Oncogene (2007) 26, 4699–4713 & 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00 www.nature.com/onc ORIGINAL ARTICLE Frequent epigenetic inactivation of SFRP and constitutive activation of Wnt signaling in gastric cancer

M Nojima1,8, H Suzuki1,2,8, M Toyota1,3,4, Y Watanabe5, R Maruyama1, S Sasaki1, Y Sasaki3, H Mita3, N Nishikawa6, K Yamaguchi6, K Hirata6, F Itoh5, T Tokino3, M Mori2, K Imai7 and Y Shinomura1

1First Department of Internal Medicine, Sapporo Medical University, Sapporo, Japan; 2Department of Public Health, Sapporo Medical University, Sapporo, Japan; 3Department of Molecular Biology, Cancer Research Institute, Sapporo Medical University, Sapporo, Japan; 4PRESTO JST, Kawaguchi, Japan; 5Division of Gastroenterology and Hepatology, Department of Internal Medicine, St Marianna University School of Medicine, Kasawaki, Japan; 6First Department of Surgery, Sapporo Medical University, Sapporo, Japan and 7Sapporo Medical University, Sapporo, Japan

Activation of Wnt signaling has been implicated in gastric Introduction tumorigenesis, although mutations in APC (adenomatous polyposis coli), CTNNB1 (b-catenin) and AXIN are seen The two known Wnt signaling pathways play essential much less frequently in gastric cancer (GC) than in roles in the regulation of cell proliferation, patterning colorectal cancer. In the present study, we investigated and fate determination during normal developmen- the relationship between activation of Wnt signaling tal processes (Taketo, 2004). The canonical pathway and changes in the expression of secreted frizzled-related operates by stabilizing b-catenin, whereas the non- (SFRP) family genes in GC. We frequently canonical pathway does not require b-catenin signaling observed nuclear b-catenin accumulation (13/15; 87%) and controls cell movement during morphogenesis. and detected the active formof b-catenin in most (12/16; When the Wnt pathway is in its resting state, b-catenin 75%) GC cell lines. CpG methylation-dependent silenc- is phosphorylated by glycogen synthase kinase 3b ing of SFRP1, SFRP2 and SFRP5 was frequently seen (GSK3b) within a protein complex that also includes among GC cell lines (SFRP1, 16/16, 100%; SFRP2, 16/ casein kinase 1, adenomatous polyposis coli (APC) and 16, 100%; SFRP5, 13/16, 81%) and primary GC Axin (Gregorieff and Clevers, 2005). Phosphorylated b- specimens (SFRP1, 42/46, 91%; SFRP2, 44/46, 96%; catenin is immediately degraded via the ubiquitin SFRP5, 30/46, 65%), and treatment with the DNA proteasome pathway. Wnt binding to Frizzled (Fz) methyltransferase inhibitor 5-aza-20-deoxycytidine rapidly results in the activation of Dishevelled (Dsh), which restored SFRP expression. Ectopic expression of SFRPs inhibits the activity of GSK3b, resulting in dephos- downregulated T-cell factor/lymphocyte enhancer factor phorylation and stabilization of b-catenin, enabling it to transcriptional activity, suppressed cell growth and accumulate within the nucleus, where it interacts with induced apoptosis in GC cells. Analysis of global expression members of the T-cell factor/lymphocyte enhancer revealed that overexpression of SFRP2 repressed Wnt factor (TCF/LEF) family of transcription factors to target genes and induced changes in the expression of stimulate the expression of target genes (Taketo, 2004). numerous genes related to proliferation, growth and The Wnt/b-catenin pathway is known to be involved apoptosis in GC cells. It thus appears that aberrant SFRP in tumorigenesis involving several cell types. For methylation is one of the major mechanisms by which Wnt instance, the first mammalian Wnt isoform was identi- signaling is activated in GC. fied based on its ability to promote mouse mammary Oncogene (2007) 26, 4699–4713; doi:10.1038/sj.onc.1210259; tumorigenesis. APC is mutated in the germ line of published online 5 February 2007 individuals with familial adenomatous polyposis coli, and acquired mutations in APC, CTNNB1 (b-catenin) Keywords: SFRP; methylation; gastric cancer; Wnt or AXIN2 initiate the majority of sporadic colorectal signaling; apoptosis cancers (CRCs) (Gregorieff and Clevers, 2005), at least in part because mutant APC and Axin are unable to assist GSK3b in phosphorylating/inactivating b-catenin. Similarly, mutations that lead to amino-acid substitu- tion of the phosphorylated residues of b-catenin stabilize Correspondence: Dr H Suzuki, First Department of Internal Medicine, the protein. Either of these disruptions can lead to Sapporo Medical University, S1, W16, Chuo-Ku, Sapporo 060-8543, constitutive activation of the Wnt pathway. Japan. Secreted frizzled-related (SFRPs) are a E-mail: [email protected] 8These authors contributed equally to this work. family of five secreted glycoproteins that have been Received 17 February 2006; revised 24 November 2006; accepted 4 December identified as modulators of the Wnt signaling pathway 2006; published online 5 February 2007 (Jones and Jomary, 2002). SFRP family members all SFRPs methylation in gastric cancer M Nojima et al 4700 contain an N-terminal domain homologous to the in APC or CTNNB1 are not often found in sporadic cysteine-rich domain (CRD) of the Wnt receptor Fz GC. For instance, our group previously examined 35 and a C-terminal domain with some homology to netrin. primary GC tissue samples and found no mutations in In contrast to Fz family proteins, SFRPs lack a exon 3 of CTNNB1, whereas nuclear and cytoplasmic transmembrane region and the cytoplasmic domain localization of b-catenin was observed in 23% of the required for signal transduction into the cells. This tumors (Sasaki et al., 2001). In the present study, enables SFRPs to downregulate Wnt signaling by therefore, our aim was to clarify the respective roles competing with Fz for Wnt binding via its CRD domain played by Wnt signaling and changes in SFRP expres- or by binding directly to Fz (Jones and Jomary, 2002). sion in GC. Up to now, it has been thought that mutations in APC, AXIN or CTNNB1 can cause constitutive signal- ing independent of the upstream signal from Wnt. In addition, we and others recently showed that SFRP Results family genes are frequent targets of aberrant DNA methylation in CRCs (Suzuki et al., 2002; Caldwell Frequent activation of the canonical Wnt signaling et al., 2004), and we further showed that restoration of pathway in GC SFRP1, SFRP2 and SFRP5 in CRC cells attenuates To investigate the Wnt signaling status in GC, we Wnt signaling, even in the presence of downstream initially analysed endogenous expression of b-catenin, as mutations (Suzuki et al., 2004). Since then, down- it is well known that canonical Wnt signaling allows regulation and methylation of SFRP genes have been stabilization of b-catenin and its accumulation in the identified in a variety of malignancies, including bladder nucleus (Taketo, 2004). Immunofluorescent analysis cancer (Stoehr et al., 2004), prostate cancer (Lodygin of the cellular distribution of endogenous b-catenin et al., 2005), endometrial cancer (Risinger et al., 2005), revealed nuclear accumulation of the protein in 13 of 15 lung cancer (Fukui et al., 2005), breast cancer (Veeck (87%) GC cells examined, whereas only six cells showed et al., 2006), mesothelioma (Lee et al., 2004) and chronic a membranous staining pattern (Table 1; Figure 1a). lymphocytic leukemia (Liu et al., 2006), which strongly Moreover, Western analysis using an anti-b-catenin suggests SFRPs function as tumor suppressor genes. antibody revealed detectable levels of endogenous b- Gastric cancer (GC) is the second most common catenin in all of the GC cell lines tested (Figure 1b). cause of death from cancer among both men and women Because activation of Wnt signaling leads to activation in the world. Wnt signaling is involved in cell prolifera- of b-catenin, we also utilized an antibody specific for tion during development of the gut, and activation of the active form of b-catenin, which is dephosphorylated the Wnt signaling pathway has been implicated in on Ser37 and Thr41. We detected active b-catenin in 12 gastric tumorigenesis (Yuasa, 2003). Indeed, persons of 16 cells, although the levels of expression varied carrying a germ-line mutation in APC have a 10-fold considerably among the cell lines (Figure 1b; Table 1). greater risk of developing GC than those without the When we then analysed TCF/LEF-regulated transcrip- mutation (Offerhaus et al., 1992). However, mutations tional activity in GC cells using a TCF/LEF-responsive

Table 1 Summary of Wnt signaling analysis results, APC and CTNNB1 status, expression and methylation status of SFRPs in GC cell lines b-Catenin staining Active-b-catenin TCF activity Genetic alterations Expression status Methylation status (OT/OF) Nuclear Membranous APC b-Catenin SFRP1 SFRP2 SFRP5 SFRP1 SFRP2 SFRP5

