Oncogene (2004) 23, 2401–2407 & 2004 Nature Publishing Group All rights reserved 0950-9232/04 $25.00 www.nature.com/onc

Frequent loss of RUNX3 gene expression in human bile duct and pancreatic cancer cell lines

Manabu Wada1, Shujiro Yazumi1, Shigeo Takaishi1, Kazunori Hasegawa1, Mitsutaka Sawada1, Hidenori Tanaka1, Hiroshi Ida1,2, Chouhei Sakakura3, Kosei Ito2, Yoshiaki Ito2 and Tsutomu Chiba*,1

1Department of Gastroenterology and Hepatology, Graduate School of Internal Medicine, Kyoto University, Kyoto 606-8507, Japan; 2Institute of Molecular and Cell Biology and Oncology Research Institute, National University of Singapore, 30 Medical Drive, Singapore 117609, Singapore; 3Department of Digestive Surgery, Kyoto Prefectural University of Medicine, Kyoto 602-8566, Japan

RUNX3, a Runt domain transcription factor involved in Introduction TGF-b signaling, is a candidate tumor-suppressor gene localized in 1p36, a region commonlydeleted in a wide Pancreatic and bile duct carcinomas, respectively, are varietyof human tumors, including those of the stomach, the fifth and the seventh leading causes of cancer deaths bile duct, and pancreas. Recently, frequent inactivation of in Japan. Their prognosis has not changed significantly RUNX3 has been demonstrated in human gastric in recent years, despite improvements in diagnosis and carcinomas. In this study, to examine the involvement of treatment. Traditional chemotherapy and radiotherapy RUNX3 abnormalities in tumorigenesis of bile duct as regimens have proved to be largely ineffective (Schnall well as pancreatic cancers, we investigated not onlythe and Macdonald, 1996). At present, surgical resection is expression but also methylation status of RUNX3 in 10 the best treatment option for early-stage tumors of the human bile duct and 12 pancreatic cancer cell lines. Seven bile duct and pancreas. Unfortunately, however, most (70%) of the bile duct and nine (75%) of the pancreatic bile duct and pancreatic tumors have usually invaded cancer cell lines exhibited no expression of RUNX3 by regional lymph nodes and have metastasized beyond both Northern blot analysis and the reverse transcriptase those nodes by the time they are diagnosed, making polymerase chain reaction. All of the 16 cell lines that did curative resection difficult. For this reason, further not express RUNX3 also showed methylation of the clarification of the mechanisms of the carcinogenesis of promoter CpG island of the gene, whereas the six cell lines bile duct and pancreatic tumors is required. that showed RUNX3 expression were not methylated or Multiple tumor-suppressor pathways are known to be onlypartiallymethylatedin the RUNX3 promoter region. abrogated in bile duct and pancreatic carcinomas. For Moreover, treatment with the methylation inhibitor 50- example, the Rb/p16 pathway is disrupted in more than aza-20-deoxycitidine activated RUNX3 mRNA expression 95% of pancreatic carcinomas, and p53 is frequently in all of 16 cancer cell lines that originallylacked RUNX3 inactivated in both bile duct and pancreatic carcinomas expression. Finally, hemizygous deletion of RUNX3,as (Goggins et al., 1999). In addition, Smad4 has also been detected byfluorescence in situ hybridization, was found in found to be inactivated in more than 50% of pancreatic 15 of the 16 cancer cell lines that lacked RUNX3 carcinomas (Yoshida et al., 1995; Schutte et al., 1997; expression. These data suggest that the inactivation of Goggins et al., 1999), and abnormalities of either Smad4 RUNX3 plays an important role in bile duct and or TGF-b receptor II (TbR-II) are often present in pancreatic carcinogenesis, and that methylation is a human bile duct cancer cell lines (Yazumi et al., 2000). common mechanism bywhich the gene is inactivated. Therefore, disregulation of the TGF-b/Smad signaling Oncogene (2004) 23, 2401–2407. doi:10.1038/sj.onc.1207395 pathway apparently plays an important role in the Published online 26 January 2004 pathogenesis of both pancreatic and bile duct carcino- mas. ONCOGENOMICS Keywords: RUNX3; bile duct cancer; pancreatic cancer; The human runt-related (RUNX) genes encode the a methylation subunit of the Runt domain transcription factor PEBP2/ CBF (Ito, 1999), and are homologues to the Drosophila genes runt (Kania et al., 1990) and lozenge (Daga et al., 1996). The mammalian and Drosophila genes share an evolutionarily conserved region of 128 amino acids, termed the Runt domain, that is required for DNA binding and heterodimerization with the b subunit PEBP2b/CBFb (Ito, 1999). RUNX3 belongs to the *Correspondence: T Chiba; E-mail: [email protected] Runt domain family, which includes RUNX1 and Received 4 August 2003; revised 16 November 2003; accepted 20 RUNX2. All the three genes are master regulators of November 2003 gene