ADAM23, a Possible Tumor Suppressor Gene, Is Frequently Silenced in Gastric Cancers by Homozygous Deletion Or Aberrant Promoter Hypermethylation

ADAM23, a possible tumor suppressor gene, is frequently silenced in gastric cancers by homozygous deletion or aberrant promoter hypermethylation

Hisashi Takada1,2,3, Issei Imoto1,2, Hitoshi Tsuda2,4, Yukihiro Nakanishi5, Takashi Ichikura6, Hidetaka Mochizuki6, Shoji Mitsufuji3, Fumie Hosoda2,7, Setsuo Hirohashi5, Misao Ohki7 and Johji Inazawa*,1,2,8

1Department of Molecular Cytogenetics, Medical Research Institute and School of Biomedical Science, Tokyo Medical and Dental University, Bunkyo-ku, Tokyo, Japan; 2Core Research for Evolutional Science and Technology (CREST) of Japan Science and Technology Corporation (JST), Hon-machi Kawaguchi, Saitama, Japan; 3Division of Gastroenterology and Hepatology, Graduate School of Medical Science, Kyoto Prefectural University of Medicine, Kawaramachi Hirokoji Kamigyo-ku, Kyoto, Japan; 4Department of Pathology II, National Defense Medical College, Namiki, Tokorozawa, Japan; 5Pathology Division, National Cancer Center Research Institute, Tsukiji Chuo-ku, Tokyo, Japan; 6Department of Surgery I, National Defense Medical College, Namiki, Tokorozawa, Japan; 7Cancer Genomics Project, National Cancer Center Research Institute, Tsukiji Chuo-ku, Tokyo, Japan; 8COE program for Frontier Research on Molecular Destruction and Reconstitution of Tooth and Bone, Tokyo Medical and Dental University, Bunkyo-ku, Tokyo, Japan

Array-based comparative genomic hybridization (CGH- Introduction array) has a powerful potential for high-throughput identification of genetic aberrations in cell genomes. We Members of the ADAM family of disintegrins and identified a homozygous loss of ADAM23 (2q33.3) in the metalloproteinases belong to the zinc protease super- course of a program to screen a panel of gastric cancer family. ADAM molecules contain both disintegrin and (GC) cell lines (1/32, 3.1%) for genomic copy-number metalloprotease domains (Seals and Courtneidge, 2003), aberrations using our custom-made CGH-array. Infre- and they modulate a variety of cell–cell and cell– quent homozygous deletion of ADAM23 was also seen in extracellular matrix interactions. These proteins have primary gastric tumors (1/39, 2.6%). ADAM23 mRNA been implicated in the control of growth factors and was expressed in normal stomach tissue, but not in the shedding of cytokines, in membrane adhesion, and in majority of GC cell lines without homozygous deletion of cell migration, as well as in processes such as muscle this gene. Expression of ADAM23 mRNA was restored to development, fertilization, and determination of cell gene-silenced GC cells after treatment with 5-aza 20-deoxy- fates. Any of these activities can be involved in cytidine. The methylation status of the ADAM23 CpG tumorigenesis (Seals and Courtneidge, 2003). Owing to island, which showed promoter activity, correlated in- their multiple functions, some members of the ADAM versely with its expression. Methylation of this CpG family have been suspected of involvement in the island was observed both in GC cell lines and in primary pathogenesis of breast cancer (O’Shea et al., 2003), GC tissues; in primary tumors with a hypermethylated prostate cancer (McCulloch et al., 2004), non-small-cell CpG island, expression of ADAM23 was lower than in lung cancer (Shintani et al., 2004), and glioblastoma adjacent noncancerous tissues. Moreover, restoration of (Kodama et al., 2004), as well as gastric cancer (GC) ADAM23 in GC cells reduced their numbers in colony- (Carl-McGrath et al., 2005). However, the association formation assays. These results suggest that genetic or between ADAM proteins and gastric carcinogenesis epigenetic silencing by hypermethylation of the ADAM23 remains unclear. Although most of these molecules are CpG-rich promoter region leads to loss of ADAM23 thought to be capable of accelerating tumorigenesis function, which may be a factor in gastric carcinogenesis. through overexpression, a few of them, including Oncogene (2005) 24, 8051–8060. doi:10.1038/sj.onc.1208952; ADAM11 and ADAM23, might be associated with published online 15 August 2005 tumorigenesis through their inactivation (Wu et al., 1997; Costa et al., 2004). Keywords: CGH-array; gastric cancer; ADAM23; As an alternative to genetic defects that may homozygous deletion; methylation contribute to human cancers, epigenetic mechanisms that involve DNA methylation and alteration of chromatin structure are important ways to silence many *Correspondence: J Inazawa, Department of Molecular Cytogenetics, genes transcriptionally, especially tumor suppressor Medical Research Institute, Tokyo Medical and Dental University, genes. In some breast tumors, the ADAM23 gene has 1-5-45 Yushima, Bunkyo-ku, Tokyo 113-8510, Japan; E-mail:[email protected] been silenced through DNA methylation (Costa et al., Received 29 March 2005; revised 7 June 2005; accepted 10 June 2005; 2004). However, no one has yet assessed the promoter published online 15 August 2005 activity and methylation status of the CpG island ADAM23 as a potential tumor suppressor for gastric cancer H Takada et al 8052 around the ADAM23 gene or other mechanisms of Identification of gene(s) involved in the homozygous gene silencing such as homozygous deletion, nor deletion at 2q33.3 evaluated potential tumor suppressor activity of the gene product. To identify all potential target genes for the homozygous In the study reported here, we identified a homo- deletion at 2q33.3 in HSC43 cells, we first tried to define zygous loss of ADAM23 (2q33.3) by screening a panel of the region by fluorescence in situ hybridization (FISH), GC cell lines for copy-number aberrations using array- using five BACs as region-specific probes (Figure 2b, c). based comparative genomic hybridization (CGH-array) No red signals specific for RP11-225L15, which is spotted analysis relied on a custom-made bacterial artificial on the MCG whole genome array-4500, or for RP11- chromosome (BAC)-based array. We observed homo- 698O13 or RP11-655B6, were detected; however two zygous deletion of ADAM23 in primary GC, albeit other BACs, RP11-90D19 and RP11-17I6, showed four infrequently, as well. However, expression of this gene green signals each on isochromosomes in HSC43 cells was also lost in the majority of GC cell lines without (Figure 2b). The extent of the homozygous deletion homozygous loss, although it was present in normal estimated from these observations was approximately stomach. Those results suggested that when ADAM23 is 1.5 Mb, and information archived by human genome inactivated epigenetically or genetically, it may be databases (http://www.ncbi.nlm.nih.gov/ and http:// involved in gastric carcinogenesis. To evaluate that genome.ucsc.edu/) indicated that five genes are located hypothesis, we examined the role of DNA methylation in the defined region (Figure 2c). Homozygous deletion within the upstream CpG island in regard to expression of ADAM23, MDH1B, and KIAA0971 (gray arrows of ADAM23 in GC, and tested the effect of restored in Figure 2c) was detected only in HSC43 (1/32, 3.1%; ADAM23 on the growth of GC cells. Figure 2d), whereas GPR1 and CPO (black arrows in Figure 2c) were retained (data not shown).

