR con- as its possible involvement in the efficacy of the 5-aza- firmed the differences observed using microarray analy- CdR treatment on cell proliferation. sis for seven genes belonging to the IFN pathways (Table 2). In particular, all seven genes were shown to be Other genes differentially expressed involved in cell differentially expressed in treated PaCa44 and CFPAC1, cycle progression as expected. However, statistically significant differences were also identified in MiaPaCa2 cells for the gadd45A, We examined the expression levels of CDC2, cyclin B1, ISG20 and ubiquitin D genes. For the first two, the fold PCNA and 14-3-3s by RT–PCR to verify their inductions observed in MiaPaCa2 were always lower behaviour after treatment with 5-aza-CdR. As reported than those in PaCa44 and CFPAC1 and close to the in Figure 3b, the expression of CDC2 and cyclin B1 was threshold level used as cutoff in the microarray analysis,

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 204 Table 2 Conﬁrmation of seven genes, whose expression was found differentially expressed by microarray analysis, using quantitative real-time RT–PCR Gene Symbol Cell line Median CT Median CT Difference Fold-diff. P-value contol treated

MiaPaCa2 0 0.07 0.07 1.05 – 0.47 IFNGR1 CFPAC1 1.62 2.96 1.34 2.53 m 0.005 PaCa44 3.52 5.26 1.74 3.34 m 0.005

MiaPaCa2 0 0.55 0.55 1.46 – 0.002 GADD45A CFPAC1 1.08 2.58 1.5 2.83 m 0.005 PaCa44 0.18 2.62 2.44 5.43 m 0.002

MiaPaCa2 0 4.32 4.32 19.97 m 0.005 UBD CFPAC1 9.21 13.43 4.22 18.64 m 0.002 PaCa44 12.77 17.78 5.01 32.22 m 0.005

MiaPaCa2 1.07 0.83 -0.24 0.85 – 0.06 CCNA2 CFPAC1 0.86 0 -0.86 0.55 . 0.002 PaCa44 2.36 0.94 -1.42 0.37 . 0.002

MiaPaCa2 0.32 0.00 -0.32 0.80 – 0.15 Mx2 CFPAC1 0.51 5.79 5.28 38.85 m 0.002 PaCa44 2.87 5.96 3.09 8.51 m 0.002

MiaPaCa2 0.00 0.12 0.12 1.08 – 0.31 G1P2 CFPAC1 1.50 4.28 2.78 6.87 m 0.002 PaCa44 1.45 4.31 2.86 7.26 m 0.002

MiaPaCa2 0.00 1.10 1.10 2.14 m 0.002 ISG20 CFPAC1 2.52 4.74 2.22 4.66 m 0.002 PaCa44 2.79 5.55 2.76 6.77 m 0.002

MiaPaCa2 17.37 17.29 0.17 1.12 – 0.1 ACTB CFPAC1 18.35 18.00 0.35 1.27 – 0.06 PaCa44 18.73 18.60 0.13 1.09 – 0.12

In the table the median number of cycles normalized and scaled for controls and treated samples is reported. The differences were evaluated using Mann–Whitney test

while ubiquitin D was expressed at a very low level (more than 1 million times less expressed) compared to b-actin.

Combined treatment with 5-aza-CdR and other anticancer drugs To verify the ability of 5-aza-CdR to sensitize the cell lines to further treatment, we combined this agent with other drugs, some of which are frequently used for pancreatic cancer chemotherapy. These included gemcitabine, 5-fluorouracil (5-FU), cisplatin, IFNg and tumour necrosis factor alpha (TNFa). As expected from the molecular response observed, the combination of 5- aza-CdR and IFNg (100 U/ml) produced an additive effect, with percentages of cells of 39 and 37% in Figure 8 Percentage of cells compared to the control observed at PaCa44 and CFPAC1 cell lines, respectively, compared 72 h after the single and combined treatments with 5-aza-CdR (2 mM) and IFNg (100 U/ml). The reductions in cell percentage to untreated controls, which were significantly lower observed after the combined treatment were always significant in than 50% (P ¼ 0.003) and 55% (P ¼ 0.05) observed with both PaCa44 and CFPAC1 cell lines, if compared to each single 5-aza-CdR alone, or 66% (P ¼ 0.0003) and 63% treatment with 5-aza-CdR (P ¼ 0.003 and 0.05, respectively) and (P ¼ 0.03) with IFNg alone. Conversely, the value for IFNg (P ¼ 0.0003 and 0.03, respectively), while such comparisons MiaPaCa2 cells was only 69%, which is similar to the were not significantly different in the MiaPaCa2 cell line effect of IFNg alone (Figure 8). Positive effects were also observed for a combination of 5-aza-CdR with cisplatin and gemcitabine. In particular, we observed a significant 5-aza-CdR alone; P ¼ 0.04) and CFPAC1 (22% com- inhibition effect with cisplatin associated with 5-aza- pared to the 36% when Cisplatin was used alone; CdR in PaCa44 (38% compared to 50% obtained with P ¼ 0.0002), while Gemcitabine was more effective in