expression in major developmental pathways. Loss of RUNX3 in bile duct and pancreatic cancer M Wada et al 2402 Interestingly, in primary cultures of RUNX3-null gastric Table 1 mRNA expression, methylation, and hemizygous deletion of epithelial cells, the cells were less sensitive to TGF-b- RUNX3 in human bile duct and pancreatic cancer cell lines induced growth inhibition and apoptosis (Li et al., RUNX3 2002). Furthermore, RUNX3 protein has been found to bind the Smad2 and Smad3 proteins (Hanai et al., 1999). Cell lines Expression Methylation Reactivation Hemizygous These data suggest a possible role for RUNX3 in deletion transducing TGF-b signaling. Bile duct cancers Recently, Li et al. (2002) demonstrated that Runx3- Mz-ChA-1 + P NE À null mice develop hyperplasia of the gastric mucosa KMBC + U NE + through activation of cellular proliferation and suppres- TGBC24TKB + U NE + RBE À M+ + sion of apoptosis in epithelial cells. They also found that SK-ChA-1 À M+ + 47% of human gastric cancer cell lines lost expression of Mz-ChA-2 À M+ + RUNX3 through a combination of hemizygous deletion TGBC1TKB À M+ + of 1p36 and methylation of the RUNX3 promoter region TGBC2TKB À M+ + HuCCT1 À M+ + (Li et al., 2002). Interestingly, 1p36, where RUNX3 TFK-1 À M+ + maps to (Bae et al., 1995), is a region commonly deleted in a wide variety of human carcinomas, including those Pancreas cancers of the bile duct and pancreas (Weith et al., 1996; Kang AsPC1 + P NE À et al., 2000; Schleger et al., 2000). In the present study, T3M4 + P NE À KLM-1 + U NE + therefore, to elucidate the roles of RUNX3 in carcino- BxPC3 À M+ À genesis of the bile duct and pancreas, we examined not PK59 À M+ + only the expression but also the methylation status of PK1 À M+ + RUNX3 in human bile duct and pancreatic cancer cell PK8 À M+ + Kp4-1 À M+ + lines. PK-45H À M+ + PK-45P À M+ + MiaPaca2 À M+ + PK9 À M+ + Results 1C3D3 + U NE À Frequent loss of RUNX3 expression U, unmethylated; P, partially methylated; M, methylated; NE, not The expression of RUNX3 mRNA was examined in 10 examined; +, positive; À, negative bile duct and 12 pancreatic cancer cell lines by Northern blot analysis. Of the 22 cancer cell lines tested, seven (70%) bile duct and nine (75%) pancreatic cell lines did of the RUNX3 promotor CpG island, and three (Mz- not express RUNX3 mRNA (Table 1). Representative ChA-1, AsPC1, and T3M4) exhibited only partial results of Northern blot analysis are shown in Figure 1a. methylation of the RUNX3 promoter region. Sequence The results of the RT–PCR analysis were consistent analysis of each PCR product revealed that the cell lines with those of the Northern blot analysis (Figure 1b, that exhibited methylation of the RUNX3 promoter Table 1). On the other hand, both the human embryonic CpG islands possessed only a methylated allele, whereas pancreas-derived cell line 1C3D3 and all the four those that showed no methylation of the RUNX3 samples of normal biliary mucosa tested expressed promoter region had only an unmethylated allele. RUNX3 mRNA (Figure 1). Moreover, three cell lines (Mz-ChA-1, AsPC1, and T3M4) that exhibited partial methylation of the RUNX3 DNA methylation of the exon 1 region of RUNX3 promoter region had both methylated and unmethylated alleles (Figure 2). The human embryonic pancreas- The pattern of DNA methylation is often altered in derived cell line 1C3D3 had no methylation of RUNX3 cancer cells. Growing evidence suggests that aberrant promotor CpG island, and normal biliary mucosa DNA methylation of CpG islands around the promoter exhibited partial methylation of the RUNX3 promoter regions can have the same effect on the inactivation of region (Figure 2). tumor-suppressor genes as do mutations in the coding regions. The observation that the RUNX3 gene promo- Reactivation of RUNX3 expression ter consists of a CpG island suggests that DNA methylation may play a role in inhibiting RUNX3 Reactivation of RUNX3 was attempted in the 16 cancer expression. To study DNA methylation in the RUNX3 cell lines in which expression of the gene could not be gene promoter, genomic DNA was isolated, treated with detected with either Northern blot analysis or RT–PCR sodium bisulfite, and analysed by using MSP. As shown by incubating the cell lines with AZA alone, trichostatin in Figure 2 and Table 1, all 16 cancer cell lines that did A (TSA) alone, or AZA plus TSA. Representative not express RUNX3 had methylation of RUNX3 results of RT–PCR performed after incubation are promotor CpG island. In contrast, of the six cell lines shown in Figure 3. Incubation with TSA only reacti- that had RUNX3 gene expression, three (KMBC, vated the expression of RUNX3 in nine of the 16 cancer TGBC24TKB, and KLM-1) exhibited no methylation cell lines, whereas incubation with AZA alone or with

