OVEREXPRESSION OF CYCLOOXYGENASE-2 PROTEIN IS LESS FREQUENTIN GASTRIC CANCERS WITH MICROSATELLITE INSTABILITYHiroyuki YAMAMOTO, Fumio ITOH*, Hiroshi FUKUSHIMA, Yuji HINODA and Kohzoh IMAI

First Department of Internal Medicine, Sapporo Medical University, Chuo-ku, Sapporo, Japan

Overexpression of cyclooxygenase-2 (COX-2) has beenreported in gastric cancers. However, the relationship be-tween expression of COX-2 and clinico-pathological or geno-typic features has not been elucidated. To address the issue,expression of COX-2 protein was analyzed in 100 gastriccancers as well as 7 gastric cancer cell lines by using immuno-blot analysis. Overexpression of COX-2 in cancer tissuescompared with matched non-cancerous tissues was found in70% of cases and was significantly associated with lymphaticinvolvement, lymph node metastasis and advanced tumorstage. Interestingly, overexpression of COX-2 was less fre-quent in gastric cancers with microsatellite instability (MSI)than in those without MSI (8/20 vs. 62/80, p F 0.01). Expres-sion of COX-2 protein was detected in some gastric cancercell lines without MSI at various levels, but not in those withMSI. Our results suggest that overexpression of COX-2 mayplay an important role in tumor progression of gastric cancerand also support the notion that gastric cancers with andwithout MSI represent distinctive pathways of carcinogen-esis. We also observed a reduction of MSI phenotype afteraspirin or sulindac treatment in a hMLH1-defective gastriccancer cell line SNU-1, which lacks COX-2 expression. Int. J.Cancer (Pred. Oncol.) 84:400–403, 1999.r 1999 Wiley-Liss, Inc.

Various human malignant tumors, such as colorectal, gastric,breast, and lung cancers have been shown to form larger amountsof prostaglandins than their normal tissue counterparts. Epidemio-logical and clinical results have suggested that non-steroidalanti-inflammatory drugs (NSAIDs) may reduce the risk of colorec-tal, gastric, breast, and lung cancers (reviewed by Taketo, 1998).Cyclooxygenase-2 (COX-2), which has been identified as beingoverexpressed in colorectal and gastric cancers (Ristimakiet al.,1997; Soydanet al., 1997; Fujitaet al., 1998; Taketo, 1998, andreferences therein), is one of the major targets of NSAIDs (Taketo,1998). Accumulating evidence indicates that overexpression ofCOX-2 is associated with colorectal carcinogenesis (Taketo, 1998).Size- and invasion-dependent increase in COX-2 levels has beenreported in colorectal cancers (Fujitaet al., 1998). Colorectalcancers with microsatellite instability (MSI) have been shown toexpress low levels of COX-2 (Karneset al., 1998). Since colorectalcancers with and without MSI represent distinctive pathways ofcarcinogenesis, the finding is of particular interest (Ionovet al.,1993; Rampinoet al., 1997; Olschwanget al., 1997; Yamamotoetal., 1998).

The expression of COX-2 has been briefly examined in gastriccancers. Using immunoblotting in 6 paired gastric cancers, Soydanet al. (1997) reported that substantial amounts of COX-2 werefound in cancers but not in normal mucosa. Ristimakiet al. (1997)reported that levels of COX-2 mRNA were higher in cancers thanpaired normal mucosa in 11 patients with gastric cancer. They havealso observed cytoplasmic staining of COX-2 in cancer cells butnot in the surrounding stroma (Ristimaki,et al., 1997). In contrastto colorectal cancers, however, we lack knowledge of the relation-ship between expression of COX-2 and clinico-pathological orgenotypic features in gastric cancers. The molecular and clinico-pathological distinctions between cancers with and without MSIare also evident in gastric cancers, and therefore those with andwithout MSI appear to represent distinctive pathways of carcinogen-esis (Yamamotoet al., 1997; Oliveira,et al., 1998). In this regard,to determine whether specific clinico-pathological or genotypicfeatures are associated with overexpression of COX-2 is importantfor an improved understanding of the involvement of COX-2 in