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 205 expression profile induced by 5-aza-CdR treatment in cell models. Analysis of cell growth and cell cycle distribution revealed differential responses of individual cell lines to treatment with this drug. We selected the relatively unresponsive MiaPaCa2 cell line, as well as the PaCa44 and CFPAC1 cell lines, which show strong inhibition of cell proliferation associated with arrest or delay in S or G2/M phases. Low rates of induction of apoptosis confirmed that 5-aza-CdR treatment produces a pre- dominantly cytostatic effect in these pancreatic cancer cell lines, as observed for other cancer cell lines (Davidson et al., 1992; Caraglia et al., 1994; Karpf et al., 1999; Sato et al., 2003a). We confirmed the effect of 5-aza-CdR on methylation status in all the treated cell lines. The level of DNMT1, Figure 9 Percentage of cells observed at 72 h after the single and assayed by RT–PCR, was downregulated in response to combined treatments with 5-aza-CdR compared to control the treatment only in PaCa44 and CFPAC1 cell lines observed at 72 h after the single and combined treatments with 5- aza-CdR (2 mM) and TNFa (3 ng/ml), gemcitabine (PaCa44: 64 mM, (confirmed by microarray analysis). Among the other CFPAC1: 15 nM, MiaPaCa2: 0.25 mM), cisplatin (PaCa44: 20 mM, methyltransferases, DNMT3b showed a different profile CFPAC1: 10 mM, MiaPaCa2: 4,5 mM) and 5-fluorouracil (PaCa44: of splicing variants in the various cell lines. These 32 mM, CFPAC1: 64 mM, MiaPaCa2: 16 mM). The comparisons variants may have alternative catalytic activity and made between the strongest single treatment against the combined treatment showed that the combination of 5-aza-CdR with different target specificity (Robertson et al., 1999, 2000; Cisplatin reduced significantly the percentage of cell compared to Saito et al., 2002) and may be responsible for a different PaCa44 treated with only 5-aza-CdR (P ¼ 0.04) and CFPAC1 DNA methylation pattern (Beaulieu et al., 2002). treated only with Cisplatin (P ¼ 0.0002). Equally, the combined Recently, Rhee et al. (2002) have demonstrated that treatment with Gemcitabine and 5-aza-CdR in PaCa44 was DNMT3b play an essential role in the maintenance of stronger than the treatment with Gemcitabine alone (P ¼ 0.002). (n) Significantly inhibited by combined treatment DNA methylation and gene silencing in cooperation with DNMT1. In particular, DNMT3b3 variants allow de novo methylation of various target sequences (Weisenberger et al., 2004) and their continued expres- combination with 5-aza-CdR only in PaCa44 (28% sion in MiaPaCa2 cells after exposure to 5-aza-CdR compared to the 56% with Gemcitabine alone; may explain the different effect on gene expression P ¼ 0.002) (Figure 9). profile in this cell line. In fact, the analysis of a few specific methylation sites gave evidence for their frequent hypermethylation particularly in the MiaPa- Discussion Ca2 cell line, which confirms the results of other studies where this cell line was among those with a higher Previous studies have proven convincingly the role of number of hypermethylated genes (Ueki et al., 2000; aberrant methylation in the initiation and progression of Sato et al., 2003a, c). Thus, we may speculate that several tumour types, including pancreatic adenocarci- the variability in the methylation status observed noma. In fact, methylation of CpG islands or sites is among the three cell lines under study could partially frequently associated with silencing of tumour-suppres- explain the difference in the response to 5-aza-CdR sor genes that may represent a selective advantage in treatment. terms of growth and survival for the affected cells. The tumour suppressor genes p16INK4 and p53, which Pancreatic adenocarcinoma is characterized clinically are critically involved in pancreatic cancer tumorigen- by a wide resistance to all current chemotherapy esis, are also implicated in the response to 5-aza-CdR treatments. The aim of this study was to assess the treatment (Bender et al., 1998; Suh et al., 2000; Nieto efficacy of 5-aza-CdR treatment in inhibiting cell growth et al., 2004; Zhu et al., 2004). The analysis of the or inducing sensitization to further drugs in pancreatic expression of p16 protein confirmed its upregulation cancer cell models. Several clinical trials of 5-aza-CdR after the treatment in both PaCa44 and CFPAC1. administration conducted in diverse tumour types have However, the S and G2/M delay observed, together with shown various degrees of efficacy, particularly in solid results of previous studies that showed a p16INK4- tumours (for reviews, see Santini et al., 2001; Aparicio independent growth inhibition of tumour cells induced and Weber, 2002). On the other hand, a recent study has by 5-aza-CdR (Bender et al., 1998), suggests that its suggested that 5-aza-CdR treatment may stimulate effect is not limited to the re-expression of p16INK4. invasion in some pancreatic cancer cell lines by The analysis of global gene expression profiles further reactivation of invasion-promoting genes (Sato et al., corroborates the observation of a different biological 2003b). In the present study, we sought to identify the behaviour for the MiaPaCa2 cell line in comparison to molecular mechanisms behind differential effects in the other two. Starting from the similarity between the pancreatic cancer by analysing the alteration in gene PaCa44 and CFPAC1 gene expression profiles, we

