Cancer and Metastasis Reviews

, Volume 30, Issue 3, pp 387–395

Cyclooxygenase-2 and Gastric Cancer


  • Alexandra Thiel
    • Department of Pathology, HUSLAB and Haartman InstituteHelsinki University Central Hospital and University of Helsinki
    • Genome-Scale Biology, Research Program UnitUniversity of Helsinki
  • Johanna Mrena
    • Department of Pathology, HUSLAB and Haartman InstituteHelsinki University Central Hospital and University of Helsinki
    • Department of Gastrointestinal SurgeryCentral Hospital of Central Finland
    • Department of Pathology, HUSLAB and Haartman InstituteHelsinki University Central Hospital and University of Helsinki
    • Genome-Scale Biology, Research Program UnitUniversity of Helsinki
    • Genome-Scale Biology, Research Program Unit, Biomedicum HelsinkiUniversity of Helsinki

DOI: 10.1007/s10555-011-9312-1

Cite this article as:
Thiel, A., Mrena, J. & Ristimäki, A. Cancer Metastasis Rev (2011) 30: 387. doi:10.1007/s10555-011-9312-1


Gastric cancer remains a leading cause of cancer-related deaths worldwide, although its incidence has been steadily declining during recent decades. Expression of cyclooxygenase-2 (COX-2) is elevated in gastric carcinomas and in their precursor lesions. COX-2 expression associates with reduced survival in gastric cancer patients, and it has also been shown to be an independent factor of poor prognosis. Several molecular mechanisms are involved in the regulation of COX-2 expression in gastric cancer cell lines, including signal transduction pathways activated by Helicobacter pylori. In gastric tumor models in vivo the role of COX-2 seems to be predominantly to facilitate tumor promotion and growth.


Cyclooxygenase-2Gastric cancerCelecoxibCarcinogenesisProstaglandin E2

1 Introduction

Incidence of gastric cancer has decreased during the past decades especially in the Western world, but it is still an important cause of morbidity and mortality worldwide [1]. The etiology of gastric cancer has a significant environmental component, including H. pylori infection, consumption of salted and nitrated foods and cigarette smoking [2, 3]. In addition to environmental factors, genetic alterations including susceptible genetic variants and also epigenetic alterations play a crucial role in the process of gastric carcinogenesis [4, 5]. The present review is intended to focus on the role of cyclooxygenase-2 (COX-2), the key regulatory enzyme in prostanoid synthesis and the primary target of non-steroidal anti-inflammatory drugs (NSAIDs), in gastric carcinogenesis [6, 7].

2 COX-2 in human gastric carcinogenesis

Epidemiologic cohort and case–control studies have suggested that regular and long-term use of aspirin and other NSAIDs reduces mortality from digestive tract malignancies, including gastric cancer [813]. As a result, the cyclooxygenase enzyme is considered a potential therapeutic target in cancer prevention and treatment. In humans, COX-2 expression, but not that of COX-1, is elevated in gastric cancer tissues [14, 15]. Importantly, COX-2 is already present in noninvasive gastric dysplasias [1618]. Thus, it seems likely that COX-2 plays a role in early gastric carcinogenesis. Chronic atrophic gastritis caused by H. pylori activates synthesis of growth factors, cytokines, and gastrin leading to elevated COX-2 expression [19]. Furthermore, CagA-positive H. pylori infection associates with COX-2 expression in human gastric cancer [20], and in gastric cancer cells H. pylori exposure elevates COX-2 expression [21]. H. pylori may thus play an important role in induction of COX-2 synthesis during chronic gastritis that is a precancerous condition for gastric cancer.