gastric carcinogenesis as well as of the 2 distinctive pathways ofgastric carcinogenesis. Moreover, a marked reduction of MSIphenotype during exposure to NSAIDs, aspirin or sulindac, hasbeen noted in colon cancer cell lines deficient for the DNAmismatch repair genes,hMLH1 or hMSH2(Ruschoffet al., 1998).With respect to chemoprevention and treatment, it is of interest toexamine whether similar effects are present in gastric cancer cells.

In our current study, we analyzed 100 gastric adenocarcinomasfor the expression of COX-2 protein using immunoblot analysis.The results were matched with clinico-pathological characteristicsand MSI status. We also examined gastric cancer cell lines with andwithout MSI for expression of COX-2 and the effects of NSAIDtreatment on MSI phenotype in ahMLH1-deficient gastric cancercell line SNU-1.

MATERIAL AND METHODS

Tissue samples and analysis of MSI status

The origin and analysis of MSI status of gastric cancer tissuesamples have been described previously (Yamamotoet al., 1999).Briefly, cancer samples exhibiting microsatellite length alterationsin 40% or more of loci examined were defined as MSI-positive(MSI1) (Boland et al., 1998). The markers used were BAT25,BAT26, D2S123, D5S346 and D17S250. These markers werechosen based on the recommendation at the National CancerInstitute workshop (Bolandet al., 1998).

Cell cultures and NSAID treatmentThe cell lines examined were as follows: AZ521, MKN-28,

MKN45 and MKN74 as MSI2 gastric cancer cell lines and SNU-1,SNU-638 and NUGC-3 as MSI1 gastric cancer cell lines. The MSIstatus of each cell line has been described previously and/or wasdetermined by microsatellite analysis at BAT26 (Boyeret al.,1995). SNU-638 was kindly provided by Dr. H. Kim (College ofMedicine, Yonsei University, Seoul, Korea). Other cells werepurchased from the Japanese Cancer Research Resources Bank(Osaka, Japan) or the ATCC (Rockville, MD). The cells werecultured either in DMEM or RPMI 1640 medium with 10% fetalbovine serum. Effects of long-term aspirin or sulindac treatment onMSI frequency in SNU-1 cells, containing a homozygous non-sense mutation at codon 226 inhMLH1, were analyzed aspreviously described with some modifications (Ruschoffet al.,1998). Five loci described above were used for analysis of MSIfrequency.

Immunoblot analysis of COX-2 expressionFrozen tissues or the cells were homogenized in lysis buffer [50

mM Tris-HCl (pH7.4), 1% Triton X-100, 1 mM diethyldithiocarba-mic acid, 1 mM EDTA, 1% Tween 20, 10 µM leupeptin and 1 mMphenylmethylsulfonyl fluoride] on ice.After sonication and centrifu-gation (10,000g at 4°C), the protein content of the supernatant was

Grant sponsor: Ministry of Education, Science, and Culture, Japan; Grantsponsor: Ministry of Health and Welfare, Japan.

*Correspondence to: First Department of Internal Medicine, SapporoMedical University, S-1, W-16, Chuo-ku, Sapporo 060-8543, Japan. Fax:181-11-613-1141. E-mail: fitoh@sapmed.ac.jp

Received 21 December 1998; Revised 28 January 1999

Int. J. Cancer (Pred. Oncol.):84,400–403 (1999)

r 1999 Wiley-Liss, Inc.

Publication of the International Union Against CancerPublication de l’Union Internationale Contre le Cancer

measured by the method of Bradford. Solubilized proteins (100 µg)were separated by SDS-PAGE and transferred to polyvinylidenedifluoride transfer membrane. After being blocked with 5% block-ing agent, the membrane was incubated with an anti-human COX-2monoclonal antibody (MAb) (Cayman, Ann Arbor, MI) anddeveloped by the enhanced chemiluminescene system (Amersham,Arlington Heights, IL). The membrane was then stripped andreincubated with a polyclonal anti-actin antibody (Sigma, St.Louis, MO). The autoradiographs were scanned using a densitom-eter. The tumor-nontumor ratio of COX-2 expression was correctedfor that of actin expression.