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 206 superimposed their differentially expressed genes to at both the transcriptional and protein level only in identify common pathways involved in the biological PaCa44 and CFPAC1 cells, where it was associated with response induced by 5-aza-CdR treatment. Gene ontol- STAT1 expression. Although the p21WAF1 protein is ogy analysis of common downregulated genes demon- frequently responsible for G1 arrest, numerous pieces of strated that ‘cell cycle’, ‘mitosis’, ‘DNA metabolism’ and evidence suggest a role also in S/G2 transition. p21WAF1 ‘cytokinesis’ terms were over-represented, in agreement is thought to induce G2 arrest by interacting with the with the delay in cell cycle progression observed. Cdc2–cyclinB1 complex together with PCNA, and thus Conversely, upregulated genes showed over-represented prevents incorporation of Cdc25C that can activate the terms such as ‘defence response’, ‘response to pest/ Cdc2–cyclinB1 complex by dephosphorylation (Ando pathogen/parasite’ and ‘superoxide metabolism’. Inter- et al., 2001). However, inhibition of S/G2 transition may estingly, several upregulated genes were found to belong also involve interaction of p21WAF1 with the Cdc2– to the IFN pathway and IFN cytokines may modulate cyclinA complex (Dulic et al., 1998). Furthermore, immune responses and exert antiproliferative effects in p21WAF1 can inhibit the interaction between PCNA and some cell types by inhibiting DNA replication and the DNA polymerase delta, which is essential for DNA protein synthesis (Clemens and McNurlan, 1985). This replication and repair, and consequently inhibit DNA may explain the frequent occurrence of terms such as synthesis (Rousseau et al., 1999; Lu et al., 2002). ‘DNA metabolism’, ‘RNA metabolism’, ‘amino-acid Downregulation of DNMT1, which competes with metabolism’, ‘nucleobase biosynthesis’ and ‘nucleobase p21WAF1 for the same binding site on PCNA (Chuang metabolism’ for our downregulated genes. et al., 1997; Fang and Lu, 2002), may further enhance Previous studies on the effects of 5-aza-CdR on gene the inhibitory effect of 21WAF1on both DNA synthesis expression profile in colon and bladder cancer cell lines and cell growth. have already described the induction of IFN-responsive Gadd45A, that includes three CpG-rich regions genes as the major cellular response to this treatment, among its exons and introns (data not shown), is suggesting their involvement in the observed growth another key gene found upregulated in 5-aza-CdR- inhibition (Karpf et al., 1999; Liang et al., 2002). treated cells, whose expression can be increased by IFNg Furthermore, inhibition of DNMT1 activity, using a treatment (Brysk et al., 1995) as well as by p53 activity conditional mutation in a mouse model, has shown a (Carrier et al., 1994). It has been associated with cell similar response at the gene expression profile level, growth suppression and G2/M cell cycle arrest by indicating that the activation of this pathway is altering cyclin B1 nuclear localization and inhibiting associated with the specific effect of 5-aza-CdR on Cdc2 activity in a p53-dependent manner (Jin et al., DNMT1 activity rather than a secondary cytotoxic 2002). However, Hirose et al. (2003) have observed outcome (Jackson-Grusby et al., 2001). In addition, in a induction of G2/M arrest caused by Gadd45A upregu- recent study, the epigenetic silencing of multiple genes lation after chromatin structure reorganization by that belong to the IFN pathway has been shown to be treatment with histone deacetylase inhibitors without involved in cellular immortalization, which was then the need for functioning p53. reverted by treatment with 5-aza-CdR, suggesting that Hence, we speculate that upregulation of p21WAF1 and this alteration may be associated with an early event in Gadd45A may be responsible for the cell cycle arrest or tumorigenesis (Kulaeva et al., 2003). delay observed after treatment with 5-aza-CdR by Our study produced analogous results, emphasising altering the formation and activation of Cdc2–CyclinB1 the role of IFN pathway activation also in the complex. Moreover, the p53 mutations present in all cell pancreatic cancer cell model. Specifically, the treatment lines under study (Moore et al., 2001) may inhibit increased the expression of IFNGR1 and STAT1 genes, apoptotic pathways, explaining the low apoptotic rate as well as others known to be induced by IFNs, such as induced by the treatment. We may also speculate that Mx2, G1P2, ISG20 and UBD. IFNGR1 is part of a downregulation of several genes associated with the S complex that mediates the activation of the transcription and G2/M phases of the cell cycle, found with the factor STAT1a by tyrosine phosphorylation on Tyr701 microarray analysis, is due to this cell cycle delay rather after its interaction with the IFNg cytokine (Shuai et al., than a direct effect of the drug treatment. In fact, we 1993). The STAT1 protein is a transcription factor (for a may expect that the differential expression observed for review, see Ramana et al., 2000) that mediates the several genes may be independent of their own antiproliferative response to interferon stimulation methylation status. (Bromberg et al., 1996). This negative effect on cell In agreement with a recent report (Sato et al., 2003b), proliferation is coupled particularly with p21WAF1 upre- the gene expression profile analysis also demonstrated gulation (Chin et al., 1996; Hobeika et al., 1998; that treatment with 5-aza-CdR induces the expression of Kominsky et al., 1998; Chen et al., 2000), but down- several genes associated with tumour invasion and regulation of CCNA2 (cyclin A) (Sibinga et al., 1999), or metastasis, such as CTSB, DAF, MMP15, t-PA and u- cyclin D and Cdc25A (Tiefenbrun et al., 1996), has also PAR, as well as the apoptotic suppressor BIRC3. been described. However, we found upregulation of other genes that In our study, we observed hypermethylation of the may inhibit matrix degradation (TIMP1 and PAI-1)or SIE-1 site in the p21WAF promoter region in all three cell stimulate the immune response (HLA-E, LY6E or lines, which was partially reverted by 5-aza-CdR SSX4). Further studies are needed in order to evaluate treatment and coupled with increased p21WAF expression if the balance between these positive and negative effects