MKN1 + ÀÀ0.74 Wild Wild ÀÀ+M M U MKN7 ++ À +++ 2.31 Wild Wild ÀÀÀMMM MKN74 ++ + +++ 9.94 Mutant Wild ÀÀÀMMM SH101 ++ À + 1.83 Wild Wild ÀÀÀMMU SNU1 ++ ÀÀ0.33 Wild Wild ÀÀÀU/M U/M M SNU638 ++ À + 6.94 Wild Mutant ÀÀÀM M U/M KatoIII + À +++ 97.4 Wild Amplified ÀÀÀ+a MMM JRST ++ À ++ 3.28 Wild Wild ÀÀÀM M U/M AZ521 + ++ ++ NAb Mutant Wild ÀÀÀMMM MKN28 À + + 0.91 Wild Wild ÀÀÀMMM MKN45 + + ++ 1.67 Wild Wild ÀÀÀMMM NUGC3 + ++ + 1.17 Wild Wild ÀÀ+ M M U/M NUGC4 À + À 2.54 NA NA ÀÀ+M M U AGS ++ À ++ 39.1 Wild Mutant ÀÀÀMMM NCI-N87 ++ ÀÀ1.73 Wild Wild ÀÀÀMMM SNU16 ND ND + ND Wild Wild ÀÀÀU/M U/M U/M

Abbreviations: APC, adenomatous polyposis coli; GC, gastric cancer; M, methylated; NA, not available; ND, not done; RT–PCR, reverse transcriptase–polymerase chain reaction; SFRP, secreted frizzled-related protein; TCF, T-cell factor; U, unmethylated; U/M, both of methylated and unmethylated DNA was detected by MSP. aWeak basal expression was detectable by RT–PCR but 5-aza-dC treatment further increased the expression. bNot available because of lowtransfection efficiency.

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4701 a MKN7 MKN74

β-catenin DAPI merge β-catenin DAPI merge

SH101 SNU638

β-catenin DAPI merge β-catenin DAPI merge

AZ521 NUGC3

β-catenin DAPI merge β-catenin DAPI merge

b SNU16 NUGC4 SNU638 KatoIII JSRT AZ521 NUGC3 AGS NCI-N87 MKN28 MKN45 MKN1 MKN7 MKN74 SH101 SNU1

β-catenin

active-β-catenin

GAPDH

c

97.4 39.1

15

10

5 Relative TCF Activity 0

AGS MKN1 MKN7 SNU1 JRST MKN74 SH101 KatoIII MKN28 MKN45 SNU638 NUGC3 NUGC4 NCI-N87 Figure 1 Evaluation of the Wnt pathway activity in GC cell lines. (a) Fluorescence immunohistochemical analysis of the intracellular distribution of b-catenin in GC cell lines. Cells were stained with anti-b-catenin antibody (green). Nuclei were visualized using 40,6- diamidino-2-phenylindole (DAPI; blue). (b) Western analysis of b-catenin in a set of GC cell lines using anti-b-catenin and anti-active- b-catenin antibodies. The same blot was reprobed for GAPDH as a loading control. (c) Luciferase assay carried out using a TCF/LEF- responsive reporter (pGL3-OT) and a negative control (pGL3-OF). Results are expressed as the pGL3-OT/pGL3-OF activity ratio after correction for transfection efficiency using renilla luciferase activity. Shown are means of four replications; the error bars represent standard deviations. Note that the TCF/LEF activities for the two cell lines with very elevated activities are not shown to scale.

luciferase reporter vector (pGL3-OT), we found that in observe some correspondence between luciferase activi- seven of the 14 (50%) cell lines tested, the luciferase ties and active b-catenin protein levels. activity of pGL3-OT was more than twofold higher than Our results confirmed activation of the Wnt signaling that of pGL3-OF, a negative control vector (Figure 1c; pathway in half of the GC cell lines tested, although the Table 1). Although the levels of TCF/LEF activity apparent degree of activation varied. By contrast, to our varied significantly among cell lines, we were able to knowledge, mutations in APC or CTNNB1 have been

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4702 reported for only four cell lines (Table 1) (Caca et al., eight cell lines are shown in Figure 2; Table 1). By 1999; Yokozaki, 2000; Sasaki et al., 2001; Ku and Park, contrast, expression of all three genes was detected in 2005). Of those, two (MKN74 and AZ521) were found normal stomach mucosa (Figure 2). Treatment with to harbor the same mutation at codon 1450 of APC,a a DNA methyltransferase inhibitor, 5-aza-20-deoxy- CGA-TGA transition, which leads to truncation of cytidine (5-aza-dC), rapidly restored SFRP expression, the protein (Sasaki et al., 2001). A GGA-GAA strongly indicating epigenetic silencing of SFRP1, transition at codon 34 (G34E) in CTNNB1 was found SFRP2 and SFRP5 in GC cells (representative data in AGS cells (Caca et al., 1999), and an ACC-GCC for eight cell lines are shown in Figure 2). change at codon 41 (T41A) in CTNNB1 was detected in We next analysed DNA methylation of SFRP genes. SNU638 cells (Ku and Park, 2005). In addition, We previously identified dense CpG islands in the 50 CTNNB1 amplification was recently demonstrated regions of SFRP1, SFRP2 and SFRP5 (Figure 3a), and in KatoIII cells (Suriano et al., 2005). The two cell lines so designed two sets of methylation-specific PCR (MSP) with APC mutations and KatoIII cells all showed primers and bisulfite sequencing primers for SFRP1 typical features of constitutive activation of Wnt and SFRP2, respectively (Figure 3a). For SFRP5,we signaling, that is, dephosphorylation and nuclear designed one MSP primer set and two bisulfite sequen- accumulation of b-catenin and upregulation of TCF/ cing primer sets. Our MSP analysis revealed methylation LEF-regulated transcription (Figure 1). In SNU638 of SFRP1 and SFRP2 in all the GC cell lines tested cells, however, only a low level of active b-catenin was (representative data for eight cell lines are shown in observed (Figure 1b). It is plausible that the anti-active- Figure 3b; Table 1). For both genes, the two sets of MSP b-catenin antibody failed to react with the mutant b- primers yielded nearly the same results, confirming that catenin, despite confocal microscopic and reporter the gene promoters were extensively methylated in GC assays confirming constitutive activation of Wnt signal- cells (Figure 3b). SFRP5 was methylated in 13 of 16 ing in SNU638 cells (Figure 1c; Table 1). Of note is fact (81%) GC cell lines, and there were good correlations that activation of the Wnt signaling pathway also was between methylation and the expression status (repre- found in cell lines without APC or CTNNB1 gene sentative data for eight cell lines are shown in Figure 3b; abnormalities (Table 1). Thus, our results suggest the Table 1). One exception was SH101 cells, in which involvement of mechanisms other than alterations in SFRP5 was silenced but no methylation was detected APC or CTNNB1 in the activation of Wnt signaling (Figure 3b; Table 1). in GC cells. In selected samples, we verified the MSP results using bisulfite sequencing (Figure 3c). The sequencing analysis revealed that the CpG islands of the three SFRP genes Expression and methylation analysis of SFRP genes in were extensively methylated in three GC cell lines GC cell lines that did not express the genes (MKN74, SNU638 We previously demonstrated frequent inactivation of and KatoIII), whereas a cell line expressing SFRP5 SFRP genes in CRC and suggested its involvement in (MKN1) showed no methylation at any of the cytosines the constitutive activation of Wnt signaling. We there- analysed (Figure 3c). fore hypothesized that a loss of SFRPs also may One potential mechanism for the tumor-specific lack contribute to the activation of Wnt signaling in GC. of SFRP mRNA expression is the inability of the cells to To test that idea, we first used reverse transcriptase– support SFRP transcription owing to a loss of critical polymerase chain reaction (RT–PCR) to analyse SFRP transactivation factors or the expression of repressor expression in a set of GC cell lines. Surprisingly, none of factors. To assess this possibility, we constructed a set of the 16 GC cell lines tested showed any expression of luciferase reporter vectors that contained upstream SFRP1 or SFRP2 mRNA, and 12 of the 16 lines also regions of SFRP1 and SFRP2 (Figure 4a and c). We did not express SFRP5 mRNA (representative data for then carried out luciferase assays in HEK293 cells and

O 2 MKN1 MKN7 MKN74 SH101 SNU1 SNU638 KatoIII JSRT dH O 2 5-aza-dC -+-+-+-+-+-+ -+- normaldH stomach SFRP1

SFRP2

SFRP5

GAPDH Figure 2 RT–PCR analysis of SFRP1, SFRP2 and SFRP5 expression in the indicated GC cell lines, with or without treatment with 5-aza-dC, and in normal stomach mucosa. RT–PCR for GAPDH was carried out for all samples to ensure the cDNA quality; dH2O indicates no RNA added.