Relationship between expression of COX-2and clinicopathological features

Expression of COX-2 was compared with the clinico-pathologi-cal features and MSI status. Statistical analysis was carried outusing the chi-square test or Fisher’s exact test, when the sampleswere insufficient for chi-square test, with ap value of less than 0.05being considered significant.

RESULTS

Immunoblot analysis of COX-2 in gastric cancer tissuesand matched non-cancerous tissues

Figure 1 shows representative results on 10 gastric cancerswithout (upper panel) or with (bottom panel) MSI. The COX-2protein was detected in the vast majority (95%) of cancer tissues,while being either low or undetectable in matched non-canceroustissues. Samples with a tumor-non-tumor ratio of.2.0 or withdetectable COX-2 protein only in cancer tissues were judged asbeing tumors with overexpression of COX-2, which was thusobserved in 70% of 100 cases. The results were compared withclinicopathological features and MSI status (Table I). Overexpres-sion of COX-2 was associated with lymphatic involvement (p 50.028), lymph node metastasis (p 50.002) and advanced tumorstage (p 5 0.004). Interestingly, overexpression of COX-2 was lessfrequent in gastric cancers with MSI than in those without MSI(8/20vs.62/80,p 5 0.003). There were no significant differences inlymphatic involvement, lymph node metastasis and tumor stagebetween gastric cancers with and without MSI (data not shown).

COX-2 expression in gastric cancer cell lines and effectsof NSAID treatment on MSI phenotype

The expression of COX-2 protein in gastric cancer cell lines withand without MSI is shown in Figure 2. The highest level of COX-2was detected in MKN-45. COX-2 protein was also detected inAZ521 and MKN-74, but not in MKN-28 or 3 MSI1 cell lines,SNU-1, SNU-638 or NUGC-3. MSI frequency in SNU-1 cells after16 weeks of aspirin or sulindac treatment was compared with thatof untreated ones. A marked reduction of MSI frequency was

TABLE I – ASSOCIATION OF COX-2 EXPRESSION WITHCLINICO-PATHOLOGICAL CHARACTERISTICS

Variables Categories Lown 5 30

Highn 5 70

pvalue

Gender male 16 45 N.S.1

female 14 25Location proximal 9 22 N.S.

middle 9 24distal 12 24

Histological type wellmoderate

59

818

N.S.

poor 16 44Vascular involvement negative

positive237

5020

N.S.

Lymphatic involvement negativepositive

1911

2644

0.028

Depth of invasion (T) 12

510

323

N.S.

3 11 304 4 14

Lymph node metastasis (N) 01

1310

927

0.002

2 7 34Distant metastasis (M) 0

1282

619

N.S.

TNM stage I 8 3 0.004II 8 13III 12 41IV 2 13

Microsatellite instability negativepositive

1812

628

0.003

1N.S.: not significant.

FIGURE 1 – Expression of COX-2 protein in gastric cancers with and without MSI. Results obtained for 10 representative cancer and matchednon-cancerous tissues are shown. Equal concentrations of protein (100 µg) were loaded in each lane, resolved by SDS-PAGE and electrotransferedto membrane. The membrane was then incubated with 5% blocking agent and subsequently incubated with antibodies against COX-2 and actin.Protein products were visualized with enhanced chemiluminescence. Actin served as an internal control. N and T, matched samples isolated fromnormal and tumor tissue, respectively.

401COX-2 PROTEIN AND MSI IN GASTRIC CANCER

observed at all 5 loci examined in SNU-1 cells after treatment witheither aspirin or sulindac.