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 207 results in an enhancement of invasion and metastasis Cell cycle profile in vivo. Cell DNA content was measured in triplicate at 3-day intervals Interestingly, the treatment induced or inhibited the up to 9 days under the same conditions. Cells were plated at expression of several genes found, respectively, down- or 2 Â 105/well in six-well culture plates (Orange Scientific), and upregulated in primary pancreatic cancer in other washed and treated with hypotonic fluorochrome solution at studies (see Supplementary Tables 3 and 4). In appropriate intervals (0.1% sodium citrate, 0.1% Triton X- particular, among the induced genes, DUSP6 has been 100, 20 mg/ml RNase and 50 mg/ml propidium iodide) (Sigma). suggested to exert tumour-suppressive effects in vitro in The cells were incubated for 30 min at room temperature in the pancreatic cancer by inhibiting phosphorylation of ERK dark before analysis. The fluorescence intensity of propidium (Furukawa et al., 2003). Among the downregulated set, iodide was measured by a FACSCAN flow cytometry (Becton Dickinson). Doublets were discriminated by using a peak we also found the gene CAD, described as upregulated fluorescence gate. Cell cycle distribution was determined by in intraductal papillary mucinous neoplasms of the Modfit software (Home Verity). pancreas (Sato et al., 2004), whose activity seem to be directly regulated by ERK2. This gene catalyses the Apoptosis initial, rate-limiting step in the de novo synthesis of pyrimidine nucleotides (Jones, 1980) and its inhibition Apoptosis was assessed by detection of histone-associated may potentially explain the downregulation of several DNA fragments 6 days after 5-aza-CdR treatment. Cells were Plus genes related to nucleotide synthesis (such as TK1, analysed using the Cell-Death Detection ELISA assay TYMS, GMPS, DTYMK) observed after treatment (Roche Molecular Biochemicals). The assay is based on a quantitative sandwich-enzyme-immunoassay principle using with 5-aza-CdR. anti-histone-biotin and anti-DNA-peroxidase (POD) mono- Finally, we combined 5-aza-CdR treatment with other clonal antibodies. POD was determined photometrically 0 chemotherapy agents in order to verify if the induction (A405 nm–A492 nm) with 2,2 azino-bis(3-ethylbenzthiazoline)-6- of genome-wide demethylation led to chemosensitiza- sulfonic acid (ABTS) as substrate. This allows the determina- tion. We observed that IFNg treatment gave an additive tion of the apoptotic enrichment of nucleosomes in the effect in combination with methyltransferase inhibitor cytoplasmic fraction of cell lysates. where its pathway was upregulated, confirming that its activation represents an important mechanism for the Combined treatment biological alterations induced by 5-aza-CdR treatment. Cells (4 Â 103/well) were plated in 96-well culture plates (Nunc Similar interactions have been described between 5-aza- Inc.) and treated after 24 h with 2 mM 5-Aza-CdR. A second CdR and IFN-a in haematological tumour cell lines treatment was added using the following drugs and conditions: (Dore and Momparler, 1992; Reid et al., 1992). 100 U/ml of IFNg (Sigma) maintained for 72 h, 3 ng/ml of TNFa Combination of 5-aza-CdR treatment with gemcitabine (Sigma) maintained for 72 h; cisplatin (Pharmacia and Upjohn or cisplatin also showed more marked inhibitory effects S.p.A.), gemcitabine (Eli Lilly Italia S.p.A.) and 5-FU (Teva than the single treatments alone. Hence, combined Pharma Italia S.r.l.) were added at approximately IC50 treatment with 5-aza-CdR and drugs commonly used concentrations (see Figures 8 and 9 for details) after 24 h for pancreatic cancer chemotherapy could improve their and maintained for the last 48 h. The level of growth inhibition effect in inhibiting tumour growth. was assessed at 72 h as described above. In order to provide evidence for the possible positive effect of drug association, we tested the difference between the strongest single treatment against the combined treatment using the Welch two-sample t- Materials and methods test, considering a P-valueo0.05 as significant.

Cell lines 5mC content in genomic DNA MiaPaca2, PaCa44 (both originating from primary tumours) Genomic DNA was obtained from treated cell lines at the third and CFPAC1 (originating from liver metastasis) adenocarci- and sixth days by digestion with proteinase-K (Sigma) and noma cells were cultured at 371Cin5%CO2 by using RPMI RNase-A (Promega), followed by phenol–chloroform extrac- 1640 medium (GIBCO) supplemented with 10% foetal bovine tion. In all, 1 mg of DNA from each sample was then digested serum (FCS) (GIBCO), 1% gentamicin (Gentalyn 80 – with McrBC enzyme (New England Biolabs), which cleaves Schering-Plough Spa) and 1% L-glutamine (GIBCO). The DNA-containing methylcytosine on one or both strands, and medium was renewed every 2 or 3 days. When the cells became resolved on a 1.5% agarose gel. conﬂuent, they were removed from the bottom of the culture

flasks by 0.5% porcine trypsin with 0.2% EDTA in Dulbecco’s WAF1 phosphate-buffered saline (PBS). Methylation status of the SIE-1 element in p21 promoter region We tested the presence of methylcytosine in the CpG site inside Cell growth assay the sis-inducible element (SIE)-1 (a STAT-responsive element Cells (104/well) were plated in 48-well culture plates (Nunc represented by the sequence 50-CTTCCCGGAAG-30)atnt Inc.) and treated after 24 h with a single dose of 2 Â 10À6 M 5- À692 in the promoter region of p21WAF1 (Chen et al., 2000) Aza-CdR (Sigma) that was left in the medium for 24 h and using PCR associated with the methylation-sensitive restric- then removed when a new drug-free medium was added to the tion enzyme HpaII. In all, 2 mg of genomic DNA was digested plates. The cell mass was assayed by crystal violet staining at 2- overnight with 40 U of HpaII (New England Biolabs, Beverly, day intervals up to 10 days. Control cells were analysed under MA, USA) at 371C. Then, 1 ml of digested product was PCR the same conditions until they reached confluence. amplified using the following primers: p21F 50-GAGTG-