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4703 100 bp Exon 1 a SFRP1 b CpG MSP-A KatoIII MSP-B MKN1 MKN7 MKN74 SH101 SNU1 SNU638 JRST BS-A M BS-B UM U M UM UM U M U M U M U SFRP1 MSP-A 100 bp SFRP2 Exon 1 SFRP1 MSP-B CpG MSP-A MSP-B SFRP2 MSP-A BS-A BS-B SFRP2 MSP-B 100 bp Exon 1 SFRP5 SFRP5 MSP CpG MSP BS-A BS-B

c SFRP1 BS-A SFRP1 BS-B MKN74 SNU638 KatoIII

-125-100 -75 -50 -25 0 5075 100 125 150 175 200 (bp)

d SFRP2 BS-A SFRP2 BS-B MKN74 SNU638 KatoIII -375 -350 -325 -300 -275 -250 5075 100 125 150 175 200225 (bp)

e SFRP5 BS-A SFRP5 BS-B MKN1 MKN74 SNU638 KatoIII -150-125-100 -75 -50 -25 025 100125 150 175 200 225 250 275 300 (bp) Figure 3 Analysis of SFRP1, SFRP2 and SFRP5 methylation in GC cell lines. (a) Schematic representation of the 50 regions of SFRP1, SFRP2 and SFRP5. Closed boxes indicate the first exons, arrows the translation start sites. Regions analysed by MSP and bisulfite sequencing are indicated by bars below the CpG sites. Two MSP primer pairs (MSP-A and MSP-B) were designed for both SFRP1 and SFRP2, and two regions (BS-A and BS-B) of each gene were analysed by bisulfite sequencing. (b) Representative results of MSP analysis of the indicated GC cell lines. Bands in the ‘M’ lanes are PCR products obtained with methylation-specific primers; those in the ‘U’ lanes are products obtained with unmethylated-specific primers. (c) Bisulfite sequencing of SFRP1 in the indicated GC cell lines. Results from two regions (BS-A and BS-B) are shown. Open and filled circles represent unmethylated and methylated CpG sites, respectively; gray circles represent partially methylated CpG sites. The location of each CpG site relative to the transcription start site is shown below. (d and e) Bisulfite sequencing of SFRP2 (d) and SFRP5 (e).

two cancer cell lines (HCT116 and SNU638) in which with DNA methylation within the promoter-associated both of the genes were silenced and methylated. CpG islands. Following transient transfection of the three cell lines, we observed high levels of luciferase activity with the pGL3-SFRP1-P3 and P4 vectors, which contain high- Suppression of TCF/LEF activity and cell growth by density CpG regions that encompass the presumed ectopic SFRPs expression in GC cells transcription start site (Figure 4b). We also obtained To examine the relationship between the silencing of similar results using luciferase vectors containing SFRP genes and Wnt signaling in GCs, we co- SFRP2 promoter sequences, and strong promoter transfected three GC cell lines showing constitutively activity was found in a region surrounding the active Wnt signaling (MKN74, SNU638 and KatoIII) transcription start site of SFRP2 (Figure 4d). Thus, with a TCF/LEF-responsive reporter gene and each the SNU638 GC cell line retains the ability to support SFRP expression vector. MKN74 cells are known to promoter-driven transcription, and we conclude that the harbor an APC mutation (Sasaki et al., 2001), whereas absence of SFRP gene expression is strongly associated CTNNB1 is mutated in SNU638 cells and amplified in

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4704 200 bp a SFRP 1 200 bp Sp1 c SFRP 2 Sp1 Exon 1 Exon 1

CpG CpG SFRP1-P1 -950 -328 SFRP2-P1 -859 -378 SFRP1-P2 -950 +13 SFRP2-P2 -859 +13 SFRP1-P3 -950 +228 SFRP2-P3 -859 +276 SFRP1-P4 -349 +228 SFRP2-P4 -398 +276 SFRP1-P5-8 +228 SFRP2-P5-8 +276

relative lucactivity (fold) relative lucactivity (fold) bd 0102030405060 010203040 pGL3-Basic pGL3-Basic pGL3-SFRP1-P1 pGL3-SFRP2-P1 pGL3-SFRP1-P2 pGL3-SFRP2-P2 pGL3-SFRP1-P3 pGL3-SFRP2-P3 HEK293 HEK293 pGL3-SFRP1-P4 pGL3-SFRP2-P4 pGL3-SFRP1-P5 pGL3-SFRP2-P5

0 50 100 150 200 250 0 50 100 150 200 pGL3-Basic pGL3-Basic pGL3-SFRP1-P1 pGL3-SFRP2-P1 pGL3-SFRP1-P2 pGL3-SFRP2-P2 pGL3-SFRP1-P3 pGL3-SFRP2-P3 HCT116 HCT116 pGL3-SFRP1-P4 pGL3-SFRP2-P4 pGL3-SFRP1-P5 pGL3-SFRP2-P5 0 10 203 04 0 010203040 pGL3-Basic pGL3-Basic pGL3-SFRP1-P1 pGL3-SFRP2-P1 pGL3-SFRP1-P2 pGL3-SFRP2-P2 pGL3-SFRP1-P3 pGL3-SFRP2-P3 SNU638 SNU638 pGL3-SFRP1-P4 pGL3-SFRP2-P4 pGL3-SFRP1-P5 pGL3-SFRP2-P5

Figure 4 Cells not expressing SFRP1 or SFRP2 support transcription from an exogenous promoter-reporter construct. (a) Diagram of the 50 CpG island of SFRP1. Various portions of the SFRP1 50-flanking regions shown as bars below were PCR amplified and cloned into the pGL3-basic vector. (b) A set of reporter vectors were transiently transfected into HEK293, HCT116 and SNU638 cells. Luciferase activity is shown for each construct expressed relative to the activity of the pGL3-basic vector after correction for transfection efficiency using the renilla luciferase activity. Shown are the means of four replications; the error bars represent standard deviations. Cell lines are indicated on the left. (c) Different portions of the SFRP2 promoter were cloned. (d) Results of luciferase assays carried out using SFRP2 promoter-reporter constructs in HEK293, HCT116 and SNU638 cells.

KatoIII cells. We found that overexpression SFRP1, GC cells, we transfected each SFRP vector into three SFRP2 or SFRP5 suppressed TCF/LEF activity in all of GC cell lines (SNU638, MKN45 and NUGC3) and then the GC cells tested (Figure 5a). carried out colony formation assays. After selection with The CRD situated in the N-terminal region of SFRPs G418 for 2 weeks, we found that introduction of SFRPs reportedly competes with Fz for Wnt binding, thereby markedly suppressed colony formation in all the cell modulating Wnt signaling (Jones and Jomary, 2002). lines tested (Figure 6a and b). Moreover, flowcyto- We confirmed this by carrying out reporter assays using metric analysis confirmed induction of apoptosis in SFRP2 deletion mutants lacking either the CRD or SNU638 and JRST cells transiently transfected with the C-terminal netrin-like domain (Figure 5b). SFRP1, SFRP2 or SFRP5 (Figure 6c). Apoptosis was SFRP2 lacking its CRD failed to inhibit TCF/LEF also induced by SFRPs in the DLD1 colon cancer cell activity, whereas the mutant lacking the netrin-like line, which carries an APC mutation (Figure 6c). domain yielded results similar to full-length SFRP2 (Figure 5c). Apparently, the CRD is necessary and sufficient to repress Wnt signaling in GC cells. Changes in gene expression profiles induced by We previously showed that overexpression of SFRPs overexpression of SFRP induces growth inhibition and apoptosis in CRC cells The present results, which are consistent with our earlier (Suzuki et al., 2004). To examine the effects of SFRPs in findings (Suzuki et al., 2004), demonstrate that ectopic

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4705 a MKN74 SNU638 KatoIII 120 120 120 pGL3-OT 100 100 100 80 80 80 pGL3-OF 60 60 60 40 40 40 20 20 20 0 0 0 relative luciferase activity relative luciferase activity relative luciferase activity

Vector Vector Vector SFRP1 SFRP2 SFRP5 SFRP1 SFRP2 SFRP5 SFRP1 SFRP2 SFRP5

b 30 153 SFRP2 1 CRD 296

Del-1 84 296 Del-2 139 296 Del-3 1 167

c MKN74 SNU638 KatoIII 140 120 140 120 100 120 100 80 100 80 80 60 60 60 40 40 40 20 20 20 0 0 0 relative luciferase activity relative luciferase activity relative luciferase activity