COX-2 expression is associated with the clinical outcome of several cancers, like colorectal [22], esophageal [23], breast [24], ovarian [25], and pancreatic cancers [26]. In gastric cancer we and others have shown that COX-2 expression is associated with intestinal histological subtype, proximal location, large tumor size, and advanced stage [2740]. Association of COX-2 with survival has been controversial, since some studies show no such link [3234]. A non-independent association of COX-2 with poor survival has been reported in patients with advanced stage [30, 41], as well as an independent association in a selected group of low-stage patients [36]. However, we were the first to report that COX-2 expression is an independent prognostic factor in a consecutive patient series with gastric cancer [27]. In addition, we had a special interest in low-stage disease, and whether COX-2 could identify those patients with non-advanced, operable, but potentially aggressive cancer. We found that low-stage patients with high tumoral COX-2 expression were at higher risk for gastric cancer-related death than were the patients with low content of COX-2 [27]. Finally, in an extended multivariate model with eight prognostic markers and clinicopathological factors, COX-2 expression is an independent prognostic factor along with p53, stage, and intent of surgery [42]. In conclusion, COX-2 expression is a marker for poor prognosis in gastric cancer at all stages, but especially in low-stage patients.

The approach to routinely prevent cancer by application of selective COX-2 inhibitors is not feasible, as this treatment increases the risk for cardiovascular events [4346]. However, the risk of cardiovascular toxicity or gastric ulcers was not increased when patients with atrophic gastritis, intestinal metaplasia, or dysplasia received 200 mg celecoxib twice daily [47]. When mice bearing metastasis originating from scirrhous gastric carcinoma were treated with a combination of S-1 (a per oral drug containing tegafur that is converted to fluorouracil in the body) and a selective COX-2 inhibitor, their survival was significantly longer [48]. The updated safety analysis of the APC Trial suggested a link between the increased risk for cardiovascular toxicity after celecoxib use and the baseline history of atherosclerotic heart disease [49].

3 In vitro models

3.1 Mechanisms of COX-2 regulation

COX-2 is regulated via multiple pathways in gastric cancer cell lines. These cell lines may originate from one of the two main histological types of gastric cancer, i.e., intestinal (for example MKN-28) or diffuse type (for example TMK-1) [50], basal, and stimulated expression of COX-2 and its enzymatic activity have been shown to be facilitated by the phosphatidylinositol 3-kinase/Akt/glygogen synthase kinase-3β (PI3K/Akt/GSK-3β) pathway in MKN-28 and TMK-1 gastric cancer cell lines [51]. In addition, inhibition of GSK-3β promoted COX-2 expression by stabilizing the transcript and to a lesser extent by stimulating transcription. Mitogen-activated protein kinases (MEK 1/2, p38, JNK) or the mammalian target of rapamycin were not involved in regulation of COX-2 overexpression in these cell lines [51]. Another mechanism of COX-2 upregulation has been found in SC-M1 gastric adenocarcinoma cells, where the activated Notch1 signal pathway led to elevated COX-2 expression [52]. In these cells the activated form of the Notch1 receptor, the Notch1 receptor intracellular domain, was overexpressed, which enhanced COX-2 promoter activity through C-promoter-binding factor 1, a nuclear mediator of Notch signaling. Importantly, the Notch1 pathway-induced COX-2 expression enhanced properties connected to cancer progression in SC-N1 cells, such as colony formation, migration, and invasion. Furthermore, expression of COX-2 and Jagged 1, a Notch ligand, were correlated in gastric cancer tissues [52].