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 208 TAGGGTGTAGGGAGATTG-30 and p21R 50-TCTGGCA- Image and data analysis GGCAAGGATTTAC-30, for 25 cycles of 941C for 45 s, 601C The scanned images were analysed using the ImaGene v4.2t for 40 s and 721C for 60 s, and a final extension of 5 min at software (BioDiscovery, Los Angeles, CA, USA). Quality 721C. Exon 8 of KIT gene was amplified as positive control measures were used to identify negative (background4signal), using the following primers: KIT8F 50-GTTTTCCAG- poor ((mean signalÀmean background)/SD 40.45) and CACTCTGACATATGG-30 and KIT8R 50-CTCTGCTCA- signal empty spots (signal/SD 42.3). Only the spots that were GTTCCTGGACAAA-30. Cycling conditions were 941C for signal not flagged and above background (total signal 3 min, followed by 35 cycles of 941C for 20 s, 551C for 25 s and 41.4 Â background in both channels) were retained. The 721C for 20 s. PCR products were subjected to electrophoresis mean signal subtracted for the background was used in the on 2% agarose gel and stained with ethidium bromide. subsequent analysis. We used the robust, locally linear lowess function to perform a within-print tip group normalization Methylation-specific PCR (MSP) followed by a scale normalization (Yang et al., 2002) using In all, 2 mg of DNA was BamHI digested and treated with marrayNorm package in R (Ihaka and Gentleman, 1996). The sodium bisulfite as described (www.mcdb.ucla.edu/Research/ log-square ratios of the normalized values from treated and Jacobsen/BisulfiteGenomicSequencing.html), with slight mod- pooled samples were compared to the ratios from control ification consisting in the addition of urea to the bisulphite samples and pool using SAM (Tusher et al., 2001). Missing reaction (Paulin et al., 1998). Then, 1 ml of bisulfite-treated data were integrated by the k-nearest neighbour algorithm DNA was the template for PCR using primers and the only when a single value per gene was missing. The input condition previously described for 14-3-3s (Ferguson et al., criteria were based on fold change greater than 1.5 between 2000) and RARRES1 (Youssef et al., 2004) genes. treated and control cells and a delta value expected to originate a data set with less than 2% of false positives (using the median of the number of falsely called genes) in order to 5-Aza-CdR treatments for gene expression profiles reduce the error due to multiple testing. The overall correlation Cells were exposed to 2 Â 10À6 M 5-Aza-CdR 24 h after passage between samples was obtained using the ‘cor’ function and in complete culture medium. At 24 h after drug addition, the Spearman correlation as distance metric in the R program, culture medium was replaced with drug-free medium. Each cell while the statistical significance of the differences between line was treated in triplicate, in parallel with a triplicate treated samples were investigated by the Mann–Whitney test. control. The culture density was set to reach 80% of A cluster of samples was generated using the ‘cluster’ and confluence at 6 days after treatment, as well as in the control ‘mva’ packages applying an agglomerative method, average- dishes. linkage and Euclidian distance on a set of genes for which a value was available in each sample. RNA extraction Gene ontology analysis Total RNA was extracted using TRIZOL (Gibco BRL, Life Technologies Inc., Frederick, MA, USA) according to the In order to identify biological processes significantly involved manufacturer’s protocol. RNA was resuspended in DEPC- in the drug effect, genes considered differentially expressed treated water and quality verified by agarose gel electrophor- between treated and control cell lines were processed through esis stained by SYBRGold dye (Molecular Probes). two programs (FatiGO and MAPPFinder) in which the gene products are annotated to the gene ontologies by diverse association tables (GOAnnotation@EBI and LocusLink/ cDNA microarray loc2go). In particular, the FatiGO Data mining program The 10K cDNA microarrays used in this study were obtained (http://fatigo.bioinfo.cnio.es/) (Al-Shahrour et al., 2004), was through the Cancer Research UK/Ludwig/Wellcome Trust interrogated using the Ensembl gene identifier (http://www.en- consortium at the Sanger Centre, Cambridge, UK. The 10K sembl.org/), comparing the fifth level terms of genes up- or slides contained 10 752 elements comprising 9464 human downregulated against 1000 randomly selected genes from our I.M.A.G.E clones and 468 chromosome 22 gene-specific microarray after removing the genes under analysis. In PCR products, together with positive and negative controls. addition, the Gene Ontology terms were obtained using The spotting patterns and the complete annotated list LocusLink ID to query the MAPPFinder2 program (http:// of cDNAs are available at the following web site: www.GenMAPP.org) (Doniger et al., 2003), that also expands http://www.sanger.ac.uk/Projects/Microarrays/informatics/ each term to the parent terms using the GO DAG structure file hver1.2.1.shtml. for each individual ontology. Only the terms with an absolute z score over 2 and at least 5, but less than 28, genes meeting the Experimental design, cDNA preparation, hybridisation and criteria (for upregulated and belonging to the specific term) or scanning less than 50 (for downregulated and belonging to the specific term), as well as a significant P-value (calculated by a non Each treated and control sample was cultured in triplicate for 6 parametric bootstrapping approach), have been reported in days and then labelled by direct incorporation of Cy5-dCTP two separate tables (Supplementary Tables 3 and 4). fluorochrome (Amersham Pharmacia Biotech). Each of the treated and control samples was cohybridized with the Western blot analysis appropriate reference sample, which consisted of the pool of the total RNA isolated from the three control dishes of the The proteins were extracted using TRIZOL (Gibco BRL) same cell line labelled with Cy3-dCTP. This design was according to the manufacturer’s protocol and dissolved in 1% followed for each cell line. The details of the hybridisation SDS. Protein concentration was measured using Bio-Rad and washing protocols are available (http://www.cgal.icnet.uk/ protein assay (Bio-Rad) and BSA (Sigma) to generate a exprotocols/protocols.html). After washing, fluorescence in- standard curve. In all, 20 mg of proteins was fractionated by tensities were scanned by dual-laser Affymetrix 428 Scanner electrophoresis in 8 or 15% SDS–polyacrylamide 1 mm gels (Affymetrix, High Wycombe, UK Ltd). and electroblotted onto nitrocellulose membranes. Blocking

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 209 Table 3 Primers and conditions for RT–PCR Gene Primers Size (bp) PCR condition

Cyclin B1 Sense – 50-AAGAGCTTTAAACTTTGGTCTGGG-30 319 Annealing : 601C Â 20 s Anti-sense – 50-CTTTGTAAGTCCTTGATTTACCATG-30 No. of cycles: 34 CDC2 Sense – 50-GGAGAAGGTACCTATGGAGTTG-30 275 Annealing : 601C Â 20 s Anti-sense – 50-GAATCCATGTACTGACCAGGAG-30 No. of cycles: 28 PCNA Sense – 50-GCGCTAGTATTTGAAGCACCAA-30 469 Annealing : 601C Â 45 s Anti-sense – 50-CGATCTTGGGAGCCAAGTAGTA-30 No. of cycles: 25 14-3-3 sigma Sense – 50-GTGTGTCCCCAGAGCCATGG-30 279 Annealing : 601C Â 45 s Anti-sense – 50-ACCTTCTCCCGGTACTCACG-30 No. of cycles: 26 DNMT1 Sense – 50-ATGCCGGCGCGTACCGCCCCAG-30 501 Annealing : 581C Â 30 s Anti-sense – 50-ATGGTGGTTTGCCTGGTGC-30 No. of cycles: 25 DNMT2 Sense – 50-AAGCTGTAAGCCAGCCCATATAC-30 148 Annealing : 601C Â 20 s Anti-sense – 50-TCAGCAGTGAACAGAACCTACATG-30 No. of cycles: 30 DNMT3a Sense – 50-GGGGACGTCCGCAGCGTCACAC-30 279 Annealing : 651C Â 30 s Anti-sense – 50-CAGGGTTGGACTCGAGAAATCGC-30 No. of cycles: 28 DNMT3b Sense – 50-CCTGCGAATTACTCACGCCCC-30 420 Annealing : 651C Â 30 s Anti-sense – 50-GTCTGTGTAGTGCACAGGAAAGCC-30 No. of cycles: 28 Beta-actin Sense – 50-GGAGTCCTGTGGCATCCACG-30 321 Annealing: 601C Â 30 s Anti-sense – 50-CTAGAAGCATTTGCGGTGGA-30 No. of cycles: 20