Del-1 Del-2 Del-3 Del-1 Del-2 Del-3 Del-1 Del-2 Del-3 Vector Vector Vector SFRP2 SFRP2 SFRP2

Figure 5 The Wnt signaling pathway in GC cells can be suppressed by overexpression of SFRPs. (a) Relative luciferase activity obtained using a TCF/LEF-responsive reporter (pGL3-OT) and a negative control (pGL3-OF) in GC cells transfected with the indicated SFRP expression construct. Results are shown relative to a value of 100 assigned to cells transfected with an empty vector (indicated as ‘Vector’) after correction for transfection efficiency using renilla luciferase activity. Shown are means of four replications; error bars represent standard deviations. (b) Schematic representation of SFRP2 deletion mutants; the CRD is shown as a gray box. Amino-acid numbers are also shown. (c) Effects of SFRP2 mutants on TCF/LEF transcription. Results are shown relative to a value of 100 assigned to cells transfected with an empty vector; full-length SFRP2 served as a positive control. expression of SFRPs inhibits Wnt signaling and induces cellular proliferation, growth, apoptosis and Wnt apoptosis and growth suppression in GC and CRC cells. signaling are listed in Table 2. However, another earlier study showed that direct Notably, four Wnt signaling target genes were down- inhibition of TCF/LEF activity by a dominant-negative regulated by SFRP2. LEF1 (Hovanes et al., 2001), mutant (TCF4B) did not affect colon cancer cell growth MMP7 (Brabletz et al., 1999), cyclin D1 (Tetsu and (Naishiro et al., 2001). Therefore, in order to further McCormick, 1999) and CD44 (Wielenga et al., 1999) clarify the effect of SFRP, we used oligonucleotide are all reported to be direct or indirect targets of microarrays to assess global changes in the gene b-catenin-TCF/LEF, confirming that Wnt signaling expression profile induced by SFRP2. We utilized is blocked by SFRP2 in SNU638 cells (Table 2). SNU638 cells because of their high transfection effi- We also found that significant numbers of genes related ciency with electroporation (>70%) and their high to cellular proliferation, growth or apoptosis are TCF/LEF activity, which was successfully suppressed repressed or induced by SFRP2. For example, by SFRP2 overexpression (Figure 5). GADD45B and GADD45A are two stress response The global changes in gene expression induced by genes, the expression of which induces apoptosis overexpression of SFRP2 in SNU638 cells were deter- (Takekawa and Saito, 1998), and PPP1R15A, also mined using the U133 Plus 2.0 Array known as GADD34, has been implicated in growth (Affymetrix, Santa Clara, CA, USA) with approxi- suppression and apoptosis. In addition, BCL2-asso- mately 55 000 gene/expressed sequence tag (EST) ciated athanogene 3 (BAG3), which is also implicated transcriptions. As compared with an empty vector- in apoptosis, is an heat shock protein 70 (Hsp70) co- transfected counterpart, 411 transcripts were down- chaperone (Takayama and Reed, 2001), and we noted regulated and 928 transcripts were upregulated by that a number of Hsp-70 genes are also upregulated by ectopic SFRP2 expression. Of these, genes related to SFRP2 (Table 2).

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4706 a Empty SFRP1 SFRP2 SFRP5 vector

SNU638

MKN45

NUGC3

b 140 SNU638 140 MKN45 140 NUGC3 120 120 120 100 100 100 80 80 80 60 60 60 40 40 40 % colony numbers % colony numbers 20 20 % colony numbers 20 0 0 0 r 1 2 5 r 1 2 5 r 1 2 5 o o o t P P P t P P P t P P P c c c R R R R R R R R R e e e F F F F F F F F F V V V S S S S S S S S S

c Empty SFRP1 SFRP2 SFRP5 vector 8.4% 14.5% 14.9% 16.6%

SNU638

10.7% 20.9% 25.1% 24.4%

JRST

5.6% 18.2% 21.9% 20.3%

DLD1

Figure 6 Growth suppression and apoptosis induced by overexpression of SFRPs. (a) Representative results from a colony formation assay carried out using the indicated GC cell lines. Cells were transfected with a vector encoding the indicated SFRP or an empty vector and selected for 2 weeks with G418. (b) Relative colony formation efficiencies of GC cells transfected with SFRP or control plasmid (Vector). Shown are means of three replications; error bars represent standard deviations. (c) Representative flowcytometric analysis. Cells were harvested and analysed 72 h after transfection. Apoptotic cells are indicated as the sub-G1 fraction, and percentages are shown on the top. Results from the DLD1 CRC cell line are also shown.

Frequent methylation and downregulation of SFRPs their methylation status in 46 primary GC specimens. in primary GC We also obtained adjacent non-tumorous gastric muco- To determine the extent to which SFRPs are aberrantly sa specimens from 32 of the 46 GC patients. In pri- methylated in primary tumors, we used MSP to examine mary GCs, we found significant frequencies of SFRP

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4707 Table 2 Genes downregulated or upregulated by overexpression of SFRP2 in SNU638 cells Probe ID Genbank Symbol Description Fold Change change P-value

Downregulated Wnt target genes 243362_s_at AA992805 LEF1 Lymphoid enhancer-binding factor 1 Transcription 0.379 0.99923 204259_at NM_002423 MMP7 Matrix metalloproteinase 7 (matrilysin, Collagen catabo- 0.660 0.99981 uterine) lism 208711_s_at BC000076 CCND1 Cyclin D1 (PRAD1: parathyroid Cell cycle, cell 0.660 0.99998 adenomatosis 1) growth 212063_at BE903880 CD44 CD44 antigen (homing function and 0.707 0.99866 Indian blood group system)

Downregulated genes 208381_s_at NM_003901 SGPL1 Sphingosine-1-phosphate lyase 1 Lipid metabolism, 0.109 0.99989 apoptosis 221666_s_at BC004470 ASC Apoptosis-associated speck-like protein Apoptosis, cell cy- 0.233 0.99851 containing a CARD cle 208173_at NM_002176 IFNB1 Interferon, beta 1, fibroblast Signal transduc- 0.379 0.99880 tion, cell prolifera- tion 214453_s_at NM_006417 IFI44 Interferon-induced protein 44 Invasive growth 0.406 0.99990 205387_s_at NM_000737 CGB Chorionic gonadotropin, beta polypep- Apoptosis 0.467 0.99997 tide 217373_x_at AJ276888 MDM2 Mdm2, transformed 3T3 cell double Oncogenesis, cell 0.500 0.99996 minute 2, p53 binding protein proliferation 202086_at NM_002462 MX1 Myxovirus resistance 1, interferon-in- Apoptosis 0.536 0.99998 ducible protein p78 211607_x_at U48722 EGFR Epidermal growth factor receptor Cell proliferation 0.536 0.99979 201508_at NM_001552 IGFBP4 Insulin-like growth factor binding Cell growth, cell 0.574 0.99969 protein 4 proliferation 206665_s_at NM_001191 BCL2L1 BCL2-like 1 Apoptosis 0.616 0.99938 201641_at NM_004335 BST2 Bone marrowstromal cell antigen 2 Cell proliferation, 0.616 0.99989 cell–cell signaling 209890_at AF065389 TM4SF9 Transmembrane 4 superfamily member Cell proliferation, 0.660 0.99979 9 cell adhesion 210984_x_at U95089 EGFR Epidermal growth factor receptor Cell proliferation 0.660 0.99994 213816_s_at AA005141 MET Met proto-oncogene (hepatocyte Signal transduc- 0.660 0.99979 growth factor receptor) tion, cell prolifera- tion 213979_s_at BF984434 CTBP1 C-terminal binding protein 1 Cell proliferation 0.660 0.99998 216041_x_at AK023348 GRN Granulin Signal transduc- 0.660 0.99998 tion, cell prolifera- tion 215101_s_at BG166705 CXCL5 Chemokine (C-X-C motif) ligand 5 Signal transduc- 0.660 0.99998 tion, cell prolifera- tion 200607_s_at BG289967 RAD21 RAD21 homolog (S. pombe) Apoptosis, cell cy- 0.707 0.99976 cle 203148_s_at NM_014788 TRIM14 Tripartite motif-containing 14 Cell growth 0.707 0.99998 31874_at Y07846 GAS2L1 Growth arrest-specific 2 like 1 Cell growth 0.707 1.00000 202286_s_at J04152 TACSTD2 Tumor-associated calcium signal Signal transduc- 0.707 0.99998 transducer 2 tion, cell prolifera- tion 209969_s_at BC002704 STAT1 Signal transducer and activator of Cell cycle, tran- 0.707 0.99851 transcription 1, 91 kDa scription 201601_x_at NM_003641 IFITM1 Interferon induced transmembrane Cell cycle, cell pro- 0.707 0.99998 protein 1 (9–27) liferation 214022_s_at AA749101 IFITM1 Interferon induced transmembrane Cell cycle, cell pro- 0.707 0.99931 protein 1 (9–27) liferation 232816_s_at AW571709 DDX11 DEAD/H (Asp-Glu-Ala-Asp/His) box Cell cycle, cell pro- 0.707 0.99923 polypeptide 11 liferation