Oncogene Loss of RUNX3 in bile duct and pancreatic cancer M Wada et al 2403

Figure 1 Expression of RUNX3 mRNA in bile duct and pancreatic cancer cell lines. (a) Northern blot analysis of mRNA (5 mg) from each cell line was performed by using the 32P-labeled Figure 2 MSP analysis of RUNX3 in bile duct and pancreatic RUNX3-specific probe and the GAPDH-specific probe. The gastric cancer cell lines. The PCR products in the left lanes (M) show the cancer cell line AGS was used as a negative control expressing presence of methylated templates of each gene, whereas the RUNX3 mRNA (lane 2). Lanes 3–6, bile duct cancer cell lines; products in the right lanes (U) indicate the presence of lanes 7–12, pancreatic cancer cell lines. The human embryonic unmethylated templates. (a) Bile duct cancer cell lines. (b) Normal pancreas-derived cell line 1C3D3 was also used (lane 1). human biliary mucosa. (c) Pancreatic cancer cell lines. (d) Human Rehybridization with GAPDH, as shown in the lower panel, was embryonic pancreas-derived cell line. Representative data are used to confirm approximately equal loading of mRNAs. In shown. (e) Sequence analysis of each PCR product from bisulfite- addition to the human embryonic pancreas-derived cell line treated methylated and/or unmethylated DNA. Each PCR product 1C3D3, cancer cell lines Mz-ChA-1, KMBC, AsPC1, and KLM- between À258 and À68 of RUNX3 gene was cloned into a TA 1 expressed RUNX3 mRNA. (b) RT–PCR analysis of RUNX3 cloning Kit. The methylated status of each PCR product was resulted in 340-bp DNA products, as expected. GAPDH was used confirmed by sequencing more than three clones. Representative as internal control. (c) RT–PCR analysis of RUNX3 and GAPDH data of T3M4 were shown. The boxes indicate 43 CpG sites. In of four samples of normal human biliary mucosa (lanes 13–16). most of CpG sites, the C residues in the PCR product amplified by Representative data are shown the unmethylated primer are not methylated, and those residues in the PCR product amplified by the methylated primer are methylated, except for two CpG sites in each allele (italic C and AZA plus TSA reactivated RUNX3 in all the 16 of them. T in the unmethylated and the methylated allele, respectively) (*). Wild, unmethylated, and methylated represent the sequence of The signal intensity was stronger, however, when the cell PCR products amplified by wild type, unmethylated, and lines were incubated with AZA plus TSA than when methylated primers, respectively with AZA or TSA alone (Figure 3). DNA methylation, therefore, appeared to be responsible for the absence of RUNX3 expression in the bile duct and pancreatic cancer cell lines we studied.

Hemizygous deletion of RUNX3 gene The integrity of RUNX3 was examined by FISH analysis. Most of the 10 bile duct and 12 pancreatic cancer cell lines tested, as well as the human embryonic pancreas-derived cell line 1C3D3 and the gastric cancer cell line AGS, tested previously (Ishino et al., 1998), were aneuploid, having two, three, or more copies of Figure 3 Reactivated expression of RUNX3 mRNA in bile duct chromosome 1, as revealed by a centromere-specific (SK-ChA-1 and Mz-ChA-1) and pancreatic cancer cells (PK59 and probe (Figure 4). When we compared the number of BxPC3). Cancer cells were cultured in the presence of AZA (1 mM) copies of chromosome 1 to that of RUNX3, nine (90%) and/or TSA (500 nM). Control experiments were performed with neither AZA nor TSA (mock trial). Total RNAs extracted from the of the 10 bile duct cancer cell lines and eight (67%) of control and treated cells were analysed by RT–PCR using a the 12 pancreatic cancer cell lines exhibited a hemi- RUNX3 primer set and a GAPDH primer set. Representative data zygous deletion of RUNX3. It is noteworthy that, in the are shown. AZA, 50-aza-20-deoxycytidine, an inhibitor of DNA six cell lines with RUNX3 expression, its hemizygous methyltransferase; TSA, trichostatin A, an inhibitor of histone deletion was present in only three (50%), whereas deacetylase hemizygous deletion of RUNX3 was detected in 15 (94%) of the 16 cell lines that did not express it. showed no methylation of RUNX3. In contrast, Interestingly, the three cancer cell lines that had the remaining three RUNX3-expressing cell lines hemizygous deletion of RUNX3 in the presence of that exhibited no hemizygous deletion (Mz-ChA-1, RUNX3 gene expression (KMBC, 24TKB, KLM-1) AsPC1, T3M4) revealed partial methylation of RUNX3,