buffers were either 5% of non-fat dry milk or 1.5% of BSA UBD (Hs00197374_m1), Mx2 (Hs00159418 m1), G1P2 plus 0.05% Tween20 in PBS. The primary antibodies were all (Hs00192713_m1) and ISG20 (Hs00158122_m1). cDNAs were from Upstate (Lake Placid, NY, USA) and included: anti- prepared from the three independent RNA samples from STAT1-CT (rabbit polyclonal IgG), anti-P-STAT1 (phospho control and treated cells used for the microarray experiments. Ser 727) (rabbit polyclonal IgG), anti-p21waf1/cip1 (mouse Quantitative RT–PCR was performed in 25 ml containing 1 ml IgG derived from two clones) and anti-human p16 INK4a of five times dilution of the reverse transcription product (16 kDa) (mouse monoclonal IgGk). Secondary antibodies according to the manufacturer’s protocol using the ABI (anti-rabbit anti-mouse) conjugated with HRP were from PRISM 7000 Sequence Detection System (Applied Biosys- Santa Cruz Biotechnology (Santa Cruz, CA, USA). The tems). Each sample reaction was repeated twice and the b-actin fragments were detected using ECL kit (Amersham) and transcript level was used as a reference control. BioMax films (Kodak). In order to make all the data comparable, the number of cycles was normalized by subtracting the related mean number RT–PCR of cycles for the b-actin (Hs99999903_m1), and scaled subtracting the lowest median among our six experimental cDNAs were synthesized from 1 mg of total RNA using an conditions. The statistical significance of the differences oligo dT primer and the Superscript II reverse transcription kit between controls and treated samples was investigated by the (GibcoBRL) according to the manufacturer’s protocol. In all, Mann–Whitney test. All the analysis was performed using the 1 ml of the cDNA was used for each PCR in a 20 ml reaction R program. volume containing 200 mM dNTPs, 1 mM sense and antisense primers, 1.25 mM MgCl2,2mlof10Â PCR buffer and 0.5 U of Taq polymerase (Roche). PCR conditions and primers used are reported in Table 3. b-Actin was used as the endogenous Abbreviations control. The PCR products were analysed by electrophoresis 5-aza-CdR, 5-aza-20-deoxycytidine; DNA MTase, DNA on a 2% agarose gel, followed by staining with ethidium methyltransferase; IFN, interferon. bromide.

Quantitative real-time RT–PCR Acknowledgements This study was supported by grants from Cancer Research Among the induced genes, seven belonging to the IFN UK, National Translational Cancer Research Network (UK pathway were chosen to be validated by quantitative RT– Department of Health), Associazione Italiana Ricerca Cancro PCR, using Assays-on-Demand t Gene Expression reagents (AIRC), Milan, Italy, Fondazione Giorgio Zanotto, Verona, (Applied Biosystems). These were CCNA2 (Hs00153138_m1), Italy and Ministero Italiano Universita` e Ricerca, Rome, Italy. GADD45A (Hs00169255_m1), IFNGR1 (Hs00166223_m1),

References

Adorjan P, Distler J, Lipscher E, Model F, Muller J, Ando T, Kawabe T, Ohara H, Ducommun B, Itoh M and Pelet C, Braun A, Florl AR, Gutig D, Grabs G, Okamoto T. (2001). J. Biol. Chem., 276, 42971–42977. Howe A, Kursar M, Lesche R, Leu E, Lewin A, Antequera F and Bird A. (1993). Proc. Natl. Acad. Sci. USA, Maier S, Muller V, Otto T, Scholz C, Schulz WA, Seifert 90, 11995–11999. HH, Schwope I, Ziebarth H, Berlin K, Piepenbrock C and Antequera F, Boyes J and Bird A. (1990). Cell, 62, Olek A. (2002). Nucleic Acids Res., 30, e21. 503–514. Al-Shahrour F, Diaz-Uriarte R and Dopazo J. (2004). Aparicio A and Weber JS. (2002). Curr. Opin. Investig. Drugs, Bioinformatics, 20, 578–580. 3, 627–633.