Upregulated genes 223121_s_at AW003584 SFRP2 Secreted frizzled-related protein 2 Development 6208.375 0.00003 213418_at NM_002155 HSPA6 Heat shock 70 kDa protein 6 (HSP70B0) 73.5 0.00002 117_at X51757 HSPA6 Heat shock 70 kDa protein 6 (HSP70B0) 52.0 0.00000 203844_at NM_000551 VHL Von Hippel-Lindau syndrome Transcription, cell 7.5 0.00035 proliferation 215271_at BF432086 TNN Tenascin N Cell growth 6.5 0.00149 223850_at AF116720 DNAJA2 DnaJ (Hsp40) homolog, subfamily A, Cell cycle, cell pro- 6.5 0.00120 member 2 liferation

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4708 Table 2 (continued ) Probe ID Genbank Symbol Description Gene ontology Fold Change change P-value

208792_s_at M25915 CLU Clusterin Lipid metabolism, 5.3 0.00002 apoptosis 202581_at NM_005346 HSPA1B Heat shock 70 kDa protein 1B mRNA catabolism 4.9 0.00002 200800_s_at NM_005345 HSPA1A Heat shock 70 kDa protein 1A mRNA catabolism 4.3 0.00002 208791_at M25915 CLU Clusterin Lipid metabolism, 4.0 0.00013 apoptosis 217911_s_at NM_004281 BAG3 BCL2-associated athanogene 3 Apoptosis 3.7 0.00002 230031_at AW052044 HSPA5 Heat shock 70 kDa protein 5 3.2 0.00002 (glucose-regulated protein, 78 kDa) 223919_at AB045831 P53AIP1 p53-regulated apoptosis-inducing Apoptosis 2.8 0.00149 protein 1 210587_at BC005161 INHBE Inhibin, beta E Cell growth 2.6 0.00035 200799_at NM_005345 HSPA1A Heat shock 70 kDa protein 1A mRNA catabolism 2.3 0.00002 209101_at M92934 CTGF Connective tissue growth factor Cell growth, cell 2.3 0.00005 adhesion 228846_at AW071793 MAD MAX dimerization protein 1 Transcription, cell 2.1 0.00055 proliferation 207978_s_at NM_006981 NR4A3 Nuclear receptor subfamily 4, group A, Transcription, cell 2.0 0.00004 member 3 growth 225434_at AW245401 DEDD2 Death effector domain containing 2 Apoptosis 2.0 0.00002 200628_s_at M61715 WARS Tryptophanyl-tRNA synthetase Cell proliferation 1.9 0.00027 201466_s_at NM_002228 JUN v-jun sarcoma virus 17 oncogene Transcription, cell 1.9 0.00165 homolog growth 207574_s_at NM_015675 GADD45B Growth arrest and DNA-damage-in- Apoptosis 1.9 0.00002 ducible, beta 209304_x_at AF087853 GADD45B Growth arrest and DNA-damage-in- Apoptosis 1.9 0.00007 ducible, beta 36004_at AF074382 IKBKG Inhibitor of kappa light polypeptide Trancription, 1.9 0.00147 gene enhancer in B-cells, kinase gamma apoptosis 211936_at AF216292 HSPA5 Heat shock 70 kDa protein 5 (glucose- 1.7 0.00010 regulated protein, 78 kDa) 206569_at NM_006850 IL24 Interleukin 24 Apoptosis, immune 1.7 0.00087 response 207536_s_at NM_001561 TNFRSF9 Tumor necrosis factor receptor Apoptosis, immune 1.7 0.00031 superfamily, member 9 response 211839_s_at U22386 CSF1 Colony stimulating factor 1 Cell proliferation 1.7 0.00165 (macrophage) 219398_at NM_022094 CIDE-3 Cell death activator CIDE-3 Apoptosis 1.7 0.00108 241840_at AW204148 HSPA9B Heat shock 70 kDa protein 9B 1.6 0.00077 (mortalin-2) 200921_s_at NM_001731 BTG1 B-cell translocation gene 1, Cell proliferation 1.6 0.00019 anti-proliferative 201465_s_at BC002646 JUN v-jun sarcoma virus 17 oncogene Transcription, cell 1.6 0.00108 homolog (avian) growth 202014_at NM_014330 PPP1R15A Protein phosphatase 1, regulatory Apoptosis, cell cy- 1.6 0.00039 (inhibitor) subunit 15A cle 202643_s_at AI738896 TNFAIP3 Tumor necrosis factor, alpha-induced Apoptosis 1.6 0.00002 protein 3 37028_at U83981 PPP1R15A Protein phosphatase 1, regulatory (in- Apoptosis, cell cy- 1.6 0.00031 hibitor) subunit 15A cle 201464_x_at BG491844 JUN v-jun sarcoma virus 17 oncogene Transcription, cell 1.5 0.00019 homolog (avian) growth 214683_s_at AI251890 CLK1 CDC-like kinase 1 Cell cycle, cell pro- 1.5 0.00002 liferation 218145_at NM_021158 TRIB3 20 open reading frame 97 Apoptosis 1.5 0.00031 240319_at AI023795 CARD12 Caspase recruitment domain family, Apoptosis 1.4 0.00015 member 12 200690_at AA927701 HSPA9B Heat shock 70 kDa protein 9B 1.3 0.00024 (mortalin-2) 202644_s_at NM_006290 TNFAIP3 Tumor necrosis factor, alpha-induced Apoptosis 1.3 0.00011 protein 3 227345_at AI738556 TNFRS Tumor necrosis factor receptor Apoptosis 1.3 0.00087 F10D superfamily, member 10d 35150_at X60592 TNFRSF5 CD40 antigen Apoptosis, immune 1.3 0.00047 response 203725_at NM_001924 GADD45A Growth arrest and DNA-damage-in- Apoptosis, cell cy- 1.2 0.00021 ducible, alpha cle

Shown are genes related to Wnt signaling, apoptosis and cell growth and proliferation.

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4709 methylation (SFRP1, 42/46, 91%; SFRP2, 44/46, 96%; examined the expression status of SFRP1 using real SFRP5, 30/46, 65%). Notably, in many cases, we also time RT–PCR with complementary DNA (cDNA) detected methylation in the adjacent non-tumorous prepared from non-tumorous gastric mucosa from GC tissue (Figure 7a), but as MSP is not a quantitative patients (n ¼ 15) and primary tumor tissues harboring method of analysing methylation, we used bisulfite SFRP1 methylation (n ¼ 16). We found substantial sequencing to further examine SFRP1 methylation in reductions in the levels of expression of SFRP1 mRNA selected samples. We found that non-tumorous gastric in tumor tissues, but not in the adjacent non-tumorous mucosal tissues also harbor methylation of the gene, but mucosa (P ¼ 0.0033) (Figure 7c). Immunohisto che- the level of methylation of each cytosine was lower than mical analysis revealed an absence of SFRP1 protein that in tumor tissues (Figure 7b). expression in GC tissues (representative result in To confirm the association between SFRP1 methy- Figure 7d), whereas positive cytoplasmic staining for lation and the gene’s silencing in primary GCs, we SFRP1 was detected in gastric mucosal glands in

a GC15 GC28 GC37 GC42 GC46 GC49 NT NT NT NT NT NT U MU MUMUM UM UM UM UMUM UMUM UM SFRP1

SFRP2

SFRP5

bc P = 0.0033 0.1000 T

GC37 0.0100 N

0.0010 SFRP1 / GAPDH T

GC46 0.0001

N tumor normal stomach

d e

Figure 7 Methylation and expression of SFRPs in primary GCs. (a) Representative results of MSP analysis of SFRP1, SFRP2 and SFRP5 in primary GC tissues and adjacent non-tumorous gastric mucosa from the same patients. For SFRP1 and SFRP2, results obtained using MSP-B primers are shown. (b) Results of bisulfite sequencing of SFRP1 carried out using BS-A primers in GC tumors and corresponding adjacent non-tumorous gastric mucosa tissues. (c) Summarized results of real-time RT–PCR of SFRP1 in a set of primary GC tumors (n ¼ 16) and corresponding non-tumorous gastric mucosal tissues from the same GC patients (n ¼ 15). Results are normalized to internal GAPDH expression. (d and e) Immunohistochemical staining of SFRP1 in GC tissue (e) and adjacent non- tumorous gastric mucosa (e) from a GC patient.