The COX-2 promoter region contains several cis elements, among them two nuclear factor-κB (NF-κB) consensus sites [53]. In AGS gastric cancer cells, NF-κB was found to regulate COX-2 expression [54]. COX-2 expression and its enzymatic activity were reduced upon NF-κB inhibition with antisense oligonucleotides of p50 (a NF-κB subunit) or with dominant negative IκBα. Inhibition of COX-2, either mediated by NF-κB inhibition or by treatment with non-specific (indomethacin) or specific COX-2 (NS-398) inhibitors, resulted in suppression of cell proliferation [54]. Another transcription factor, activator protein-1 (AP-1), was found to be responsible for upregulation of COX-2 transcription in gastrin-treated AGS-E gastric cancer cells (this cell line stably expresses the gastrin receptor CCK2) [55]. This transcriptional regulation was mediated via p38 upstream of the PI3K/Akt pathway. Gastrin also increased COX-2 mRNA stability through p38 signaling. Human antigen R (HuR) is a mRNA-binding factor that binds to AU-rich elements and thus can stabilize transcripts [56]. Interestingly, also HuR expression, cytoplasmic location as well as binding to COX-2 mRNA were enhanced after gastrin treatment in a p38-dependent manner [55]. This finding of HuR-mediated COX-2 stability correlates with our results in TMK-1 gastric cancer cells, where inhibition of HuR with siRNA reduced the expression of COX-2 protein [27].

It is well established that H. pylori contributes to COX-2 expression [15]. Expression of COX-2 and mPGES-1 transcripts as well as PGE2 levels increased after infection of primary mouse epithelial cells with Helicobacter felis, a close relative to Helicobacter pylori [57]. Also in rat gastric epithelial cells treatment with a H. pylori water extract (only containing bacterial proteins but not bacterial cells) led to an increase in COX-2 and PGE2 levels that peaked 24 h after treatment and declined at 48 h [58]. One of the underlying molecular mechanism of the COX-2 upregulation upon infection with H. pylori (P12 wild type) was unraveled in AGS and MKN-28 gastric cancer cells. Transcription factors USF1, USF2, and CREB were found to bind to the CRE/Ebox site of the COX-2 promoter and these transcription factors were induced by the MEK/ERK1/2 cascade [59]. In another study using AGS gastric cancer cells, H. pylori (patient isolate) promoted COX-2 transcription through TLR2/TRL9 that activated the MAPK pathways (ERK1/2, p38, JNK) and resulted in the activation of CRE and AP-1 on the COX-2 promoter [60]. COX-2 expression has also been shown to be induced by NF-κB, which is activated by the TLR2/TLR9 and c-Src or TLR2/TLR9 and NIK (mitogen-activated protein kinase kinase kinase 14) pathway in this model [61]. In MKN-45 gastric cancer cells the p38MAPK/ATF-2 pathway was necessary for increased COX-2 expression after H. pylori (NCTC11637 standard strain) infection [62]. Thus H. pylori clearly induces COX-2, but the mechanism seems to be dependent on the H. pylori strain properties and the recipient cells.

MicroRNAs (miRNAs) are small non-coding RNAs that upon binding to their target mRNA can repress translation or lead to mRNA degradation. In the recent years dysregulated miRNAs have also been found in gastric cancer, and they are linked to many processes, such as cell proliferation, apoptosis, and invasion [63]. Until this moment, in cancer cells, only miRNA-101 has been shown to target COX-2 directly and thus silence it [6466]. The direct interaction of miRNA-101 with COX-2 in cancer was first shown in LS-174T colon cancer cells, where co-transfection of miRNA-101 and a COX-2 3′UTR reporter led to a decreased luciferase activity [64]. When miRNA-101 was overexpressed in gastric cancer cells lines, the mRNA level of COX-2 was decreased in BGC-823, MKN-45, and AGS cells [65]. Furthermore, miRNA-101 overexpression resulted in inhibition of proliferation, migration, and invasion in these cells, and miRNA-101-overexpressing MKN-45 cells that were injected into nude mice showed a significantly reduced tumor growth compared to the control group.

15-Hydroxyprostaglandin dehydrogenase (15-PGDH) degrades and inactivates PGE2, the main prostaglandin produced by epithelial-derived tumor cells, and thus acts as a direct counterplayer to COX-2. 15-PGDH is frequently downregulated in gastric cancer [6770]. A reciprocal regulation of COX-2 and 15-PGDH was found in lung cancer cells as well [71]. In SGC7901 gastric cancer cells, COX-2 was involved in the downregulation of 15-PGDH, and in gastric cancer tissues there was a negative correlation between COX-2 and 15-PGDH [68]. However, we could not find a link between COX-2 and 15-PGDH in MKN-28 gastric cancer cells, as siRNA inhibition of either COX-2 or 15-PGDH did not alter the expression of the other protein. Furthermore, no correlation between COX-2 and 15-PGDH expression was established in our clinical material [69].