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 210 Baylin SB, Herman JG, Graff JR, Vertino PM and Issa JP. Herman JG, Merlo A, Mao L, Lapidus RG, Issa JP, Davidson (1998). Adv. Cancer Res., 72, 141–196. NE, Sidransky D and Baylin SB. (1995). Cancer Res., 55, Baylin SB, Makos M, Wu JJ, Yen RW, de Bustros A, Vertino 4525–4530. P and Nelkin BD. (1991). Cancer Cells, 3, 383–390. Hirose T, Sowa Y, Takahashi S, Saito S, Yasuda C, Shindo N, Beaulieu N, Morin S, Chute IC, Robert MF, Nguyen H Furuichi K and Sakai T. (2003). Oncogene, 22, and MacLeod AR. (2002). J. Biol. Chem., 277, 28176–28181. 7762–7773. Bender CM, Pao MM and Jones PA. (1998). Cancer Res., 58, Hobeika AC, Etienne W, Cruz PE, Subramaniam PS and 95–101. Johnson HM. (1998). Int. J. Cancer, 77, 138–145. Bird A, Taggart M, Frommer M, Miller OJ and Macleod D. Iacobuzio-Donahue CA, Maitra A, Olsen M, Lowe AW, van (1985). Cell, 40, 91–99. Heek NT, Rosty C, Walter K, Sato N, Parker A, Ashfaq R, Bird AP. (1993). Cold Spring Harb. Symp. Quant. Biol., 58, Jaffee E, Ryu B, Jones J, Eshleman JR, Yeo CJ, Cameron 281–285. JL, Kern SE, Hruban RH, Brown PO and Goggins M. Bromberg JF, Horvath CM, Wen Z, Schreiber RD and Darnell (2003). Am. J. Pathol., 162, 1151–1162. Jr JE. (1996). Proc. Natl. Acad. Sci. USA, 93, 7673–7678. Ihaka R and Gentleman R. (1996). J. Comput. Graphical Brysk MM, Selvanayagam P, Arany I, Brysk H, Tyring SK Statist., 5, 299–314. and Rajaraman S. (1995). J. Interferon Cytokine Res., 15, Jackson-Grusby L, Beard C, Possemato R, Tudor M, 1029–1035. Fambrough D, Csankovszki G, Dausman J, Lee P, Wilson Caraglia M, Pinto A, Correale P, Zagonel V, Genua G, Leardi C, Lander E and Jaenisch R. (2001). Nat. Genet., 27, 31–39. A, Pepe S, Bianco AR and Tagliaferri P. (1994). Ann. Oncol., Jin S, Tong T, Fan W, Fan F, Antinore MJ, Zhu X, 5, 269–276. Mazzacurati L, Li X, Petrik KL, Rajasekaran B, Wu M Carrier F, Smith ML, Bae I, Kilpatrick KE, Lansing TJ, Chen and Zhan Q. (2002). Oncogene, 21, 8696–8704. CY, Engelstein M, Friend SH, Henner WD and Gilmer TM. Jones ME. (1980). Annu. Rev. Biochem., 49, 253–279. (1994). J. Biol. Chem., 269, 32672–32677. Jones PA and Buckley JD. (1990). Adv. Cancer Res., 54, 1–23. Chen B, He L, Savell VH, Jenkins JJ and Parham DM. (2000). Jones PL and Wolffe AP. (1999). Semin. Cancer Biol., 9, Cancer Res., 60, 3290–3298. 339–347. Chin YE, Kitagawa M, Su WC, You ZH, Iwamoto Y and Fu Kane MF, Loda M, Gaida GM, Lipman J, Mishra R, XY. (1996). Science, 272, 719–722. Goldman H, Jessup JM and Kolodner R. (1997). Cancer Chuang LS, Ian HI, Koh TW, Ng HH, Xu G and Li BF. Res., 57, 808–811. (1997). Science, 277, 1996–2000. Karpf AR, Moore BC, Ririe TO and Jones DA. (2001). Mol. Clemens MJ and McNurlan MA. (1985). Biochem. J., 226, Pharmacol., 59, 751–757. 345–360. Karpf AR, Peterson PW, Rawlins JT, Dalley BK, Yang Q, Costello JF, Fruhwald MC, Smiraglia DJ, Rush LJ, Robert- Albertsen H and Jones DA. (1999). Proc. Natl. Acad. Sci. son GP, Gao X, Wright FA, Feramisco JD, Peltomaki P, USA, 96, 14007–14012. Lang JC, Schuller DE, Yu L, Bloomﬁeld CD, Caligiuri MA, Kominsky S, Johnson HM, Bryan G, Tanabe T, Hobeika AC, Yates A, Nishikawa R, Su Huang H, Petrelli NJ, Zhang X, Subramaniam PS and Torres B. (1998). Oncogene, 17, O’Dorisio MS, Held WA, Cavenee WK and Plass C. (2000). 2973–2979. Nat. Genet., 24, 132–138. Kulaeva OI, Draghici S, Tang L, Kraniak JM, Land SJ and Creusot F, Acs G and Christman JK. (1982). J. Biol. Chem., Tainsky MA. (2003). Oncogene, 22, 4118–4127. 257, 2041–2048. Leonhardt H, Page AW, Weier HU and Bestor TH. (1992). Cross SH and Bird AP. (1995). Curr. Opin. Genet. Dev., 5, 309– Cell, 71, 865–873. 314. Liang G, Gonzales FA, Jones PA, Orntoft TF and Thykjaer T. Davidson S, Crowther P, Radley J and Woodcock D. (1992). (2002). Cancer Res., 62, 961–966. Eur. J. Cancer, 28, 362–368. Lu X, Tan CK, Zhou JQ, You M, Carastro LM, Downey KM De Smet C, Lurquin C, Lethe B, Martelange V and Boon T. and So AG. (2002). J. Biol. Chem., 277, 24340–24345. (1999). Mol. Cell. Biol., 19, 7327–7335. Lubomierski N, Kersting M, Bert T, Muench K, Wulbrand U, Doniger SW, Salomonis N, Dahlquist KD, Vranizan K, Schuermann M, Bartsch D and Simon B. (2001). Cancer Lawlor SC and Conklin BR. (2003). Genome Biol., 4, R7. Res., 61, 5905–5910. Dore BT and Momparler RL. (1992). Anticancer Drugs, 3, Michalowsky LA and Jones PA. (1987). Mol. Cell. Biol., 7, 281–287. 3076–3083. Dulic V, Stein GH, Far DF and Reed SI. (1998). Mol. Cell. Moore PS, Sipos B, Orlandini S, Sorio C, Real FX, Lemoine Biol., 18, 546–557. NR, Gress T, Bassi C, Kloppel G, Kalthoff H, Ungefroren Eggert A, Grotzer MA, Zuzak TJ, Wiewrodt BR, Ikegaki N H, Lohr M and Scarpa A. (2001). Virchows Arch., 439, and Brodeur GM. (2000). Med. Pediatr. Oncol., 35, 603–607. 798–802. Fang JY and Lu YY. (2002). World J. Gastroenterol., 8, Nieto M, Samper E, Fraga MF, Gonzalez de Buitrago G, 400–405. Esteller M and Serrano M. (2004). Oncogene, 23, Ferguson AT, Evron E, Umbricht CB, Pandita TK, Chan TA, 735–743. Hermeking H, Marks JR, Lambers AR, Futreal PA, Ottaviano YL, Issa JP, Parl FF, Smith HS, Baylin SB and Stampfer MR and Sukumar S. (2000). Proc. Natl. Acad. Davidson NE. (1994). Cancer Res., 54, 2552–2555. Sci. USA, 97, 6049–6054. Paulin R, Grigg GW, Davey MW and Piper AA. (1998). Furukawa T, Sunamura M, Motoi F, Matsuno S and Horii A. Nucleic Acids Res., 26, 5009–5010. (2003). Am. J. Pathol., 162, 1807–1815. Pisani P, Parkin DM, Bray F and Ferlay J. (1999). Int. J. Herman JG, Jen J, Merlo A and Baylin SB. (1996). Cancer Cancer, 83, 18–29. Res., 56, 722–727. Ramana CV, Chatterjee-Kishore M, Nguyen H and Stark GR. Herman JG, Latif F, Weng Y, Lerman MI, Zbar B, Liu S, (2000). Oncogene, 19, 2619–2627. Samid D, Duan DS, Gnarra JR, Linehan WM and Baylin Reid TR, Merigan TC and Basham TY. (1992). J. Interferon SB. (1994). Proc. Natl. Acad. Sci. USA, 91, 9700–9704. Res., 12, 131–137.