Loss of ADAM23 expression in GC cell lines Results Next we determined expression levels of ADAM23, CGH-array analysis of GC cell lines MDH1B, and KIAA0971 by means of reverse transcriptase (RT)–PCR in all 32 GC cell lines and in Copy-number gains and losses were seen to some normal stomach. The HSC43 cell line as well as 13 lines degree in all of the 32 GC cell lines examined. Since without homozygous loss within 2q33.3 (41.9%) com- the most common genetic aberrations had already pletely lacked expression of ADAM23 mRNA; eight been identified in GC cell lines and primary tumors, other lines (25.8%) showed reduced expression com- we paid attention to more remarkable patterns of pared with normal stomach (Figure 2e). ADAM23 was chromosomal abnormalities, such as high-level amplifi- more or less expressed in various other tissues including cations and homozygous deletions, which are likely to stomach (Figure 2e). On the other hand, the RT–PCR be landmarks of oncogenes and tumor suppressor genes, product of MDH1B was absent less frequently, and that respectively. Tables 1 and 2 summarize the clones of KIAA0971 was not decreased at all. showing high-level amplifications (log2ratio >2.0) or homozygous deletions (log2ratio oÀ2.0), respectively. High-level amplifications were detected in 18 of the 32 Homozygous deletion of ADAM23 in LCM-treated GC cell lines, and 16 loci were represented (Table 1). primary tumors Among them, loci containing genes that had been To confirm that homozygous loss of ADAM23 was not reported already as putative oncogenes were detected an artefact that arose during establishment of the cell in more than two cell lines each: MYC, PVT1 (six lines, we performed genomic PCR using laser-capture cell lines), KSAM (four lines), KRAS (three lines), and microdissection (LCM)-treated primary gastric tumors, MET (two lines). and detected homozygous deletion of ADAM23 in one In all, 13 loci were identified as candidate regions for of the 39 tumors examined (2.6%, Figure 2f). homozygous deletions among 14 of the 32 GC cell lines (Table 2). The region containing CDKN2A/p16 (six Effect of demethylation by 5-aza 20-deoxycytidine on lines) was the most frequently deleted, and regions that ADAM23 expression harbored TEK (three lines) and WWOX (two lines; left panel in Figure 1a) were observed in at least two cell Aberrant methylation of DNA in 50 regulatory regions lines each. Genomic polymerase chain reaction (PCR) harboring a higher than expected number of CpG analysis (Figure 1b) also confirmed complete losses of dinucleotides is a key mechanism by which genes candidate genes that located within or around BACs relevant to cancer initiation and progression can be that had been detected as homozygous deletions by silenced (Baylin et al., 1998). To investigate whether analysis on the CGH-array. Among them, homozygous demethylation could restore expression of ADAM23 loss at 2q33.3, the location of ADAM23 (right panel in mRNA, we treated GC cells lacking ADAM23 expres- Figures 1a, b, and 2a), had never been documented in sion (SNU601 and SH101P4) with 10 mM of 5-aza GC before, prompting us to examine whether a tumor 20-deoxycytidine (5-aza-dCyd), a methyltransferase suppressor gene(s) involved in gastric tumorigenesis inhibitor, for 5 days. ADAM23 mRNA was induced in might lie within this region. both cell lines after treatment with 5-aza-dCyd

Oncogene ADAM23 as a potential tumor suppressor for gastric cancer H Takada et al 8053 Table 1 High-level ampliﬁcations (bg2 ratio>2.0) detected in GC cells lines by CGH-array analysis using MCG Whole Genome Array-4500 BAC Locusa Cell line (Total 32) Candidate geneb