3.2 COX-2-pathway-mediated effects

The COX-2 pathway has been shown to be involved in many processes leading to tumor progression such as angiogenesis, survival, proliferation, invasion, and immunosuppression [72]. When the conditioned medium of SGC7901 gastric cancer cells treated with COX-2 siRNA or the inhibitor NS-398 was applied to human umbilical vein endothelial cells, the proliferation, migration, and tube formation was suppressed [73]. Addition of PGE2 could partially reverse these effects. Furthermore, the angiogenesis of COX-2 siRNA or NS-398-treated SGC7901 cell tumors in nude mice was significantly suppressed as determined by microvessel area [73]. A similar approach was used by Yao et al. who also reported a reduced tumor angiogenesis of COX-2-siRNA-treated SGC7901 cells [74]. A microarray analysis was used to determine COX-2-regulated angiogenesis-related molecules that was confirmed by RT-PCR and Western blotting. VEGF, Flt-1, Flk-1/KDR, angiopoietin-1, tie-2, MMP2, and osteopontin were found to be downregulated in response to COX-2 inhibition [74]. In MKN-28 cells, the EGFR-MAPK signaling pathway was involved in the upregulation of VEGF after PGE2 application to the cells [75]. In MKN-45 cells inhibition of COX-2 with NS-398 led to a reduced proliferation and induction of apoptosis, connected with downregulation of Bcl-2 and upregulation of Bax. Notably, the effects were increased synergistically, when the CCK-2 receptor was added together with NS-398, suggesting a contribution of gastrin signaling in proliferation and apoptosis in these cells [76]. Also, in clinical gastric cancer tissue samples, COX-2 was associated with markers for apoptosis and proliferation [42]. COX-2 signaling has been found to be involved in immunosuppression in gastric cancer, where regulatory T (Treg) cells (CD4+CD25+Foxp3+) suppressed effector T cells (CD4+CD25−) [77]. Expression of Foxp3 correlated with that of COX-2 in Treg cells, and importantly, the suppression of the effector T-cell response was reversed by COX inhibitors and PGE2 receptor (EP2 and EP4) antagonists (AH6809 and AH23848). Factors and pathways involved in activation or inhibition of COX-2 in gastric cancer in vitro are summarized in Fig. 1.
Fig. 1

Schematic representation of factors involved in regulation of cyclooxygenase-2 (COX-2) in vitro that have been described in this review. Pathways that are active upon H. pylori infection are shown in the yellow box. Notch 1 receptor intracellular domain (Notch 1 IC), C-promoter-binding factor 1 (CBF1), phosphatidylinositol 3-kinase (PI3K), glycogen synthase kinase 3β (GSK3β), microRNA 101 (miRNA 101), activator protein 1 (AP-1), human antigen R (HuR), nuclear factor κB (NF-κB), mitogen-activated protein kinases (MAPKs), toll-like receptor (TLR), cAMP response element (CRE), activating transcription factor 2 (ATF-2), cellular src (C-Src), mitogen-activated protein kinase kinase kinase 14 (MAP3K14 = NIK), extracellular signal-regulated kinases (ERK), cAMP response element binding (CREB), upstream stimulatory factor (USF)

4 In vivo models

The use of mouse gastric tumor models suggests that the prostanoid pathway plays a crucial role in the promotion of gastric tumors [57, 7880]. Mice that express COX-2 and mPGES-1 simultaneously in gastric epithelial cells (K19-C2mE) developed hyperplasia with spasmolytic polypeptide (TFF2)-expressing metaplasia (SPEM) in the proximal glandular stomach [57, 81]. The hyperplastic lesions were heavily infiltrated with macrophages. When COX-2 was inhibited with the selective inhibitor NS-398 in these mice, the macrophage infiltration was suppressed to the wild type level and gastric hypertrophy was completely suppressed [57].