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4710 adjacent non-tumorous mucosa (a representative result showed moderate activation. Taken together, these is shown in Figure 7e). observations suggest that nuclear b-catenin staining It thus appears that the relatively lowlevel of may overestimate Wnt signaling activity and that methylation seen in non-tumorous gastric mucosa from detection of active b-catenin protein is a more useful GC patients is not sufficient to silence the gene way to assess Wnt signaling. It is noteworthy, however, expression, whereas the higher levels of methylation that anti-active-b-catenin antibody sometimes may not seen in the tumors are sufficient to do so. be suitable for detecting b-catenin mutants, as it likely failed to detect mutant b-catenin (T41A) in SNU638 cells. We obtained similar results using HepG2 cells, which harbor a deletion in the GSK3b phosphorylation Discussion site of CTNNB1 (data not shown). We previously showed that SFRP1 and SFRP2 are Several lines of evidence suggest the Wnt signaling methylated in approximately 90% of CRCs, and about pathway is a major contributor to gastric carcinogenesis, 60% of those tumors also showed SFRP5 methylation though mutation of APC is not seen in the majority of (Suzuki et al., 2002). In the present study, we observed GCs (Horii et al., 1992; Sasaki et al., 2001). In addition, similar frequencies of SFRP methylation in GC cell lines our group has reported that although 16 of 70 (23%) and primary tumors. We also confirmed that restoration primary GCs showed nuclear and/or cytoplasmic of SFRP expression attenuated Wnt signaling, inhibited localization of b-catenin, subsequent analysis of 35 cell growth and induced apoptosis among GC cells. We tumors revealed no mutation in CTNNB1 (Sasaki et al., therefore propose that a loss of SFRP expression and 2001). Other researchers have documented the presence mutation of downstream components act in concert to of CTNNB1 mutations in GC with varying frequencies. boost activity in the Wnt signaling pathway. In cells For example, one group reported that 19.5% (59 of 303) without APC or CTNNB1 mutations, loss of SFRPs of primary GC tumors showed nuclear accumulation of may lead to moderate activation of Wnt signaling; b-catenin, but CTNNB1 mutation was detected in only however, we also noted that some GC cells did not show 5% (4 of 77) of the tumors (Woo et al., 2001). Another Wnt signaling activity, even though SFRPs were group analysed 311 primary GCs and observed nuclear methylated. Given that without Wnt ligands, loss of b-catenin in 90 tumors (29%). but they subsequently SFRP function would not be enough to activate the identified CTNNB1 mutations in only 19 of 73 tumors pathway, it is plausible that alterations in Wnt ligands (26%) with nuclear b-catenin, and no mutations were or their receptors may also be important determinants found in 19 tumors without nuclear accumulation of the signaling activity. (Clements et al., 2002). An alternative possibility is that loss of SFRPs A recent in situ hybridization analysis revealed contributes to activation of other oncogenic signaling CTNNB1 gene amplification in KatoIII (Suriano et al., pathways. One recent study showed that SFRP1 is 2005), and we found boosted TCF/LEF transcriptional frequently methylated in prostate cancer, and ectopic activity in the same cells. Thus, CTNNB1 amplification expression of SFRP1 reduced cellular proliferation, appears to be one mechanism by which b-catenin can be even though b-catenin-TCF/LEF is not activated overexpressed, but the frequency of this change in in prostate cancer cells (Lodygin et al., 2005). In primary GCs is quite low(1 of 49 patients) (Suriano another study, SFRP4 was found to be methylated in et al., 2005). AXIN1 mutations are seen more frequently b-catenin-deficient mesothelioma cells, and overexpres- than CTNNB1 mutation in hepatocellular carcinomas in sion of SFRP4 in the cells resulted in apoptosis and which there is nuclear accumulation of b-catenin (Satoh growth suppression (He et al., 2005a). Also, as described et al., 2003), but AXIN1 mutations have not yet been above, direct inhibition of TCF/LEF by the dominant- reported in GC, and no mutations were observed in negative mutant TCF4B did not suppress CRC cell gastro-esophageal junction adenocarcinomas (Koppert growth (Naishiro et al., 2001). These observations et al., 2004). prompted us to survey the changes in the global In the present study, we first sought to assess Wnt expression profile induced by SFRP overexpression in signaling in GC cells using several independent ap- cancer cells. proaches, and then tried to compare the results with Our microarray analysis confirmed that ectopic already known genetic alterations. Using confocal expression of SFRP2 repressed Wnt signaling in microscopic analysis, we observed nuclear b-catenin in SNU638 cells by downregulating b-catenin-TCF/LEF 13 of 15 GC cells, and membranous staining in 9 cells. target genes. LEF1, cyclin D1, MMP-7 and CD44 are all Western analysis showed detectable levels of total b- reportedly Wnt target genes in human CRC cells, and catenin in all GC cell lines tested; however, levels of the overexpression of cyclin D1 and MMP-7 has been activated protein (detected with anti-active-b-catenin reported in GC (Arber et al., 1999; Ii et al., 2006). In antibody) were much more varied and were not detected addition, Saikawa et al. (2001) reported that an in four cell lines (MKN1, SNU1, NUGC4 and NCI- antisense oligonucleotide against cyclin D1 inhibited N87). TCF/LEF-responsive reporter assays revealed growth of EGF-stimulated GC cells, while MMP-7, also markedly elevated activities in cells with genetic altera- known as matrilysin, is thought to promote cancer cell tions (SNU638, MKN74, KatoIII and AGS), and some invasion through its proteolytic activity against compo- of the remaining cell types without genetic changes also nents of the extracellular matrix (Ii et al., 2006). Thus,