Oncogene Growth delay of human pancreatic cancer cells E Missiaglia et al 211 Rhee I, Bachman KE, Park BH, Jair KW, Yen RW, Schuebel Sibinga NE, Wang H, Perrella MA, Endege WO, Patterson C, KE, Cui H, Feinberg AP, Lengauer C, Kinzler KW, Baylin Yoshizumi M, Haber E and Lee ME. (1999). J. Biol. Chem., SB and Vogelstein B. (2002). Nature, 416, 552–556. 274, 12139–12146. Rideout III WM, Eversole-Cire P, Spruck III CH, Hustad Suh SI, Pyun HY, Cho JW, Baek WK, Park JB, Kwon T, Park CM, Coetzee GA, Gonzales FA and Jones PA. (1994). Mol. JW, Suh MH and Carson DA. (2000). Cancer Lett., 160, Cell. Biol., 14, 6143–6152. 81–88. Robertson KD, Keyomarsi K, Gonzales FA, Velicescu M and Tazi J and Bird A. (1990). Cell, 60, 909–920. Jones PA. (2000). Nucleic Acids Res., 28, 2108–2113. Tiefenbrun N, Melamed D, Levy N, Resnitzky D, Hoffman I, Robertson KD, Uzvolgyi E, Liang G, Talmadge C, Sumegi J, Reed SI and Kimchi A. (1996). Mol. Cell. Biol., 16, Gonzales FA and Jones PA. (1999). Nucleic Acids Res., 27, 3934–3944. 2291–2298. Tusher VG, Tibshirani R and Chu G. (2001). Proc. Natl. Acad. Rousseau D, Cannella D, Boulaire J, Fitzgerald P, Fotedar A Sci. USA, 98, 5116–5121. and Fotedar R. (1999). Oncogene, 18, 4313–4325. Ueki T, Toyota M, Skinner H, Walter KM, Yeo CJ, Issa JP, Saito Y, Kanai Y, Sakamoto M, Saito H, Ishii H and Hruban RH and Goggins M. (2001). Cancer Res., 61, Hirohashi S. (2002). Proc. Natl. Acad. Sci. USA, 99, 8540–8546. 10060–10065. Ueki T, Toyota M, Sohn T, Yeo CJ, Issa JP, Hruban RH and Sakai T, Toguchida J, Ohtani N, Yandell DW, Rapaport JM Goggins M. (2000). Cancer Res., 60, 1835–1839. and Dryja TP. (1991). Am. J. Hum. Genet., 48, 880–888. Venkatasubbarao K, Ammanamanchi S, Brattain MG, Santi DV, Norment A and Garrett CE. (1984). Proc. Natl. Mimari D and Freeman JW. (2001). Cancer Res., 61, Acad. Sci. USA, 81, 6993–6997. 6239–6247. Santini V, Kantarjian HM and Issa JP. (2001). Ann. Intern. Weisenberger DJ, Velicescu M, Cheng JC, Gonzales FA, Med., 134, 573–586. Liang G, Jones PA, Xu GL, Hu YG, Marquez VE, Nguyen Sato N, Fukushima N, Maitra A, Iacobuzio-Donahue CA, CT, Lin JC, Preciado-Lopez MA and Tsai YC. (2004). Mol. van Heek NT, Cameron JL, Yeo CJ, Hruban RH and Cancer Res., 2, 62–72. Goggins M. (2004). Am. J. Pathol., 164, 903–914. Yamada T, Ohwada S, Saitoh F, Adachi M, Morishita Y and Sato N, Fukushima N, Maitra A, Matsubayashi H, Yeo CJ, Hozumi M. (1996). Anticancer Res., 16, 735–740. Cameron JL, Hruban RH and Goggins M. (2003a). Cancer Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J and Res., 63, 3735–3742. Speed TP. (2002). Nucleic Acids Res., 30, e15. Sato N, Maehara N, Su GH and Goggins M. (2003b). J. Natl. Yoder JA and Bestor TH. (1996). Biol. Chem., 377, 605–610. Cancer Inst., 95, 327–330. Youssef EM, Chen XQ, Higuchi E, Kondo Y, Garcia-Manero Sato N, Maitra A, Fukushima N, van Heek NT, Matsubaya- G, Lotan R and Issa JP. (2004). Cancer Res., 64, shi H, Iacobuzio-Donahue CA, Rosty C and Goggins M. 2411–2417. (2003c). Cancer Res., 63, 4158–4166. Zhu WG, Hileman T, Ke Y, Wang P, Lu S, Duan W, Dai Z, Shuai K, Stark GR, Kerr IM and Darnell Jr JE. (1993). Tong T, Villalona-Calero MA, Plass C and Otterson GA. Science, 261, 1744–1746. (2004). J. Biol. Chem., 279, 15161–15166.

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

Oncogene