DISCUSSION

In the current study, gastric cancers were evaluated to determinewhether specific clinico-pathological or genotypic features areassociated with overexpression of COX-2. The post-transcriptionaland post-translational regulation of COX-2 mRNA is complex andits mRNA expression is not necessarily indicative of enzymeexpression. Indeed, substantial expression of COX-2 mRNA with-out concomitant expression of COX-2 protein has been reported(Lee et al., 1992). Therefore, we used here immunoblot analysis.Expression of COX-2 was detected in the vast majority of cancertissues at various levels, while being either low or undetectable innon-cancerous tissues. Our results are in agreement with previousones (Soydanet al., 1997; Ristimakiet al., 1997). Overexpressionof COX-2 in cancer tissues compared with matched non-canceroustissues was seen in 70% of cases, being associated with lymphaticinvolvement, lymph node metastasis and advanced tumor stage.Our results suggest that COX-2 may play an important role in theprogression of gastric cancers. Interestingly, a marked increase inCOX-2 immunoreactivity has been often observed in lymph nodemetastases in adenocarcinomas of the lung (Hidaet al., 1998). Wehave previously reported that matrix metalloproteinase-7 (MMP-7,matrilysin) is frequently overexpressed in gastric cancer tissues andthat its expression correlates with lymph node metastasis andadvanced tumor stage (Senotaet al., 1998; Adachi,et al., 1998).Overexpression of COX-2 has been shown to augment the expres-sion of membrane-type MMP and the activation of MMP-2 incolon cancer cells (Tsujiiet al., 1997). Whether there are anyrelationships between overexpression of COX-2 and MMP expres-sion in gastric cancer tissues is not known.

With respect to the genotypic features, overexpression of COX-2was less frequent in gastric cancers with MSI than in those withoutMSI. These results were corroborated by the subsequent examina-tion of COX-2 expression in gastric cancer cell lines. Although the

total number of cell lines we examined was small, expression ofCOX-2 protein was detected in some lines without MSI at variouslevels, while not in those with MSI. Since the underlying mecha-nism by which COX-2 expression is elevated in gastrointestinalcancers is not well understood, these results, coupled with those ofothers, may help to address the issue (Karneset al., 1998). Inaddition to the induction by stimulation of chemical substances,cytokines and growth factors, the elevation of COX-2 expressionhas been supposed to be secondary to other initiating events such asalterations in the key regulatory gene(s), namelyAPC and others(Williams et al., 1997; Taketo, 1998, and references therein).Gastric cancers with and without MSI appear to represent distinc-tive pathways of carcinogenesis. Therefore, one possibility is thatalterations in the key regulatory gene(s) may be frequentlyinvolved in gastric cancers without MSI but not in those with MSI.APC is one of possible candidate genes, but links betweenAPCmutations and COX-2 expression are not clear in gastric cancers.This issue as well as the incidence ofAPC mutations in MSI1gastric cancers need to be addressed. Alternatively, as speculatedfor colorectal cancers (Karneset al., 1998), other mechanisms,such as prevalent DNA methylation and/or TGF-b type II receptormutations, may be involved in the low incidence of COX-2overexpression in MSI1 gastric cancers. The distinct expression ofCOX-2 might result from several mechanisms that deserve furtherinvestigation.

Our results suggest that overexpression of COX-2 may beinvolved in tumor progression of gastric cancers and support thenotion that gastric cancers with and without MSI representdistinctive pathways of carcinogenesis. With respect to chemopre-vention and treatment, a marked reduction of MSI phenotypeduring exposure to NSAIDs, aspirin or sulindac, has been reportedin colon cancer cell lines deficient for the DNA mismatch repairgenes,hMLH1 or hMSH2(Ruschoffet al., 1998). Interestingly, areduction of the MSI phenotype after NSAIDs treatment has alsobeen observed in DLD-1/HCT15 cells which lack COX-2 expres-sion. We found similar effects of NSAIDs on the MSI phenotype ina hMLH-1-deficient gastric cancer cell line SNU-1, which lacksCOX-2 expression. Thus, a direct link between COX-2 expressionand the MSI phenotype is unlikely (Ruschoffet al., 1998). Gastriccancer is included in the broad spectrum of hereditary non-polyposis colorectal cancer (HNPCC) syndrome and often exhibitsMSI (Yamamotoet al., 1997). Thus, although selective COX-2inhibition may be effective for chemoprevention in the majority ofgastric cancers, the use of aspirin/sulindac would be necessary foran effective prophylactic therapy for gastric tumors developing inthe setting of DNA mismatch repair defects, such as in HNPCCkindreds.