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4711 our data suggest that loss of SFRPs likely contribute to 46 primary GCs and adjacent non-tumorous gastric mucosa the proliferation and invasiveness of GC cells. were obtained as described (Satoh et al., 2003). Genomic DNA The global expression analysis also revealed that was extracted using the standard phenol–chloroform proce- SFRPs induce changes in the expression of numerous dure. Total RNA was extracted by using Trizol reagent genes related to cellular growth or apoptosis. For (Invitrogen, Carlsbad, CA, USA) and then treated with a DNA-free kit (Ambion, Austin, TX, USA). Total RNA from example, GADD45B and GADD45A are induced by normal gastric mucosa from a healthy individual was environmental stresses, and they in turn induce p38/ purchased from BioChain (Hayward, CA, USA). JNK activation and apoptosis (Takekawa and Saito, 1998). In addition, PPP1R15A (GADD34) is regulated RT–PCR by genotoxic stress and nutrient deprivation during Single-stranded cDNA was prepared using SuperScript III myeloid differentiation (Hollander et al., 2003). Expres- reverse transcriptase (Invitrogen). The integrity of the cDNA sion of GADD34 correlates with apoptosis induced by was confirmed by amplifying glyceraldehydes-3-phosphate various signals, and its overexpression can elicit dehydrogenase (GAPDH). The PCR protocol entailed 5 min apoptosis (Hollander et al., 2003). By contrast, BAG3 at 951C; 35 cycles of 1 min at 951C, 1 min at 551C and 1 min at was first identified as a BAG1-related protein and is 721C; and a 7 min final extension at 721C. Primer sequences known to modulate anti-apoptotic activity. The BAG and PCR product sizes are shown in Supplementary Table 1. family proteins are Hsp70 co-chaperones and are characterized by the BAG-domain that interacts with Methylation analysis the ATPase domain of Hsp70 (Takayama and Reed, Bisulfite treatment of genomic DNA was carried out as described previously (Suzuki et al., 2002). Methylation of 2001). Interestingly, we also observed that SFRP2 SFRPs was analysed using MSP and bisulfite-sequencing induces a number of Hsp70 genes, including HSPA6, analysis as described previously (Suzuki et al., 2004). PCR HSPA1A and HSPA1B. Moreover, we carried out a was run in a volume of 25 ml containing 1 Â MSP buffer second microarray analysis, which revealed that over- (67 mM Tris-HCl (pH8.8), 16.6 mM (NH4)2SO4, 6.7 mM MgCl2 expression of SFRP2 also induced the PPP1R15A, and 10 mM 2-mercaptoethanol), 1.25 mM deoxyribonucleoside BAG3 and Hsp70 genes in the DLD1 CRC cell line (data triphosphate, 0.4 mM each primer and 0.5 U of JumpStart not shown). As the majority of the genes on our list are REDTaq DNA polymerase (Sigma, St Louis, MO, USA). The not implicated in Wnt signaling, it is not clear how PCR protocol for MSP entailed: 5 min at 951C; 35 cycles of SFRPs induce changes in the expression of these genes, 30 s at 951C, 30 s at 601C and 30 s at 721C; and a 7 min final 1 and much about howSFRPs induce apoptosis remains extension at 72 C. The PCR protocol for bisulfite-sequencing entailed: 5 min at 951C; 35 cycles of 1 min at 951C, 1 min at unclear. Nevertheless, our data and the studies of other 601C and 1 min at 721C; and a 7 min final extension at 721C. groups indicate that growth suppression and apoptosis Amplified bisulfite-sequencing PCR products were directly induced by SFRPs can be mediated by a mechanism sequenced by using an ABI3100 automated sequencer (Applied independent of the b-catenin-TCF/LEF pathway. Biosystems, Foster City, CA, USA). Primer sequences and Finally, our results suggest that blockade of Wnt PCR product sizes are given in Supplementary Table 1. ligand may exert a therapeutic effect in the treatment of GC. One recent study reported that anti-Wnt1 antibody Western blotting attenuated Wnt signaling and induced apoptosis in CRC Western analysis was carried out as described previously cells carrying APC or CTNNB1 mutations (He et al., (Murai et al., 2005). Anti-b-catenin antibody was purchased 2005b). Wnt and Fz also have been shown to be from BD Transduction Laboratories (San Diego, CA, USA). potential targets for immunotherapy in head and neck Anti-active-b-catenin antibody (clone 8E7), which is specific squamous cell carcinomas (Rhee et al., 2002). And the for b-catenin dephosphorylated on Ser37 and Thr41, was from Upstate (Lake Placid, NY, USA). Anti-GAPDH antibody was frequent SFRP methylation seen in non-cancerous from Santa Cruz Biotechnology (Santa Cruz, CA, USA). gastric mucosa from GC patients suggests that methyla- tion occurs at an early stage of gastric tumorigenesis. Immunofluorescence microscopy Taken together, these data suggest that Wnt blockade or Cells cultured on chamber slides were washed with phosphate- restoration of SFRP function may provide a useful new buffered saline (PBS) and fixed with 4% paraformaldehyde, approach to treating GCs. after which they were incubated with anti-b-catenin mono- clonal antibody (BD Transduction Laboratories) and stained with anti-mouse IgG Alexa Fluor 488 (Invitrogen). Cells were Materials and methods examined using an FV300-IX71 confocal laser scanning microscope (Olympus, Tokyo, Japan). Cell lines and tissues Thirteen GC cell lines (MKN1, MKN7, MKN28, MKN45, TCF/LEF reporter assay MKN74, SH101, SNU1, SNU638, JRST, KatoIII, AZ521, Reporter assays with pGL3-OT, a TCF/LEF-responsive NUGC3 and NUGC4) and two CRC cell lines (HCT116 and luciferase reporter plasmid, and pGL3-OF, a negative control DLD1) were obtained as described previously (Satoh et al., plasmid, were carried out as described previously (Suzuki 2003; Toyota et al., 2003). Three others (AGS, NCI-N87 and et al., 2004). Briefly, cells (5 Â 104 cells/well in 24-well plates) SNU16) were obtained from the American Type Culture were transfected with 100 ng of pGL3-OT or pGL3-OF and Collection (Manassas, VA, USA). All of the cell lines were 2 ng of pRL-CMV using Lipofectamine 2000 (Invitrogen). For cultured in the appropriate medium. To analyse restoration of co-transfection experiments, cells were transfected with 100 ng SFRP genes, cells were treated with 2.0 mM 5-aza-dC (Sigma) of one of the pCMV-HA-SFRP vectors plus the reporter for 72 h, replacing the drug and medium every 24 h. A total of plasmid. A series of SFRP expression vectors and SFRP2

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4712 deletion mutant constructs were prepared as described analysed using a FACSCalibur instrument (Becton Dickinson, previously (Suzuki et al., 2004). Firefly and renilla luciferase San Jose, CA, USA). activities were measured 48 h after transfection using a Dual- Luciferase Reporter Assay System (Promega, Madison, WI, Microarray analysis USA) and a Lumat LB 9507 luminometer (Berthold Techno- SNU638 cells (2 Â 106 cells) were transfected with 5 mg of either logies, Bad Wildbad, Germany). pCMV-HA-SFRP2 or empty vector using a Cell Line Promoter reporter assay Nucleofector kit V (Amaxa). The cells were then harvested Upstream regions of SFRP1 and SFRP2 were amplified by 24 h after transfection, and total RNA was prepared as PCR and then cloned into pCR2.1 TOPO (Invitrogen). Each described above. Synthesized cRNA was hybridized to the 0 forward PCR primer carried a 5 overhang that contained an Human Genome U133 Plus 2.0 Array (Affymetrix) with MluI recognition site, and each reverse primer contained a approximately 55 000 gene/EST transcriptions. BglII site. After verifying the sequences, fragments were cut using MluI and BglII and ligated into pGL3-Basic (Promega). Cells (5 Â 104 cells/well in 24-well plates) were transfected with Real-time RT–PCR 100 ng of one of the reporter plasmids and 2 ng of pRL-TK Real-time RT–PCR was carried out using TaqMan Gene (Promega) using Lipofectamine 2000 (Invitrogen). A pGL3- Expression Assays (Applied Biosystems) and a 7900HT Fast Basic vector without insert served as a negative control. Real-Time PCR System (Applied Biosystems) according to the Luciferase activities were measured 48 h after transfection manufacture’s instructions. SDS2.1 software (Applied Biosys- using a Dual-Luciferase Reporter Assay System (Promega). tems) was used to perform comparative DCt analysis. GAPDH served as endogenous controls. Colony formation assay Cells (2 Â 106 cells) were transfected with 5 mg of one of the Immunohistochemistry pcDNA3.1His-SFRP vectors or with empty vector using a Cell Formalin-fixed, paraffin-embedded sections of gastric carcino- Line Nucleofector kit V (Amaxa, Cologne, Germany) and a mas and adjacent non-tumorous gastric mucosa from five Nucleofector I electroporation device (Amaxa) according patients were heated in 10 mM sodium citrate in a microwave to manufacture’s instructions. A set of SFRP vectors were oven at maximum power for 15 min and then at a reduced prepared as described previously (Suzuki et al., 2004). Cells power for 15 min to achieve antigen recovery. After the were then plated on 60-mm culture dishes and selected for 14 sections were blocked in a 5% solution of bovine serum days with 0.6 mg/ml G418. Colonies were then stained with albumin in PBS also containing 0.5% Tween-20, they were Giemsa, and counted using National Institute of Health incubated with a goat anti-SFRP1 polyclonal antibody C-19 IMAGE software. (1:100 dilution; Santa Cruz) overnight at 41C. The sections were then developed using EnVision-Plus reagents (Dako, 0 Flow cytometry Glostrup, Denmark); 3,3 -diaminobenzidin served as the Flowcytometry wascarried out as described previously chromogen. Finally, the sections were counterstained with hematoxylin, and microscopic images were captured digitally. (Suzuki et al., 2004). Briefly, cells (2 Â 106 cells) were transfected with 5 mg of one of the pCMV-HA-SFRP vectors or with empty vector using the Cell Line Nucleofector kit V Acknowledgements (Amaxa) and a Nucleofector I electroporation device (Amaxa). The cells were then harvested 72 h after transfection and We thank Dr WF Goldman for editing the paper.