Chr. Band Position n Name

RP11-142F22 8q22.2 chr8:100 624 092–100 777 935 6 HSC-39, 40A, 44PE, 45, 58, SH101P4 RP11-237F24 8q24.21 chr8:128 822 387–128 822 827 6 HSC-39, 40A, 44PE, 45, 58, SH101P4 MYC RP11-79I21 9q21.33 chr9:86 382 439–86 536 303 4 HSC40A, SNU216, MKN-7, 28 RP11-7P17 10q26.12 chr10:123 070 046–123 245 773 4 HSC-39, 40A, 43, KATO-III KSAM RP11-62L18 10q26.13 chr10:123 214 091–123 388 488 4 HSC-39, 40A, 43, KATO-III KSAM RP11-89K10 8q24.21 chr8:127 636 646–127 799 456 3 HSC-44PE, 45, 58 MYC RP11-89L16 8q24.21 chr8:129 633 607–129 784 954 3 HSC-39, 40A, SH101P4 MYC RP11-95L16 7q31.2 chr7:115 396 972–115 481 579 2 OKAJIMA, MKN45 MET RP11-91I20 8q12.1 chr8:61 578 650–61 769 195 2 HSC-39, 40A RP11-381K22 9q21.2 chr9:26 352 705–26 555 895 2 HSC60, KATO-III RP11-79A4 11p11.2 chr11:48 644 474–48 801 171 2 HSC-39, 40A RP11-77M17 11q12.1 chr11:57 109 511–57 277 679 2 HSC-39, 40A RP11-75H24 11q12.1 chr11:58 480 721–58 632 565 2 HSC-39, 40A RP11-407G6 12p13.33 chr12:2 518 681–2 637 273 2 HSC-44PE, 45 RP11-64J22 12p12.1 chr12:25 789 044–25 883 554 2 SH101P4, MKN1 KRAS RP11-507J18 15q21.2 chr15:48 735 230–48 930 297 2 HSC-39, 40A RP11-79E20 2p24.1 chr2:19 899 402–20 095 951 1 OKAJIMA SDC1 RP11-7908 2p24.1 chr2:19 899 407–20 095 968 1 OKAJIMA SDC1 RP11-241P3 3p22.1 chr3:42 284 364–42 440 027 1 HSC60 RP11-119L2 3p22.1 chr3:42 430 733–42 596 936 1 HSC60 RP11-451E18 6p12.3 chr6:51 252 528–51 432 711 1 OKAJIMA RP11-78C11 7q31.1 chr7:114 102 146–114 261 700 1 MKN45 MET RP11-51M22 7q31.2 chr7:115 435 012–115 597 337 1 MKN45 MET RP11-68E12 7q31.31 chr7:120 238 702–120 327 634 1 MKN45 MET RP11-3L10 7q31.32 chr7:120 631 254–120 795 956 1 MKN45 MET RP11-90L21 7q31.32 chr7:120 708 290–120 870 296 1 MKN45 MET RP11-112P4 7q31.32 chr7:121 128 048–121 317 230 1 MKN45 MET RP11-4801 7q31.32 chr7:121 434 700–121 553 715 1 MKN45 MET RP11-90E17 10q21.2 chr10:63 955 366–64 119 462 1 RERF-GC-1B RP11-79H15 10q21.2 chr10:62 550 598–62 706 766 1 RERF-GC-1B RP11-8E20 10q21.3 chr10:64 236 668–64 237 086 1 RERF-GC-1B RP11-14I14 10q21.3 chr10:64 866 985–64 879 715 1 RERF-GC-1B RP11-282N18 10q22.1 chr10:74 329 219–74 329 597 1 MKN1 RP11-95I16 10q26.12 chr10:122 625 701–122 772 534 1 HSC43 KSAM RP11-127B19 10q26.12 chr10:122 724 367–122 903 459 1 KATO-III KSAM RP11-48A2 10q26.12 chr10:129 686 061–129 851 191 1 KATO-III KSAM RP11-47D7 11p15.4 chr11:11 602 849–11 769 753 1 MKN45 RP11-47N15 11q23.1 chr11:111 597 475–111 791 409 1 MKN45 RP11-85G7 12p12.3 chr12:18 610 139–18 636 564 1 HSC60 KRAS RP11-877E17 12p12.1 chr12:25 986 009–26 163 998 1 MKN1 SSPN (KRAG) RP11-53C3 12p12.1 chr12:26 255 248–26 383 596 1 MKN1 SSPN (KRAG) RP11-666F17 12p11.23 chr12:26 671 081–26 857 010 1 MKN1 SSPN (KRAG) RP11-69I10 20q13.11 chr20:41 467 082–41 631 733 1 RERF-GC-1B MYBL2 RP11-29H19 20q13.12 chr20:42 038 797–42 086 477 1 RERF-GC-1B MYBL2 RP11-94A18 20q13.33 chr20:58 192 126–58 357 017 1 NUGC-2 RP11-278E23 22q11.21 chr22:19 862 624–19 958 793 1 Takigawa RP11-58H17 Xp11.23 chrX:49 260 873–49 460 929 1 HSC60 RP11-467H14 Xq22.1 chrX:101 269 026–101 540 274 1 HSC39 aBased on UCSC Genome Browser, May 2004 Assembly bRepresentative candidate oncogene located around BAC

(Figure 3a). In addition, we observed elevated expres- silencing of tumor suppressor genes. A 1067-bp sion of mRNA in SNU601 cells after treatment with sequence (Region II in Figure 3b) that included exon 1 trichostatin A (TSA), a histone deacetylase inhibitor, (À278 to IVS þ 352) of ADAM23 was identiﬁed as a alone, and enhancement of expression by 5-aza-dCyd CpG island by means of the genome database (http:// given along with TSA, although treatment with TSA www.ebi.ac.uk/emboss/cpgplot/). Others have reported alone had no affect on expression of ADAM23 in frequent hypermethylation in part of a slightly CpG-rich SH101P4 cells, indicating that histone deacetylation region approximately 0.5 Mb upstream from the trans- does play some role in transcriptional silencing of cription start site (479-bp sequence containing 22 CpG ADAM23 among methylated GC cells. sites, Region I in Figure 3b) in breast tumors (Costa et al., 2004). However, no potential association between the degree of methylation of the CpG island located Promoter activity of the ADAM23 CpG islands around exon 1 with expression of ADAM23 mRNA has Hypermethylation in CpG-rich promoter or exonic been investigated, nor has the structure of the gene’s regions is strongly associated with transcriptional promoter region or its methylation status in cells lacking

Oncogene ADAM23 as a potential tumor suppressor for gastric cancer H Takada et al 8054 Table 2 Homozygous deletions (log2 ratio oÀ2.0) detected in GC cell lines by CGII-array analysis using MCG Whole Genome Array-4500 BAC Locusa Cell line (Total 32) Candidate geneb

Chr. Band Position n Name

RP11-344A7 9p21.3 chr9:21 506 373–21 676 227 6 HSC-39, 40A, 44PE, 45, CDKN2A, MTAP SNU668, SH101P4 RP11-782K2 9p21.3 chr9:22 584 981–22 585 358 3 HSC-39, 40A, SNU668 CDKN2A, MTAP RP11-57P14 9p21.2 chr9:27 142 247–27 329 315 3 HSC-39, 40A, 58 TEK RP11-81B19 9p21.2 chr9:27 342 516–27 492 813 3 HSC-39, 40A, 58 TEK RP11-330J23 9p21.3 chr9:25 292 197–25 425 886 2 HSC-39, 40A TEK RP11-82P9 9p21.2 chr9:27 662 589–27 780 460 2 HSC-39, 40A TEK RP11-61L1 16q23.1 chr16:77 311 883–77 345 302 2 OKAJIMA, MKN45 WWOX RP11-20H6 1q25.3 chr1:179 194 208–179 346 598 1 HSC41 RGSL1, RGSL2, RNASEL, RGS16 RP11-225L15 2q33.3 chr2:207 216 189–207 216 506 1 HSC43 ADAM23, MDH1B, KIAA0971 RP11-94D19 3p14.2 chr3:60 756 956–60 927 079 1 HSC60 FHIT RP11-79K22 6q16.3 chr6:101 736 584–101 908 759 1 Takigawa GRIK2 RP11-90011 6q16.3 chr6:101 893 295–102 051 789 1 Takigawa GRIK2 RP11-341G2 9p24.3 chr9:1 121 123–1 241 689 1 HSC58 ANKRD15, DMRT1, DMRT3, DMRT2, SMARCA2, VLDLR, KCNV2, KIAA0020 RP11-48M17 9p24.3–24.4 chr9:2 136 329–2 296 367 1 HSC58 ANKRD15, DMRT1, DMRT3, DMRT2, SMARCA2, VLDLR, KCNV2, KIAA0020 RP11-399M15 9p22.2–22.1 chr9:18 459 220–18 672 439 1 RERF-GC-1B SH3GL2, ADAMTSL1, FLJ35283, MGC35182 RP11-81B11 9p22.1 chr9:18 502 807–18 657 760 1 RERF-GC-1B SH3GL2, ADAMTSL1, FLJ35283, MGC35182 RP11-11J1 9p21.3 chr9:22 417 726–22 579 721 1 SNU668 CDKN2A, MTAP RP11-44115 9p21.3 chr9:22 309 530–22 479 595 1 SNU668 CDKN2A, MTAP RP11-113D19 9p21.3 chr9:20 996 400–21 158 464 1 MKN45 CDKN2A, MTAP RP11-33015 9p21.3 chr9:22 823 087–22 897 484 1 HSC40A CDKN2A, MTAP RP11-90K7 13q14.2 chr13:48 069 306–48 242 971 1 HSC43 RB1, CHC1L, CYSLTR2, FNDC3, CAB39L RP11-321F3 14q21.3 chr14:46 999 937–47 165 909 1 SNU668 MAMDC1 RP11-185J20 16p13.2 chr16:6 588 011–6 756 741 1 MKN45 A2BP1 RP11-134G22 20p12.1 chr20:15 224 211–15 251 444 1 Takigawa RP11-91N21 21q11.2 chr21:14 886 780–15 043 867 1 Takigawa RP11-278E11 Xp11.1 chrX:56 628 610–56 716 147 1 RERF-GC-1B RP11-570J18 Xq11.2 chrX:62 685 940–62 742 514 1 RERF-GC-1B RP11-569H20 Xq11.2 chrX:62 869 152–63 042 142 1 RERF-GC-1B