ACKNOWLEDGEMENTS

This work was supported by grants-in-aid from the Ministry ofEducation, Science, and Culture (to FI, YH and KI) and from theMinistry of Health and Welfare (to FI, YH and KI), Japan.

REFERENCES

ADACHI, Y., ITOH, F., YAMAMOTO, H., MATSUNO, K., ARIMURA, Y., KUSANO,M., ENDOH, T., HINODA, Y., OOHARA, M., HOSOKAWA, M. and IMAI , K.,Matrix metalloproteinase matrilysin (MMP-7) participates in the progres-sion of human gastric and esophageal cancers.Int. J. Oncol., 13,1031–1035(1998).

BOLAND, C.R., THIBODEAU, S.N., HAMILTON , S.R., SIDRANSKY, D., ESHLE-MAN, J.R., BURT, R.W., MELTZER, S.J., FODDE, R., RODRIGUEZ-BIGAS, M.A.,RANZANI , G.N. and SRIVASTAVA , S., A National Cancer Institute workshopon microsatellite instability for cancer detection and familial predisposition:development of international criteria for the determination of microsatelliteinstability in colorectal cancer.Cancer Res., 58,5248–5257 (1998).

BOYER, J.C., UMAR, A., RISINGER, J.I., LIPFORD, J.R., KANE, M., YIN, S.,BARRETT, J.C., KOLODNER, R.D. and KUNKEL, T.A., Microsatellite instabil-

ity, mismatch repair deficiency, and genetic defects in human cancer celllines.Cancer Res., 55,6063–6070 (1995).

FUJITA, T., MATSUI, M., TAKAKU , K., UETAKE, H., ICHIKAWA , W., TAKETO,M.M. and SUGIHARA, K., Size- and invasion-dependent increase in cyclooxy-genase 2 levels in human colorectal carcinomas.Cancer Res., 58, 4823–4826 (1998).

HIDA, T., YATABE, Y., ACHIWA, H., MURAMATSU, H., KOZAKI, K., NAKA -MURA, S., OGAWA, M., MITSUDOMI, T., SUGIURA, T. and TAKAHASHI , T.,Increased expression of cyclooxygenase 2 occurs frequently in human lungcancers, specifically in adenocarcinomas.Cancer Res., 58, 3761–3764(1998).

IONOV, Y., PEINADO, M.A., MALKHOSYAN, S., SHIBATA , D. and PERUCHO, M.,Ubiquitous somatic mutations in simple repeated sequences reveal a new

FIGURE 2 – Expression of COX-2 protein in gastric cancer cell lineswith and without MSI. Immunoblot analysis using total cell extract(100 µg) in each lane was carried out as described above.

402 YAMAMOTO ET AL.

mechanism for colonic carcinogenesis.Nature (Lond.), 363, 558–561(1993).

KARNES, W.E., JR., SHATTUCK-BRANDT, R., BURGART, L.J., DUBOIS, R.N.,TESTER, D.J., CUNNINGHAM, J.M., KIM, C.-Y., MCDONNELL, S.K., SCHAID,D.J. and THIBODEAU, S.N., Reduced COX-2 protein in colorectal cancerwith defective mismatch repair.Cancer Res., 58,5473–5477 (1998).