Oncogene Loss of RUNX3 in bile duct and pancreatic cancer M Wada et al 2404 Table 2 mRNA expression of Smad4 and TbR-II and mutation of the TbR-II gene in human bile duct and pancreatic cancer cell lines Cell lines Smad4 TbR-II Mutation of expression expression TbR-II

Bile duct cancers Mz-ChA-1 À + À KMBC + + À TGBC24TKB + + À RBE + + À SK-ChA-1 À + À Mz-ChA-2 + + À TGBC1TKB + + À TGBC2TKB + + À HuCCT1 À + À TFK-1 + + À

Pancreas cancers AsPC1 + + À Figure 4 FISH analysis of the RUNX3 gene and chromosome 1 in T3M4 + + À human embryonic pancreas-derived cell line 1C3D3 (a), gastric KLM-1 À + À cancer cell line AGS (b), bile duct cancer cell line Mz-ChA-1 (c), BxPC3 À + À and pancreatic cancer cell line KLM-1 (d). Typical photographs of PK59 + + À the FISH results are shown; the signals of the RUNX3 gene are red, PK1 À + À and those of the centromere are green. The number of chromosome PK8 + + À 1 signals and of RUNX3 signals are the same in the 1C3D3 (a) and Kp4-1 À + À Mz-ChA-1 cells (c), showing no hemizygous deletion. The number PK-45H À + À of RUNX3 signals is less than that of the chromosome 1 signals in PK-45P À + À AGS (b) and KLM-1 cells (d), demonstrating the presence of a MiaPaca2 + ÀÀ hemizygous deletion. An analysis of normal peripheral lympho- PK9 À + À cytes (data not shown) showed two centromere spots and two RUNX3 spots 1C3D3 + + À

suggesting the presence of both methylated and un- methylated RUNX3 genes. On the other hand, no hemizygous deletion was present in the human embryonic pancreas-derived cell line 1C3D3.

Expression of Smad4 and TbR-II genes The expression of Smad4 mRNA was examined in 10 bile duct and 12 pancreatic cancer cell lines by Northern blot analysis. Three (30%) bile duct and seven (58%) pancreatic cancer cell lines did not express Smad4 Figure 5 Expression of Smad4 and TbR-II mRNAs in human bile duct and pancreatic cancer cell lines. (a) Northern blot analysis of mRNA (Table 2). The representative results of Northern mRNA (5 mg) from each cell line was performed by using the 32P- blot analysis are shown in Figure 5a. The expression of labeled Smad4-specific probe and the GAPDH-specific probe. Lanes 1– TbR-II mRNA was examined by RT–PCR. Except for 3; bile duct cancer cell lines, lanes 4–7; pancreatic cancer cell lines. MiaPaca2, all bile duct and pancreatic cancer cell lines KMBC and MiaPaca2 expressed Smad4 mRNA. (b) RT–PCR analysis expressed TbR-II mRNA (Figure 5b, Table 2). The of TbR-II resulted in 260-bp DNA products, as expected. GAPDH was used as internal control. Except for MiaPaca2, all the other cell lines human embryonic pancreas-derived cell line 1C3D3 had expressed TbR-II mRNA. Representative data are shown both Smad4 and TbR-II mRNA expressions.