aBased on UCSC Genome Browser, May 2004 Assembly. bGenes located around BAC, whose homozygous deletion was validated by genomic PCR

ADAM23 expression. Therefore, we tested the promoter first assessed the methylation status of the CpG island in activity of the CpG-rich region (Region I) using one GC-derived cell lines with or without expression of this 480-bp fragment (Fragment 1) and the CpG island gene, by means of bisulfite-PCR experiments using (Region II) using three 1026-, 348- and 677-bp appropriate restriction enzymes for each of the four fragments around exon 1 (À341 to þ 685, À341 to PCR fragments (Regions 1, 2-A, 2-B, and 2-C in þ 7, and þ 9to þ 685, respectively; Fragments 2–4) in Figure 3d). GC cells lacking ADAM23 expression but SH101P4 and AZ-521 cells (Figure 3b). without homozygous deletion of this gene (KATO-III, A remarkable increase in transcriptional activity was SNU601, and SH101P4) were found to be aberrantly observed in constructs containing Fragment 3 (7.4- and hypermethylated in all regions, although the degree of 12.7-fold increases, respectively, in SH101P4 and AZ- hypermethylation varied among CpG sites. On the other 521 cells), whereas constructs containing Fragment 1, hand, hypomethylation was seen in ADAM23-expres- representing the 479-bp region I analysed by Costa et al. sing GC cells (AZ-521, SNU484, RERF-GC-1B, and (2004), showed some but much lower transcriptional NUGC-2). Notably, the NUGC-2 cell line, which activity (3.3- and 3.9-fold increases) regardless of the expresses ADAM23, was relatively hypermethylated in expression status for ADAM23 (Figure 3c). Therefore, Region 1 (Figure 3d); this implied that Region 2, the region around Fragment 3 seems to be the most especially Region 2-A, rather than Region 1, is a more important sequence for promoting expression of critical site for epigenetic events affecting ADAM23 ADAM23. expression. We assessed the methylation status of each CpG dinucleotide within the ADAM23 CpG island (Region Methylation status of the ADAM23 CpG island in GC II) in more detail by bisulfite-sequencing. CpG sites on cell lines and primary tumors the CpG island tended to be extensively methylated in To explore the potential role of methylation of CpG ADAM23-nonexpressing cells without homozygous de- islands in silencing the transcription of ADAM23,we letions (SNU601, SH101P4), whereas almost all CpG

Oncogene ADAM23 as a potential tumor suppressor for gastric cancer H Takada et al 8055 GC tissues (cases 11, 26, 30 and 35), which showed a decreased ADAM23 expression compared with the corresponding noncancerous tissue, while no GCs (cases 8 and 13) without reduction of ADAM23 mRNA were methylated (Figure 4c). Although hypomethylated alleles were also present due to normal tissue compo- nents contaminating the tumors, corresponding noncancerous tissues showed hypomethylation of this region (Figure 4c). These ﬁndings indicate that methylation of ADAM23 Region 2-A did not arise during passage of GC cell lines in vitro, but rather may be a true cancer-related event during the pathogenesis of GC.

Suppression of cell growth after restoration of ADAM23 expression To gain further insight into the potential role of ADAM23 protein in gastric carcinogenesis, we investigated whether restoration of ADAM23 expression would suppress growth of GC cells in which the gene had been silenced. We performed colony-formation assays using the full coding sequence of ADAM23 cloned into a mammalian expression vector. As shown in Figure 4d, 3 weeks after transfection and subsequent selection of drug-resistant colonies, the numbers of large colonies produced by ADAM23-transfected SH101P4 Figure 1 (a) Representative copy-number profiles of chromosome cells decreased compared to cells containing empty 16 in OKAJIMA cells (left panel) and chromosome 2 in HSC43 vector. cells (right panel). Arrowhead and arrow indicate candidate spots showing patterns of homozygous deletion (log2 ratio oÀ2). (b) Images from genomic PCR experiments showing GAPDH, the functional control, and representative genes that were absent in at Discussion least one cell line for all cases examined. PLC, normal peripheral lymphocytes CGH-array technology allows high-throughput and quantitative analysis of copy-number changes at high resolution throughout the genome and provides many sites were unmethylated in ADAM23-expressing cells advantages over conventional CGH and other methods (AZ-521, Figure 3e). for reliably detecting and precisely mapping chromo- Since hypermethylation around Region 2-A somal changes (Snijders et al. , 2001; Inazawa et al., (Figure 3b) would likely be associated with silencing of 2004). In the study reported here, we used our MCG ADAM23 expression in GCs, we performed methyla- Whole Genome Array-4500, covering the entire human tion-specific PCR (MSP) with primer sets targeted for genome with a total of 4523 BACs, to examine 32 GC the most frequently methylated sequence (Figure 3b, e) cell lines for copy-number alterations within smaller in primary GCs as well as the rest of our GC cell lines regions. Since the most common chromosomal imbal- without homozygous loss of ADAM23. ances had already been identified in GC, we focused on Two representative cell lines without ADAM23 the more extreme aberrations, that is, local high-level expression were mostly methylated, whereas one amplifications and homozygous deletions. ADAM23-expressing cell line was unmethylated, as The results of the new CGH-array analysis were expected (Figure 4a). Target sequence was methylated mostly in accord with our previous CGH/CGH-array in 19 (61.3%) of the 31 GC cell lines (data not shown). data (Fukuda et al., 2000; Takada et al., 2005) and with Moreover, we detected its methylation in seven of 16 other published CGH/CGH-array data derived from primary GC tissues (43.8%); four of those seven samples analysis of primary gastric tumors (Sakakura et al., were not methylated in the corresponding noncancerous 1999; Tay et al., 2003; Kimura et al., 2004; Weiss et al., stomach tissues (Figure 4a). RT–PCR using these four- 2004; Gorringe et al., 2005). That is, loci that were paired samples demonstrated that ADAM23 expression frequently detected as high-level amplifications in our was nevertheless lower in tumor tissues compared with study contained the same genes whose amplifications the noncancerous tissues (Figure 4b). To confirm the had been known to play important roles in pathogenesis results of MSP analysis quantitatively, we performed of GC as putative oncogenes, such as MYC, PVT1, bisulfite-sequencing of Region 2-A in individual alleles KSAM, MET,andKRAS (Tahara, 1995; Koo et al., from representative GC samples and their correspond- 2000; Peng et al., 2003). On the other hand, homozygous ing noncancerous tissues. Aberrant hypermethylation of losses were more frequently identified by MCG Whole Region 2-A was detected at frequencies of 20–50% in Genome Array-4500 than by MCG Cancer Array-800