LEE, S.H., SOYOOLA, E., CHANMUGAM , P., HART, S., SUN, W., ZHONG, H.,LIOU, S., SIMMONS, D. and HWANG, D., Selective expression of mitogen-inducible cyclooxygenase in macrophages stimulated with lipopolysaccha-ride.J. biol. Chem., 267,25934–25938 (1992).

OLIVEIRA , C., SERUCA, R., SEIXAS, M. and SOBRINHO-SIMOES, M., Theclinicopathological features of gastric carcinomas with microsatelliteinstability may be mediated by mutations of different ‘‘target genes’’.Amer.J. Pathol., 153,1211–1219 (1998).

OLSCHWANG, S., HAMELIN , R, LAURENT-PUIG, P., THUILLE, B., RYCKE, Y.D.,LI, Y.J., MUZEAU, F., GIRODET, J., SALMON, R.-J. and THOMAS, G.,Alternative genetic pathways in colorectal carcinogenesis.Proc. nat. Acad.Sci. (Wash.), 94,12122–12127 (1997).

RAMPINO, N., YAMAMOTO, H., IONOV, Y., LI, Y., SAWAI , H., REED, J.C. andPERUCHO, M., Somatic frameshift mutations in the BAX gene in coloncancers of the microsatellite mutator phenotype.Science, 275, 967–969(1997).

RISTIMAKI , A., HONKANEN, N., JANKALA , H., SIPPONEN, P. and HARKONEN,M., Expression of cyclooxygenase-2 in human gastric carcinoma.CancerRes., 57,1276–1280 (1997).

RUSCHOFF, J., WALLINGER, S., DIETMAIER, W., BOCKER, T., BROCKHOFF, G.,HOFSTADTER, F. and FISHEL, R., Aspirin suppresses the mutator phenotype

associated with hereditary nonpolyposis colorectal cancer by geneticselection.Proc. nat. Acad. Sci. (Wash.), 95,11301–11306 (1998).

SENOTA, A., ITOH, F., YAMAMOTO, H., ADACHI, Y., HINODA, Y. and IMAI , K.,Relation of matrilysin messenger RNA expression with invasive activity inhuman gatric cancer.Clin. exp. Metastasis, 16,313–321 (1998).

SOYDAN, A.S., GAFFEN, J.D., WEECH, P.K., TREMBLAY, N.M., KARGMAN, S.,O’NEILL, G., BENNETT, A. and TAVARES, I.A., Cytosolic phospholipase A2,cyclo-oxygenases and arachidonate in human stomach tumors.Europ. J.Cancer, 33,1508–1512 (1997).

TAKETO, M.M., Cyclooxygenase-2 inhibitors in tumorigenesis (Part II).J.nat. Cancer Inst., 90,1609–1620 (1998).

TSUJII, M., KAWANO, S. and DUBOIS, R.N., Cyclooxygenase-2 expression inhuman colon cancer cells increases metastatic potential.Proc. nat. Acad.Sci. (Wash.), 94,3336–3340 (1997).

WILLIAMS , C.S., SMALLEY , W. and DUBOIS, R.N., Aspirin use and potentialmechanisms for colorectal cancer prevention.J. clin. Invest., 100, 1325–1329 (1997).

YAMAMOTO, H., SAWAI , H. and PERUCHO, M., Frameshift somatic mutationsin gastrointestinal cancer of the microsatellite mutator phenotype.CancerRes., 57,4420–4426 (1997).

YAMAMOTO, H., ITOH, F., FUKUSHIMA, H., ADACHI, Y., ITOH, H., HINODA, Y.and IMAI , K., Frequent Bax frameshift mutations in gastric cancer with highbut not low microsatellite instability.J. exp. clin. Cancer Res., 18,(1999) inpress.

YAMAMOTO, H., ITOH, F., KUSANO, M., YOSHIDA, Y., HINODA, Y. and IMAI ,K., Infrequent inactivation of DCC gene in replication error-positivecolorectal cancers.Biochem. biophys. Res. Comm., 244,204–209 (1998).

403COX-2 PROTEIN AND MSI IN GASTRIC CANCER