Mutation of the TbR-II gene human carcinomas, including those of the bile duct and the pancreas (Weith et al., 1996; Ishino et al., 1998; We analysed the 10-adenine tract located in exon 3 of Herman et al., 2000). In addition, RUNX3 has been the TbR-II gene, the so-called BAT-RII sequence, suggested to play a role in transducing TGF-b signaling because BAT-RII is often mutated in human cancer (Li et al., 2002). Li et al. (2002) have previously shown cells due to microsatellite instability (Parsons et al., that introduction of RUNX3 genes into MKN28, a 1995). In this study, however, no BAT-RII mutation human gastric cancer cell line that is deficient in was found in any bile duct and pancreatic cancer cell RUNX3, exerts an inhibitory effect on tumor cell lines (Table 2). growth. Moreover, they found that a considerable number of human gastric cancer cell lines lost RUNX3 expression (Li et al., 2002). In this study, we have Discussion demonstrated that 70% of the human bile duct and 75% of the human pancreatic cancer cell lines we examined RUNX3 maps to chromosome 1p36, a region frequently did not express RUNX3, whether analysed by Northern altered by deletions or translocations in a wide variety of blotting or by RT–PCR. The frequency of RUNX3

Oncogene Loss of RUNX3 in bile duct and pancreatic cancer M Wada et al 2405 inactivation in bile duct and pancreatic cancer cell lines three of the six cell lines that possess RUNX3 expression is remarkably high, when compared with that in other also showed partial methylation of RUNX3 without tumor-suppressor genes, such as p53, p16,andSmad4 hemizygous deletion, indicating the presence of both (Yoshida et al., 1995; Schutte et al., 1997; Goggins et al., methylated and unmethylated alleles. 1999; Yazumi et al., 2000). Moreover, it should be noted TGF-b plays a key role in regulating the growth and that the frequency of RUNX3 silencing in bile duct and differentiation of many cell types. In TGF-b1-null pancreatic cancer cell lines was higher than that we animals, proliferation of the gastric epithelium is found previously in human gastric cancer cell lines (Li stimulated and hyperplasia occurs (Crawford et al., et al., 2002). Altogether, this evidence suggests that 1998). TGF-b is known to be a potent inhibitor of inactivation of RUNX3 plays an important role in pancreatic acinar and duct cell proliferation in vitro tumorigenesis of bile duct and pancreatic carcinomas. (Bisgaard and Thorgeirsson, 1991; Logsdon et al., 1992). Although the importance of genetic mutations in Previously, we found that Runx3À/À gastric epithelial carcinogenesis has long been recognized, epigenetic cells are resistant to the apoptosis-inducing action of inactivation of tumor-suppressor genes has begun to TGF-b. Moreover, RUNX3 protein binds the Smad2 receive great attention recently (Baylin and Herman, and Smad3 proteins (Hanai et al., 1999). Therefore, 2000; Eng et al., 2000; Ueki et al., 2000; Lee et al., 2002). RUNX3 appears to be an in vivo target of the TGF-b In particular, a considerable number of studies have signaling pathway (Li et al., 2002). In this regard, it is revealed that hypermethylation of the CpG island of the interesting to note that, in addition to RUNX3 gene, a promoter region inactivates some genes – such as Rb, considerable number of the bile duct and pancreatic CDKN2A (p16), CDKN2B, VHL, E-cadherin, MLH1, cancer cell lines tested in this study do not express BRCA1,andPTEN (Baylin and Herman, 2000; Eng Smad4 mRNA. Taken together, it is strongly suggested et al., 2000; Ueki et al., 2000; Lee et al., 2002) – as that disruption of the TGF-b/Smad signaling pathway effectively as gene mutations or deletions. That RUNX3 plays an important role in bile duct and pancreatic is frequently inactivated in gastric