Oncogene ADAM23 as a potential tumor suppressor for gastric cancer H Takada et al 8056

Figure 2 (a) Representative duplicate CGH-array image of the HSC43 cell line. A remarkable decrease in copy-number ratio (log2ratio) of ADAM23 at 2q33.3 was detected as a clear red signal. A detailed copy-number proﬁle of chromosome 2 in this cell line is shown in Figure 1a. (b) Representative FISH images from probes RP11-698O13 and RP11-655B6, which contain the ADAM23 gene (red signals) and with RP11-90D19 and RP11-17I6 as a control (green signals) hybridized to metaphase chromosomes from the HSC43 cell line. Absence of red signals indicates homozygous loss of ADAM23.(c) Map of 2q33.3 covering the region homozygously deleted in the HSC43 cell line. BACs with no signals and normal signals in HSC43 cells are shown by vertical white and black bars, respectively. The homozygously deleted region in HSC43 cells, as determined by FISH analysis, is indicated as a vertical white closed arrow. Five genes located within this region are indicated as black or gray arrows that indicate positions and directions of transcription. (d) Homozygous deletions of ADAM23, MDH1B, and KIAA0971, not but GPR1 and CPO (data not shown), in one GC cell line (arrowhead), detected by genomic PCR analysis. 1: MKN1, 2: MKN7, 3: MKN28, 4: MKN45, 5: MKN74, 6: KATO-III, 7: OKAJIMA, 8: OCUM-1, 9: HSC39, 10: HSC40A, 11: HSC41, 12: HSC42, 13: HSC43, 14: HSC44PE, 15: HSC45, 16: HSC57, 17: HSC58, 18: HSC60, 19: HSC64, 20: NUGC-2, 21: NUGC-3, 22: NUGC-4, 23: RERF-GC-1B, 24: AZ-521, 25: SNU216, 26: SNU484, 27: SNU601, 28: SNU638, 29: SNU668, 30: SNU719, 31: SH101P4, and 32: Takigawa. (e) Expression of ADAM23, MDH1B, and KIAA0971 in GC cell lines and normal stomach, detected by RT–PCR analysis. Arrowhead indicates the cell line with the homozygous deletion indicated in Figure 1d. Expression of MDH1B and KIAA0971 mRNAs, not but ADAM23, was observed to some degree in most GC cell lines. Note that 13 of the 31 cell lines without homozygous deletion of ADAM23 (41.9%) showed almost complete silencing of this gene and 8 (25.8%) showed decreased expression. St: stomach, H: heart, Br: brain, Te: testis, Sp: spleen, SI: small intestine, and Co: colon. (f) Homozygous deletion of ADAM23, detected by genomic PCR analysis in one of the 39 LCM-treated primary gastric tumors examined (For color version see online)

(Takada et al., 2005), probably because homozygous contained one potential candidate, ADAM23, whose deletions are expected to be quite small and a higher- epigenetic silencing through DNA methylation had been density array may be necessary to fully reveal them. reported in breast tumors (Costa et al., 2004). Indeed, Loci with homozygous losses can flag regions likely to RT–PCR analysis of genes within this region identified harbor tumor suppressor genes; in fact, some of the loci ADAM23 as the most probable target, because expres- we detected in our GC cell lines contained representative sion of this gene was frequently silenced in our panel of tumor suppressor genes such as FHIT, CDKN2A/p16, GC cell lines, whereas it was expressed in normal RB1 and WWOX, which had been reported already in stomach and other tissues, even though its expression connection with various tumors including GC (Akama level was various among tissues. Since genomic PCR et al., 1996; Huiping et al., 2002; Feakins et al., 2003; experiments demonstrated infrequent homozygous dele- Aqeilan et al., 2004). We chose to focus on the complete tion of ADAM23 in primary gastric tumors as well, this loss at 2q33.3 observed in one of the cell lines, because gene is likely to be inactivated mainly by epigenetic (a) this locus was hemizygously deleted in five of the mechanisms such as DNA methylation. other 31 lines (15.6%, data not shown) but it was not Costa et al. (2004) reported that methylation in a previously known to harbor any tumor suppressor genes region located approximately 0.5 kb upstream from the involved in the pathogenesis of GC, and (b) this region transcription initiation site of the ADAM23 gene was

Oncogene ADAM23 as a potential tumor suppressor for gastric cancer H Takada et al 8057