References

Arber N, Gammon MD, Hibshoosh H, Britton JA, Zhang Y, related protein 1 (SFRP 1) by promoter hypermethylation in Schonberg JB et al. (1999). Overexpression of cyclin D1 non-small-cell lung cancer. Oncogene 24: 6323–6327. occurs in both squamous carcinomas and adenocarcinomas Gregorieff A, Clevers H. (2005). Wnt signaling in the intestinal of the esophagus and in adenocarcinomas of the stomach. epithelium: from endoderm to cancer. Genes Dev 19: Hum Pathol 30: 1087–1092. 877–890. Brabletz T, Jung A, Dag S, Hlubek F, Kirchner T. (1999). He B, Lee AY, Dadfarmay S, You L, Xu Z, Reguart N et al. Beta-catenin regulates the expression of the matrix metallo- (2005a). Secreted frizzled-related protein 4 is silenced by proteinase-7 in human colorectal cancer. Am J Pathol 155: hypermethylation and induces apoptosis in beta-catenin- 1033–1038. deficient human mesothelioma cells. Cancer Res 65: 743–748. Caca K, Kolligs FT, Ji X, Hayes M, Qian J, Yahanda A He B, Reguart N, You L, Mazieres J, Xu Z, Lee AY et al. et al. (1999). Beta- and gamma-catenin mutations, but (2005b). Blockade of Wnt-1 signaling induces apoptosis in not E-cadherin inactivation, underlie T-cell factor/ human colorectal cancer cells containing downstream lymphoid enhancer factor transcriptional deregulation mutations. Oncogene 24: 3054–3058. in gastric and pancreatic cancer. Cell Growth Differ 10: Hollander MC, Poola-Kella S, Fornace Jr AJ. (2003). Gadd34 369–376. functional domains involved in growth suppression and Caldwell GM, Jones C, Gensberg K, Jan S, Hardy RG, Byrd P apoptosis. Oncogene 22: 3827–3832. et al. (2004). The Wnt antagonist sFRP1 in colorectal Horii A, Nakatsuru S, Miyoshi Y, Ichii S, Nagase H, Kato Y tumorigenesis. Cancer Res 64: 883–888. et al. (1992). The APC gene, responsible for familial Clements WM, Wang J, Sarnaik A, Kim OJ, MacDonald J, adenomatous polyposis, is mutated in human gastric cancer. Fenoglio-Preiser C et al. (2002). Beta-catenin mutation is a Cancer Res 52: 3231–3233. frequent cause of Wnt pathway activation in gastric cancer. Hovanes K, Li TW, Munguia JE, Truong T, Milovanovic T, Cancer Res 62: 3503–3506. Lawrence Marsh J et al. (2001). Beta-catenin-sensitive Fukui T, Kondo M, Ito G, Maeda O, Sato N, Yoshioka H isoforms of lymphoid enhancer factor-1 are selectively et al. (2005). Transcriptional silencing of secreted frizzled expressed in colon cancer. Nat Genet 28: 53–57.

Oncogene SFRPs methylation in gastric cancer M Nojima et al 4713 Ii M, Yamamoto H, Adachi Y, Maruyama Y, Shinomura Y. Sasaki Y, Morimoto I, Kusano M, Hosokawa M, Itoh F, (2006). Role of matrix metalloproteinase-7 (matrilysin) in Yanagihara K et al. (2001). Mutational analysis of the beta- human cancer invasion, apoptosis, growth, and angio- catenin gene in gastric carcinomas. Tumour Biol 22: 123–130. genesis. Exp Biol Med 231: 20–27. Satoh A, Toyota M, Itoh F, Sasaki Y, Suzuki H, Ogi K et al. Jones SE, Jomary C. (2002). Secreted frizzled-related proteins: (2003). Epigenetic inactivation of CHFR and sensitivity to searching for relationships and patterns. Bioessays 24: microtubule inhibitors in gastric cancer. Cancer Res 63: 811–820. 8606–8613. Koppert LB, van der Velden AW, van de Wetering M, Abbou M, Stoehr R, Wissmann C, Suzuki H, Knuechel R, Krieg RC, van den Ouweland AM, Tilanus HW et al. (2004). Frequent Klopocki E et al. (2004). Deletions of chromosome 8p and loss of the AXIN1 locus but absence of AXIN1 gene mutations loss of sFRP1 expression are progression markers of in adenocarcinomas of the gastro-oesophageal junction with papillary bladder cancer. Lab Invest 84: 465–478. nuclear beta-catenin expression. Br J Cancer 90: 892–899. Suriano G, Vrcelj N, Senz J, Ferreira P, Masoudi H, Cox K Ku JL, Park JG. (2005). Biology of SNU cell lines. Cancer Res et al. (2005). Beta-catenin (CTNNB1) gene amplification: a Treatment 37: 1–19. newmechanism of protein overexpression in cancer. Genes Lee AY, He B, You L, Dadfarmay S, Xu Z, Mazieres J et al. Cancer 42: 238–246. (2004). Expression of the secreted frizzled-related protein Suzuki H, Gabrielson E, Chen W, Anbazhagan R, van gene family is downregulated in human mesothelioma. Engeland M, Weijenberg MP et al. (2002). A genomic Oncogene 23: 6672–6676. screen for genes upregulated by demethylation and histone Liu TH, Raval A, Chen SS, Matkovic JJ, Byrd JC, Plass C. deacetylase inhibition in human colorectal cancer. Nat Genet (2006). CpG island methylation and expression of the 31: 141–149. secreted frizzled-related protein gene family in chronic Suzuki H, Watkins DN, Jair KW, Schuebel KE, Markowitz SD, lymphocytic leukemia. Cancer Res 66: 653–658. Chen WD et al. (2004). Epigenetic inactivation of SFRP Lodygin D, Epanchintsev A, Menssen A, Diebold J, genes allows constitutive WNT signaling in colorectal Hermeking H. (2005). Functional epigenomics identifies cancer. Nat Genet 36: 417–422. genes frequently silenced in prostate cancer. Cancer Res 65: Takayama S, Reed JC. (2001). Molecular chaperone targeting 4218–4227. and regulation by BAG family proteins. Nat Cell Biol 3: Murai M, Toyota M, Suzuki H, Satoh A, Sasaki Y, Akino K E237–E241. et al. (2005). Aberrant methylation and silencing of the Takekawa M, Saito H. (1998). A family of stress-inducible BNIP3 gene in colorectal and gastric cancer. Clin Cancer Res GADD45-like proteins mediate activation of the stress- 11: 1021–1027. responsive MTK1/MEKK4 MAPKKK. Cell 95: 521–530. Naishiro Y, Yamada T, Takaoka AS, Hayashi R, Hasegawa F, Taketo MM. (2004). Shutting down Wnt signal-activated Imai K et al. (2001). Restoration of epithelial cell polarity cancer. Nat Genet 36: 320–322. in a colorectal cancer cell line by suppression of beta- Tetsu O, McCormick F. (1999). Beta-catenin regulates expression catenin/T-cell factor 4-mediated gene transactivation. Cancer of cyclin D1 in colon carcinoma cells. Nature 398: 422–426. Res 61: 2751–2758. Toyota M, Sasaki Y, Satoh A, Ogi K, Kikuchi T, Suzuki H Offerhaus GJ, Giardiello FM, Krush AJ, Booker SV, et al. (2003). Epigenetic inactivation of CHFR in human Tersmette AC, Kelley NC et al. (1992). The risk of upper tumors. Proc Natl Acad Sci USA 100: 7818–7823. gastrointestinal cancer in familial adenomatous polyposis. Veeck J, Niederacher D, An H, Klopocki E, Wiesmann F, Betz B Gastroenterology 102: 1980–1982. et al. (2006). Aberrant methylation of the Wnt antagonist Rhee CS, Sen M, Lu D, Wu C, Leoni L, Rubin J et al. (2002). SFRP1 in breast cancer is associated with unfavourable Wnt and frizzled receptors as potential targets for immu- prognosis. Oncogene 25: 3479–3488. notherapy in head and neck squamous cell carcinomas. Wielenga VJ, Smits R, Korinek V, Smit L, Kielman M, Fodde R Oncogene 21: 6598–6605. et al. (1999). Expression of CD44 in Apc and Tcf mutant Risinger JI, Maxwell GL, Chandramouli GV, Aprelikova O, mice implies regulation by the WNT pathway. Am J Pathol Litzi T, Umar A et al. (2005). Gene expression profiling of 154: 515–523. microsatellite unstable and microsatellite stable endometrial Woo DK, Kim HS, Lee HS, Kang YH, Yang HK, Kim WH. cancers indicates distinct pathways of aberrant signaling. (2001). Altered expression and mutation of beta-catenin Cancer Res 65: 5031–5037. gene in gastric carcinomas and cell lines. Int J Cancer 95: Saikawa Y, Kubota T, Otani Y, Kitajima M, Modlin IM. 108–113. (2001). Cyclin D1 antisense oligonucleotide inhibits cell Yokozaki H. (2000). Molecular characteristics of eight gastric growth stimulated by epidermal growth factor and induces cancer cell lines established in Japan. Pathol Int 50: 767–777. apoptosis of gastric cancer cells. Jpn J Cancer Res 92: Yuasa Y. (2003). Control of gut differentiation and intestinal- 1102–1109. type gastric carcinogenesis. Nat Rev Cancer 3: 592–600.

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

Oncogene