cancer cell lines by carcinogenesis. Furthermore, the fact that some cancer hypermethylation of CpG islands in the exon 1 region cell lines do not express both RUNX3 and Smad4 gives further evidence of epigenetic inactivation (Li et al., mRNAs may indicate that multiple defects are required 2002). In the present study, we found that all of the 16 to disrupt TGF-b signaling pathway. cancer cell lines that did not express RUNX3 mRNA In conclusion, we have demonstrated for the first had hypermethylation in the CpG island of the RUNX3 time that RUNX3 expression is frequently lost in human promoter region, whereas no or partial methylation of bile duct and pancreatic cancer cell lines, and that the same region was observed in the six cancer cell lines methylation of the promoter region and hemizygous that retained RUNX3 expression. In addition, no and deletion are responsible for inactivating the gene. We only partial methylation were observed in the human propose that inactivation of RUNX3 plays an important embryonic pancreas-derived cell line and normal biliary role in tumorigenesis of bile duct and pancreatic mucosa, respectively, that expressed RUNX3. More carcinomas. strikingly, in all of the cancer cell lines that did not express RUNX3, the gene was reactivated by AZA (a specific inhibitor of methyltransferase) alone or by the combination of AZA and TSA (an inhibitor of histone Materials and methods deacetylase). Consequently, DNA methylation of the Cell lines and cell culture CpG island of the RUNX3 promoter region appears to be the mechanism by which the gene is inactivated in bile The KMBC human bile duct cancer cell line was kindly duct and pancreatic cancer cell lines. provided by Drs H Yano and M Kojiro (Kurume University, It is noteworthy that Herman et al. (2000), in Kurume, Japan); the RBE cell line was provided by Dr M allotyping intrahepatic cholangiocarcinoma cells, could Enjoji (Kyushu University, Fukuoka, Japan), and the SK- ChA-1, Mz-ChA-1, Mz-ChA-2 cell lines were kind gifts from detect only one methylated allele among 1000 unmethy- Dr A Knuth (Krankenhaus Nordwest, Frankfurt, Germany). lated ones using MSP. This finding suggests the Human bile duct cancer cell lines TGBC1TKB, TGBC2TKB, possibility that the hypermethylated loci of RUNX3 TGBC24TKB, and human pancreatic cancer cell lines T3M4 could be detected in clinical samples, such as stool, and Kp4-1 were obtained from Riken Gene Bank (Tokyo, blood, bile, or pancreatic juice, and that such detection Japan). Human pancreatic cancer cell lines MiaPaca2, AsPC1, might be clinically applicable as an early diagnostic test and BxPC3 were purchased from American Type Culture for bile duct and pancreatic carcinomas. Collection (Rockville, MD, USA). Human bile duct cancer cell It may be noted in this study that 15 out of 16 cancer lines HuCCT1 and TFK-1 and human pancreatic cancer cell cell lines that did not express RUNX3 gene had both lines, PK1, PK8, KLM-1, PK-45H, PK-45P, PK59, and PK9 methylation and hemizygous deletion of RUNX3 gene. were obtained from the Cell Resource Center of the Biomedical Research Institute of Development, Aging and Therefore, although hemizygous deletion may not Cancer at Tohoku University (Sendai, Japan). The human always indicate loss of heterozygosity (LOH) of RUNX3 embryonic pancreas-derived cell line 1C3D3 was established gene, whether or not silencing of RUNX3 is due to by Dr Ishikawa (Tokyo Jikeikai Medical University, Tokyo, biallelic methylation or combination of methylation and Japan) and provided from the RIKEN cell bank. Normal LOH of RUNX3 is an interesting question to be human biliary mucosa was obtained from pancreatoduode- clarified. In this context, we also demonstrated that nectomized specimens of four patients with pancreas cancer.