Figure 3 (a) Representative results of RT–PCR analysis to reveal ADAM23 expression in SNU601 and SH101P4 cells with or without treatment with 5-aza-dCyd (10 mM) for 5 days and/or TSA (10 ng/ml) for 24 h. (b) Schematic map of the 480-bp CpG-rich region (À1027 to –548, Region I) previously analysed by Costa et al. (2004) and a 1067-bp CpG island (À278 to þ 789, Region II) in ADAM23. CpG sites are indicated by vertical ticks on the expanded axis. Exon 1 is indicated by an open box, and the transcription start site is marked at þ 1. The fragments examined in a promoter assay (Fragments 1–4) are indicated by heavy black lines. The regions examined in bisulfite-PCR analysis and bisulfite-sequencing (Region 1, and Region 2 divided into three regions, namely 2-A, 2-B, and 2-C) are indicated by horizontal gray bars. For bisulfite-PCR analysis, restriction sites for HhaI and BstUI are indicated respectively by black or gray downward arrowheads. Arrows at the bottom show the positions of primers for MSP. (c) Promoter activity of the ADAM23 CpG-rich region and CpG island. pGL3 basic empty vectors (mock) and constructs containing one of four different sequences upstream or around exon 1 of ADAM23 (Fragments 1–4) were transfected into SH101P4 and AZ-521 cells. Luciferase activities were normalized versus an internal control. The data presented are the means7s.e.m. of three separate experiments, each performed in triplicate. (d) Representative results of bisulfite-PCR analyses of ADAM23 Regions 1, 2-A, 2-B and 2-C in GC cell lines after digestion with HhaIorBstUI. Arrows show fragments specifically restricted in the sites recognized as methylated CpGs. (e) Results of bisulfite-sequencing of the ADAM23 Region 2, examined in ADAM23-expressing cell lines ( þ ) and ADAM23- nonexpressing cell lines (À). Using three fragments (Regions 2-A, 2-B and 2-C), 124 of the 127 CpG sites within the CpG island (Region II) were sequenced. Each square indicates a CpG site: open squares, unmethylated; solid squares, methylated strongly associated with its altered expression in fragment seems to be inversely correlated with expres- breast tumor cell lines. However, a site that fulfills sion of the gene; and (c) hypermethylation of this the accepted criteria for CpG islands exists in another region, along with decreased gene expression, occurs in area of the ADAM23 gene, and little was known about primary gastric tumors compared with noncancerous that sequence. In the present study, we demonstrated gastric tissue. These factors indicated that aberrant that (a) an actual CpG island upstream of the hypermethylation within a CpG island having promoter transcriptional start site of the ADAM23 gene is the activity is likely to be one of the important mechanisms most important sequence for promoting expression of leading to inactivation of the ADAM23 gene in gastric this gene; (b) the methylation status around that neoplasm.

Oncogene ADAM23 as a potential tumor suppressor for gastric cancer H Takada et al 8058

Figure 4 (a) Representative results of MSP analysis of ADAM23 Region 2-A in primary GC tissues (T) and corresponding noncancerous stomach tissues (N). Parallel amplification reactions were performed using primers specific for unmethylated (U) or methylated (M) DNA. (b) Expression of ADAM23 in six primary cases of GC and paired noncancerous tissues. In cases 11, 26, 30 and 35, where tumors had shown aberrant hypermethylation of ADAM23 Region 2-A (Figure 3b, e), tumors showed only traces of ADAM23 expression compared with the noncancerous tissues; cases 8 and 13, where both tumors and noncancerous tissues were unmethylated in Region 2-A, showed no differences in ADAM23 expression. RT–PCR products were calculated in a semiquantitative manner by densitometry. Expression level of ADAM23 in each sample was normalized on the basis of the respective GAPDH intensity. % expression indicates a relative expression level of ADAM23 in tumor samples compared with corresponding noncancerous tissues. (c) Methylation status of ADAM23 Region 2-A, determined by bisulfite-sequencing in tumor samples (T) and corresponding noncancerous tissues (N) of representative cases analysed in (a) and (b). See legend for Figure 3e for interpretation. (d) Effect of restoration of ADAM23 on growth of GC cells. An Myc-tagged construct containing ADAM23 (pCMV-Tag3-ADAM23) or empty vector (pCMV-Tag3-mock) as a control was transfected into SH101P4 cells, which lack expression of the ADAM23 gene because the CpG island is methylated. Western blot analysis using 10 mg of protein extract and anti-Myc antibody demonstrated that cells were transiently transfected with pCMV-Tag3-ADAM23 expressed Myc-tagged ADAM23 protein (arrow in upper panel). At 3 weeks after transfection and subsequent selection of drug-resistant colonies in six-well plates (Imoto et al., 2003), the colonies formed by ADAM23-transfected cells were less numerous than those formed by mock-transfected cells (middle panel). Lower panel, quantitative analysis of colony formation. Colonies larger than 2 mm were counted, and results are presented as the means7s.e. (bars) of three separate experiments, each performed in triplicate

According to analysis of primary samples, hyper- however, the unique structure of ADAM molecules methylation of ADAM23 CpG island with reduced allows us to postulate an ability of ADAM23 to expression of this gene seems to occur in a tumor- function as an adhesion molecule. The disintegrin-like speciﬁc manner; nevertheless, some cases showed its domain of ADAM23 promotes attachment of human methylation both in tumors and in the corresponding cells of neural origin, such as neuroblastoma or noncancerous tissues (data not shown). Since methyla- astrocytoma cells (Cal et al., 2000). Therefore loss of tion of several genes appears to follow a chronological ADAM23, which is likely to play an important role with order during carcinogenesis in stomach tissue (Kang regard to cell–cell and cell–extracellular matrix interac- et al., 2003), transcriptional silencing of tumor-related tions in gastric tissue as well, might be essential for the genes, including ADAM23, by aberrant methylation of progression of GC. Additional studies can be designed CpG islands is likely to be established during an early to unravel the carcinogenic consequences of loss of stage in the multistep process of gastric carcinogenesis. ADAM23 function in gastric tissue. The exact roles of ADAM proteins, including ADAM23, in carcinogenesis remain largely unknown. Our colony-formation assays using ADAM23-trans- Materials and methods fected GC cells revealed growth-suppressing and/or antiproliferative activity of the ADAM23 protein; Cell lines and primary tumors neither had ever been demonstrated before. So far, little A total of 32 GC cell lines were employed. Features of those information is available about the functional conse- cell lines except Takigawa were described in the previous quences of ADAM23 loss in terms of carcinogenesis; reports (Fukuda et al., 2000; Takada et al., 2005). All cell lines