Importantly, when in addition to COX-2 and mPGES-1, the oncogenic Wnt pathway was activated, the K19-Wnt1/C2mE mice (Gan mice) developed gastric adenocarcinomas by 20 weeks of age [78]. Upon Wnt signaling the undifferentiated progenitor cell population expanded while the factor mainly responsible for increased tumor angiogenesis and increased proliferation was the elevated PGE2 signaling. Interestingly, tumors of the Gan mice have a similar gene expression profile as human intestinal gastric cancer [82]. When Gan mice were treated with celecoxib and ZD1839, an EGFR inhibitor, the tumor volume was decreased by 90% and 76%, respectively, and a combination of both drugs led to a complete regression of the tumors [83]. Additionally, treatment of these mice with an EP4 inhibitor (RQ-00015986/CJ-42794) led to a 76% regression of mean tumor size [84]. Ligands for EGFR and metalloproteinases (that shed the ectodomains of EGFR ligands and thus activate them) were upregulated directly and indirectly by the PGE2 pathway through EP4 in Gan mouse tumors [83]. The activation of the EGFR pathway by PGE2 signaling might be responsible for tumor cell proliferation, as both COX-2 and EGFR inhibition decreased the number of Ki-67-positive cells [83].

In the trefoil factor 1 (TFF1) knockout model the mice develop gastric adenomas at the pyloric antrum [85]. COX-2 is expressed in stromal cells in these adenomas, and we used different approaches to investigate the importance of COX-2 at different stages of tumor development [86]. Treatment with the COX-2 selective inhibitor celecoxib (1,600 ppm for 3 months), that started before the adenomas had developed, induced ulceration and inflammation at the site of the adenoma exclusively [86]. Importantly, the same phenotype of disrupted and inflamed adenoma was observed when TFF1−/− mice were depleted of the COX-2 gene, and thus confirmed that the underlying mechanism was COX-2 specific [87]. Also, the treatment of fully developed adenomas with celecoxib (8–14 weeks) lead to adenoma regression and ulceration (Fig. 2). These findings underline the requirement for COX-2-derived products for tumor integrity and promotion.
Fig. 2

Histology of the gastric tissues of TFF1−/− mice according to treatment. a Intact pyloric adenoma of a mouse treated for 12 weeks with control food. The duodenal Brunner glands are visible in the left side of the figures. b Deep injury and strong chronic transmural inflammation of a mouse treated for 14 weeks with celecoxib (1,600 ppm). Residual adenoma tissue is marked with an arrowhead. Original magnification ×40 (a, b) and ×100 (enlargements)

Models of nonneoplastic lesions with hamartoma development are the Lkb1+/− and K19-Nog/C2mE mice [79, 88]. In the Lkb1+/− model the mice develop gastrointestinal hamartomatous polyps [88]. A significant fraction of these polyps expressed high levels of COX-2 at early stages of tumor growth and tumor burden was reduced by more than 50% in mice that were COX-2 heterozygous or deficient. Importantly, the reduction was due to a decrease in large polyps (>2 mm) as the number of small polyps did not change. When the Lkb1+/− mice were treated with celecoxib before or after the onset of polyposis the number of large polyps was significantly decreased in the treatment groups [88]. In this model COX-2 has a role in tumor promotion but not in tumor initiation, since some early polyps did not contain elevated levels of COX-2 and the number of small polyps remained unchanged after COX-2 inhibition. A possible mechanism might be the COX-2-dependent promotion of angiogenesis in the polyps of Lkb1+/− mice, as a significant decrease in microvessel density was observed after COX-2 inhibition [88]. The K19-Nog/C2mE mice express noggin, an endogenous antagonist of bone morphogenetic protein, in addition to COX-2 and mPGES-1, and develop large hamartomatous tumors in the glandular stomach [79]. Celecoxib treatment of the K19-Nog/C2mE hamartomas for 3 weeks lead to a 42% decreased tumor size with induced necrotic areas, and also the number of blood vessels decreased significantly [79]. These two models suggest that promotion of angiogenesis by PGE2 signaling might be important for hamartoma development.