Oncogene Loss of RUNX3 in bile duct and pancreatic cancer M Wada et al 2406 All of the cell lines were cultured in DMEM or RPMI-1640 Methylation-specific PCR (MSP) medium (Gibco-BRL, Tokyo, Japan), except for T3M4 MSP was performed as previously described (Herman et al., and 1C3D3, which were maintained in F-12 medium 2000). Briefly, genomic DNA that had been denatured with (Gibco-BRL, Tokyo, Japan). All the three media were NaOH was treated with 3 M sodium bisulfite for 16 h and supplemented with 10% fetal bovine serum and penicillin- purified. The DNA was subjected to PCR by using the streptomycin, expect for 1C3D3, which was supplemented with following primer sets (CpG Ware Primer Design Software, 5% fetal bovine serum and 10% newborn bovine serum and Intergen Company, New York): untreated DNA, Rx3-5W 2.5% horse serum. (50-CCCCGCGCGGGCCCCGCCGC-30) and Rx3-3W (50-GGGGGCGGCCGGCGCGGGCG-30); detection of un- RNA extraction and reverse transcriptase–polymerase chain methylated DNA, Rx3-5U (50-TTTTGTGTGGGTTTTGT- reaction (RT–PCR) TGT-30) and Rx3-3U (50-AAAAACAACCAACACAAACA- 30); detection of methylated DNA, Rx3-5M (50-TTTCG- Total cytoplasmic RNA was isolated by using TRIzol Reagent CGCGGGTTTCGTCGC-30) and Rx3-3M (50-GAAAAC- (Gibco-BRL, Tokyo, Japan). For the RT–PCR procedure, GACCGACG CGAACG-30). Each PCR product was cloned cDNA was synthesized from 5 mg of total RNA by using an into a TA cloning Kit (Promega, WI, USA), and was oligo-dT primer (Superscript kit, Gibco-BRL, Tokyo, Japan), sequenced. according to the manufacturer’s manual. For human RUNX3 transcripts, 50-AGACGGCACCGGCAGAAG-30 was used as a sense primer and 50-TGTAGGGGAAGGCAGCTGAC-30 Reactivation of RUNX3 expression was used as an antisense primer to amplify a region Bile duct or pancreatic cancer cells (1 Â 104), in which RUNX3 corresponding to 541–841 bp of the human RUNX3 sequence, was not expressed, were cultured in 6-cm dishes for 2 days. The using the cDNA as a template. To confirm the integrity cells were then incubated with 1 mM of 50-aza-20-deoxycitidine of the prepared RNA, the same cDNA was subjected to (AZA), an inhibitor of DNA methyltransferase, for 24 h at PCR amplification of the GAPDH cDNA. GAPDH 371C. TSA, an inhibitor of histone deacetylase, was added at a 0 cDNA was amplified by using the sense primer 5 -ACCA- concentration of 500 nM, and the cells were incubated for CACTTTCTACAATGAGCTG C-30, and the antisense another 24 h at 371C. In other cultures, the cells were primer 50-CTTCTCTTTAATGTCACGCACG-30. The PCR incubated with either AZA or TSA alone. In control reactions were performed in a thermal cycler (Perkin- experiments, the cells were incubated without AZA or TSA Elmer, Tokyo, Japan). The cDNA was denatured for (mock trial). After the incubation, the cells were harvested, the 1 min at 941C, then for 35 cycles of 30 s at 941C, 30 s at total RNA was extracted, and RT–PCR was performed as 551C, and 2 min at 721C. For human TbR-II transcripts, described above. 50-TGGAGAAAGAATGACGAGAACA-30 was used as 0 the sense primer and 5 -AAGATGATGTTGTCATTG- Fluorescence in situ hybridization (FISH) CACTC-30 was used as the antisense primer. The cDNA was denatured for 10 min at 951C, then for 35 cycles of 30 s at To detect the copy number changes in RUNX3, FISH was 941C, 30 s at 551C, and 1 min at 721C. The PCR products were carried out as described previously (Ishino et al., 1998). The electrophoresed on 1% agarose gels and stained with ethidium probes used were pUC1.77, which is specific for the bromide. pericentromeric regions of chromosome 1, and RP11-84-D-1, an RUNX3 BAC clone that contains 169 kb of DNA, including all of the RUNX3 exons. The pUC1.77 and the RUNX3 BAC Northern blot analysis probes (1 mg each) were labeled with bio-16-dUTP and dig-11- The mRNA was isolated from total RNA by using the dUTP, respectively, by using a nick translation kit (Boehringer Poly A Tract mRNA Isolation System (Promega, Madison, Mannheim, Tokyo, Japan), and 1 ml of Cot-1 was added to 9 ml WI, USA). Aliquots of mRNA (5 mg) were electrophoresed of the hybridization probe solution. The final mixture was on 1.25% formaldehyde agarose gels and transferred onto denatured at 751C for 10 min, cooled on ice for 5 min, and then nitrocellulose membranes (Schleicher and Schuell, mixed with an equal volume of 4 Â SSC containing 20% Postfach, Germany). To make the probes, the RUNX3, dextran sulfate. The bile duct or pancreatic cancer cells were GAPDH, and Smad4 cDNAs were amplified as described fixed in a mixture of methanol and acetic acid (3 : 1) and placed above, and sequenced to ensure no PCR infidelity. After on microscope slides. The hybridization mixture was then being prehybridized at 651C, the membranes were hybridized added to the cells on the slides. The slides were covered with overnight at 651C with probes radiolabeled with [32P] Parafilm and were incubated in a humidified box for 16–24 h at dCTP. The membranes were then washed with 0.2 Â SCC 37oC. After being washed successively in 50% formamide/2 Â and 0.1% SDS. The expression of mRNA was quantified and SSC for 15 min, 2 Â SSC for 15 min, and 1 Â SSC for 15 min, normalized by measuring the GAPDH mRNA level using a the slides were counterstained with 1 mg/ml of 4,6-diamidino-2- Fujix BAS 2000 image analyzer (Fujix, Tokyo, Japan). Two phenylindole (DAPI), and the hybridization signals were independent experiments were performed to validate the counted. The signals numbered more than 200 interphase findings. nuclei for each slide. We considered the hemizygous deletion of RUNX3 to have occurred if more than 20% of nuclei exhibited a RUNX3-to-centromere signal ratio of less than one, as DNA extraction and sequencing analysis previously described (Ishino et al., 1998). Genomic DNAs were isolated by using DNeasy Tissue Kit (QIAGEN, Hilden, Germany). DNA fragments containing Acknowledgements exon 3 of TbR-II were amplified as described (Takenoshita We thank Ms Pamela Paradis Tice, ELS (D), for editing et al., 1998) and directly sequenced with 50-ACAGTTTGC- the manuscript. This work was supported by a Grant-in- CATGACCCCAAG-30 as a forward sequence primer and Aid for Scientific Research from the Ministry of Culture 50-GCACTCATCAGAGCTACAGG-30 as a reverse sequence and Science of Japan (11470130, 11877090, 11670496, and primer. 13670510).