Oncogene ADAM23 as a potential tumor suppressor for gastric cancer H Takada et al 8059 were maintained in RPMI-1640 supplemented with 10% phosphate dehydrogenase (GAPDH) was amplified at the same fetal bovine serum and 100 units/ml penicillin/100 mg/ml time to estimate the efficiency of cDNA synthesis. streptomycin. Primary tumor samples were obtained from 55 patients who Drug treatment were undergoing surgery at the National Defense Medical College in Saitama or the National Cancer Center Hospital in Cells were treated with 10 mM of 5-aza-dCyd for 5 days and/or Tokyo, Japan, with prior written consent from each patient in with 10 ng/ml TSA for varying periods. For the synergistic the formal style and after approval by the local ethics study, 10 mM of 5-aza-dCyd was present in the cultures for committees. Samples from 39 patients with well-differentiated 5 days, and 10 ng/ml TSA was added for the last 24 h. adenocarcinomas were paraffin-embedded for LCM after 24 h methanol fixation, as described elsewhere (Noguchi et al., Promoter reporter assay 1997). From the other 16 patients we obtained paired samples from cancerous and adjacent noncancerous tissues, which were We obtained by PCR three DNA fragments (1026-, 348- and frozen immediately in liquid nitrogen and stored at À801C 677-bp, Fragments 2–4) around the CpG island of ADAM23 until required. None of the patients had been administered predicted by the CpGPLOT program (http://www.ebi. ac.uk/ preoperative radiation, chemotherapy, or immunotherapy. emboss/cpgplot/) and a 480-bp slightly CpG-rich region (Fragment 1) upstream from exon 1 of ADAM23 as a target for DNA methylation (Figure 3b). Each fragment was ligated CGH-array analysis into the pGL3-Basic vector (Promega, Madison, WI, USA), Hybridizations were carried out as described elsewhere and an equal amount of each construct was introduced into (Sonoda et al., 2004) with minor modifications. Briefly, test cells along with an internal control vector (pRL-hTK, and reference DNAs were labeled, respectively, with Cy3- and Promega), using FuGENE 6 (Roche Diagnostics, Tokyo, Cy5-dCTP, precipitated together with ethanol in the presence Japan). A pGL3-Basic vector without insert served as a of Cot-1 DNA, redissolved in a hybridization mixture (50% negative control. Firefly luciferase and Renilla luciferase formamide, 10% dextran sulfate, 2 Â standard saline citrate activities were each measured 36 h after transfection, using (SSC), 4% sodium dodecyl sulfate (SDS), pH 7), and the Dual-Luciferase Reporter Assay System (Promega); denatured at 751C for 10 min. After a 40-min preincubation relative luciferase activities were calculated and normalized at 421C, each mixture was applied to array slides and versus Renilla luciferase activity. incubated at 501C for 10 min, 461C for 10 min, and 431C for 60 h in a hybridization machine (GeneTAC; Harvard Bisulfite-PCR analysis and bisulfite-sequencing Bioscience, Holliston, MA, USA). Hybridized slides were Genomic DNAs were treated with sodium bisulfite using an washed once in a solution of 50% formamide, 2x SSC (pH 7.0) EZ DNA Methylation kit (Zymo Research, Orange, CA, 1 1 for 10 min at 50 C and in 1 Â SSC for 10 min at 42 C, then USA), and subjected to PCR using primer sets designed to scanned with a GenePix 4000B (Axon Instruments, Foster amplify the regions of interest. To analyse the CpG island City, CA, USA). Acquired images were analysed with GenePix (Region II in Figure 3b), that sequence was divided into three Pro 4.1 imaging software (Axon Instruments). Fluorescence PCR fragments (Regions 2-A, 2-B, and 2-C in Figure 3b). ratios were normalized so that the mean of the middle third of For bisulfite-PCR analysis, PCR products were digested log2ratios across the array was zero. Average ratios that with HhaIorBstUI, which recognizes sequences unique to the deviated significantly (>2 s.d.) from zero were considered methylated alleles but cannot recognize unmethylated alleles, abnormal. and electrophoresed. For bisulfite-sequencing, the PCR products were subcloned and then sequenced. Fluorescence in situ hybridization Metaphase chromosomes were prepared from normal male Methylation-specific PCR lymphocytes and from each GC cell line. FISH analyses were Genomic DNA treated with sodium bisulfite was amplified performed as described previously (Fukuda et al., 2000), using using primers specific to the methylated and unmethylated BACs located around the region of interest as probes. forms of DNA sequences of interest. DNAs from cell line AZ- 521 and from peripheral blood lymphocytes of a healthy male, Screening of primary tumors for homozygous deletions which were recognized as unmethylated forms by bisulfite- by genomic PCR sequencing analysis, were used as negative controls for MSP assay; DNAs from cell lines SNU601 and SH101P4, recog- Methanol-fixed, paraffin-embedded tissues were prepared for nized as methylated, served as positive controls for methylated LCM with a PixCell II LCM system (Arcturus Engineering, alleles. PCR products were visualized on 3% agarose gels Mountain View, CA, USA). Each genomic DNA was isolated stained with ethidium bromide. in lysis buffer (10 mM Tris-HCl at pH 7.5, 1 mM EDTA, 0.5% SDS) and amplified by adaptor-ligation-mediated PCR after Transient transfection, Western blotting, and colony-formation end-filling, as described by Tanabe et al. (2003). assays We screened DNAs from 39 primary GCs for homozygous losses by genomic PCR. All the relevant primer sequences are A plasmid expressing a Myc-tagged ADAM23 (pCMV-Tag3- available on request. ADAM23) was obtained by cloning the full coding sequence of ADAM23 into the pCMV-Tag3 vector (Stratagene, La Jolla, CA, USA) in-frame, along with the Myc-epitope. pCMV- Reverse transcriptase–polymerase chain reaction Tag3-ADAM23, or the empty vector (pCMV-Tag3-mock) Single-stranded cDNAs were generated from total RNAs control, were transfected into cells for colony-formation assays using the SuperScriptt First-Strand Synthesis System (Invi- as described elsewhere (Imoto et al., 2003). Expression of trogen, Carlsbad, CA, USA), and amplified with primers ADAM23 protein in transiently transfected cells was specific for each gene. The gene encoding glyceraldehyde-3- confirmed 48 h after transfection by Western blot analysis

Oncogene ADAM23 as a potential tumor suppressor for gastric cancer H Takada et al 8060 using anti-Myc antibody (Cell Signaling Technology, Beverly, Center Research Institute) for providing GC cell lines, and Ai MA, USA), as described elsewhere (Yuki et al., 2004). After 3 Watanabe for technical assistance. weeks of incubation with appropriate concentrations of G418 This work was supported by Grants-in-Aid for Scientiﬁc in six-well plates, cells were ﬁxed with 70% ethanol and stained Research on Priority Areas (C) from the Ministry of with crystal violet. Education, Culture, Sports, Science, and Technology, Japan; by a Grant-in-Aid from Core Research for Evolutional Science Acknowledgements and Technology (CREST) of the Japan Science and Techno- We are grateful to Professor Yusuke Nakamura (Human logy Corporation (JST); by a Center of Excellence (COE) Genome Center, The Institute of Medical Science, The program for Frontier Research on Molecular Destruction and University of Tokyo) for his continuous encouragement Reconstitution of Tooth and Bone; by the program for throughout this work. We thank Professor Jae-Gahb Park promotion of Fundamental Studies in Health Sciences of the (Laboratory of Cell Biology, Cancer Research Institute, Seoul Pharmaceuticals and Medical Devices Agency (PMDA); and National University College of Medicine) and Dr Kazuyoshi by the Third Term Comprehensive Control Research for Yanagihara (Central Animal Laboratory, National Cancer Cancer of the Ministry of Health, Labour and Welfare.