The role of COX-2 in hypertrophic gastric mucosa was studied in gastrin transgenic mice [89]. These hypergastrinemic (ACT-GAS) mice develop gastric mucosal hypertrophy that can progress to noninvasive intramucosal adenocarcinoma at later stages (80 weeks old mice). COX-2 levels were elevated in interstitial cells in hypertrophic tissues in 16 weeks old mice and also in tumor tissues of 80 weeks old mice. Treatment of 5 weeks old mice for 11–19 weeks with celecoxib reduced the mucosal thickness and this was due to the reduction of foveolar thickness but not glandular thickness. The total gastric cell count was decreased upon celecoxib treatment, but the number of proliferative cells remained unchanged. However, the number of apoptotic cells increased upon treatment with celecoxib. PGE2 levels were increased in ACT-GAS mice compared to wild-type mice, but the PGE2 level returned to basal level after celecoxib treatment. Importantly, the treatment of already existing mucosal hypertrophy with celecoxib also lead to a significant reduction of the mucosal thickness. The increased levels of PGE2 in the ACT-GAS mice might directly promote prolonged survival of foveolar cells, possibly through EP4 receptor signaling [89].

COX-2 expression also plays a role in chemically and H. pylori-induced gastric cancer development in mice [84, 9092]. Whereas H. pylori or N-methyl-N-nitrosurea (MNU) treatment alone only rarely lead to tumor development in C57BL/6 mice, the combination of both resulted in adenomas or adenocarcinomas in 11 out of 16 mice [90]. When the (relatively) selective COX-2 inhibitor nimesulide was administered long-term, gastric tumorigenesis was significantly attenuated in H. pylori and N-methyl-N-nitrosurea-treated mice and apoptosis was increased in these tumors [90]. In a different approach, wild-type or COX-2 transgenic mice were treated with MNU. Whereas COX-2 overexpression alone did not lead to tumor development, the tumor incidence after MNU treatment was almost double in the COX-2 transgenic mice, suggesting a role of tumor promotion for COX-2 [91]. Also, gastric tumor multiplicity was higher in K19-C2mE mice that were treated with MNU and H. pylori compared to treated wild-type mice [92].

5 Conclusions

COX-2 expression is elevated in gastric carcinomas and in their precursor lesions, and it provides valuable clinical information as a prognostic factor. Several pathways can promote COX-2 expression in gastric cancer cells (Fig. 1), including H. pylori infection and subsequent release of inflammatory mediators. In rodent models of gastric tumorigenesis, tumor initiation seems to be induced by a special event (TFF1 deletion, Wnt overexpression, reduced Lkb1 expression, and MNU treatment), whereas COX-2-derived products seem to play a role predominantly in tumor promotion. Epidemiologic studies suggest that the use of aspirin and other NSAIDs reduces mortality from gastric cancer, which has also been demonstrated by a recent meta-analysis [93]. Two placebo-controlled randomized trials have been done on the effect of COX-2 inhibitors on gastric lesions. Leung et al. showed that rofecoxib treatment did not have an effect on intestinal metaplasia regression [94], whereas in the study by Zhang et al. celecoxib induced a significant regression of gastric precancerous lesions [95]. These data should encourage further prospective clinical trials investigating the clinical use of COX-2 inhibitors in patients with gastric neoplastic lesions.

Copyright information

© Springer Science+Business Media, LLC 2011