Oncogene Loss of RUNX3 in bile duct and pancreatic cancer M Wada et al 2407 References

Bae SC, Takahashi E, Zhang YW, Ogawa E, Shigesada K, Yamamoto H, Yamashita N, Yano T, Ikeda T, Itohara S, Namba Y, Satake M and Ito Y. (1995). Gene, 159, 245–248. Inazawa J, Abe T, Hagiwara A, Yamagishi H, Ooe A, Baylin SB and Herman JG. (2000). Trends Genet., 16, 168–174. Kaneda A, Sugimura T, Ushijima T, Bae SC and Ito Y. Bisgaard HC and Thorgeirsson SS. (1991). J. Cell. Physiol., (2002). Cell, 109, 113–124. 147, 333–343. Logsdon CD, Keyes L and Beauchamp RD. (1992). Am. Crawford SE, Stellmach V, Murphy-Ullrich JE, Ribeiro SM, J. Physiol., 262, G364–G368. Lawler J, Hynes RO, Boivin GP and Bouck N. (1998). Cell, Parsons R, Myeroff LL, Liu B, Willson JK, Markowitz SD, 93, 1159–1170. Kinzler KW and Vogelstein B. (1995). Cancer Res., 55, Daga A, Karlovich CA, Dumstrei K and Banerjee U. (1996). 5548–5550. Genes Dev., 10, 1194–1205. Schleger C, Arens N, Zentgraf H, Bleyl U and Verbeke C. Eng C, Herman JG and Baylin SB. (2000). Nat. Genet., 24, (2000). J. Pathol., 191, 27–32. 101–102. Schnall SF and Macdonald JS. (1996). Semin. Oncol., 23, Goggins M, Kern SE, Offerhaus JA and Hruban RH. (1999). 220–228. Ann. Oncol., 4 (Suppl 10), 4–8. Schutte M, Hruban RH, Geradts J, Maynard R, Hanai J, Chen LF, Kanno T, Ohtani-Fujita N, Kim WY, Guo Hilgers W, Rabindran SK, Moskaluk CA, Hahn SA, WH, Imamura T, Ishidou Y, Fukuchi M, Shi MJ, Stavnezer Schwarte-Waldhoff I, Schmiegel W, Baylin SB, J, Kawabata M, Miyazono K and Ito Y. (1999). J. Biol. Kern SE and Herman JG. (1997). Cancer Res., 57, Chem., 274, 31577–31582. 3126–3130. Herman JG, Graff JR, Myohanen S, Nelkin BD and Baylin Takenoshita S, Tani M, Miyoshi H, Matsui M, Nagamachi Y, SB. (2000). Mod. Pathol., 13, 627–631. Bennett WP, Hagiwara K, Harris CC and Yokota J. (1998). Ishino S, Hashimoto N, Fushiki S, Date K, Mori T, Fujimoto Carcinogenesis, 19, 803–807. M, Nakagawa Y, Ueda S, Abe T and Inazawa J. (1998). Ueki T, Toyota M, Sohn T, Yeo CJ, Issa JP, Hruban RH and Cancer, 83, 360–366. Goggins M. (2000). Cancer Res., 60, 1835–1839. Ito Y. (1999). Genes Cells, 12, 685–696. Weith A, Brodeur GM, Bruns GA, Matise TC, Mischke D, Kang YK, Kim YI and Kim WH. (2000). Mod. Pathol., 13, Nizetic D, Seldin MF, van Roy N and Vance J. (1996). 627–631. Cytogenet. Cell Genet., 72, 114–144. Kania MA, Bonner AS, Duffy JB and Gergen JP. (1990). Yazumi S, Ko K, Watanabe N, Shinohara H, Yoshikawa K, Genes Dev., 10, 1701–1713. Chiba T and Takhashi R. (2000). Int. J. Cancer, 86, Lee S, Kim WH, Jung HY, Yang MH and Kang GH. (2002). 782–789. Am. J. Pathol., 161, 1015–1022. Yoshida S, Todoroki T, Ichikawa Y, Hanai S, Suzuki H, Hori Li QL, Ito K, Sakakura C, Fukamachi H, Inoue K, Chi XZ, M, Fukao K, Miwa M and Uchida K. (1995). Cancer Res., Lee KY, Nomura S, Lee CW, Han SB, Kim HM, Kim WJ, 55, 2756–2760.

Oncogene