References

Akama Y, Yasui W, Kuniyasu H, Yokozaki H, Akagi M, Noguchi M, Furuya S, Takeuchi T and Hirohashi S. (1997). Tahara H, Ishikawa T and Tahara E. (1996). Jpn J Cancer Pathol Int, 47, 685–691. Res., 87, 824–830. O’Shea C, McKie N, Buggy Y, Duggan C, Hill AD, Aqeilan RI, Kuroki T, Pekarsky Y, Albagha O, Trapasso F, McDermott E, O’Higgins N and Duffy MJ. (2003). Int J Baffa R, Huebner K, Edmonds P and Croce CM. (2004). Cancer, 105, 754–761. Clin Cancer Res., 10, 3053–3058. Peng DF, Sugihara H, Mukaisho K, Tsubosa Y and Hattori T. Baylin SB, Herman JG, Graff JR, Vertino PM and Issa JP. (2003). J Pathol., 201, 439–450. (1998). Adv Cancer Res., 72, 141–196. Sakakura C, Mori T, Sakabe T, Ariyama Y, Shinomiya T, Cal S, Freije JM, Lopez JM, Takada Y and Lopez-Otin C. Date K, Hagiwara A, Yamaguchi T, Takahashi T, (2000). Mol Biol Cell, 11, 1457–1469. Nakamura Y, Abe T and Inazawa J. (1999). Genes Carl-McGrath S, Lendeckel U, Ebert M, Roessner A and Chromosomes Cancer, 24, 299–305. Rocken C. (2005). Int J Oncol., 26, 17–24. Seals DFand Courtneidge SA. (2003). Genes Dev., 17, 7–30. Costa FF, Verbisck NV, Salim AC, Ierardi DF, Pires LC, Shintani Y, Higashiyama S, Ohta M, Hirabayashi H, Sasahara RM, Sogayar MC, Zanata SM, Mackay A, Yamamoto S, Yoshimasu T, Matsuda H and Matsuura N. O’Hare M, Soares F, Simpson AJ and Camargo AA. (2004). Cancer Res., 64, 4190–4196. (2004). Oncogene, 23, 1481–1488. Snijders AM, Nowak N, Segraves R, Blackwood S, Brown N, Feakins RM, Nickols CD, Bidd H and Walton SJ. (2003). Conroy J, Hamilton G, Hindle AK, Huey B, Kimura K, Hum Pathol., 34, 1276–1282. Law S, Myambo K, Palmer J, Ylstra B, Yue JP, Gray JW, Fukuda Y, Kurihara N, Imoto I, Yasui K, Yoshida M, Jain AN, Pinkel D and Albertson DG. (2001). Nat Genet., Yanagihara K, Park JG, Nakamura Y and Inazawa J. 29, 263–264. (2000). Genes Chromosomes Cancer, 29, 315–324. Sonoda I, Imoto I, Inoue J, Shibata T, Shimada Y, Chin K, Gorringe KL, Boussioutas A, Bowtell DD and Melbourne Imamura M, Amagasa T, Gray JW, Hirohashi S and Gastric Cancer Group. Peter Mac Micro Array Facility. Inazawa J. (2004). Cancer Res, 64, 3741–3747. (2005). Genes Chromosomes Cancer, 42, 247–259. Tahara E. (1995). World J Surg., 19, 484–490. Huiping C, Kristjansdottir S, Bergthorsson JT, Jonasson JG, Takada H, Imoto I, Tsuda H, Sonoda I, Ichikura T, Magnusson J, Egilsson V and Ingvarsson S. (2002). Eur J Mochizuki H, Okanoue T and Inazawa J. (2005). Cancer Cancer, 38, 728–735. Sci., 96, 100–110. Imoto I, Yuki Y, Sonoda I, Ito T, Shimada Y, Imamura M Tanabe C, Aoyagi K, Sakiyama T, Kohno T, Yanagitani N, and Inazawa J. (2003). Cancer Res., 63, 5691–5696. Akimoto S, Sakamoto M, Sakamoto H, Yokota J, Ohki M, Inazawa J, Inoue J and Imoto I. (2004). Cancer Sci., 95, Terada M, Yoshida T and Sasaki H. (2003). Genes 559–563. Chromosomes Cancer, 38, 168–176. Kang GH, Lee HJ, Hwang KS, Lee S, Kim JH and Kim JS. Tay ST, Leong SH, Yu K, Aggarwal A, Tan SY, Lee CH, (2003). Am J Pathol., 163, 1551–1556. Wong K, Visvanathan J, Lim Dennis, Wong WK, Kimura Y, Noguchi T, Kawahara K, Kashima K, Daa T and Soo KC, Kon OL and Tan P. (2003). Cancer Res., 63, Yokoyama S. (2004). Mod Pathol., 17, 1328–1337. 3309–3316. Kodama T, Ikeda E, Okada A, Ohtsuka T, Shimoda M, Weiss MM, Kuipers EJ, Postma C, Snijders AM, Pinkel D, Shiomi T, Yoshida K, Nakada M, Ohuchi E and Okada Y. Meuwissen SG, Albertson D and Meijer GA. (2004). Cell (2004). Am J Pathol., 165, 1743–1753. Oncol., 26, 307–317. Koo SH, Kwon KC, Shin SY, Jeon YM, Park JW, Kim SH Wu E, Croucher PI and McKie N. (1997). Biochem Biophys and Noh SM. (2000). Cancer Genet Cytogenet., 117, 97–103. Res Commun., 235, 437–442. McCulloch DR, Akl P, Samaratunga H, Herington AC and Yuki Y, Imoto I, Imaizumi M, Hibi S, Kaneko Y, Amagasa T Odorico DM. (2004). Clin Cancer Res., 10, 314–323. and Inazawa J. (2004). Cancer Sci., 95, 503–507.

Oncogene