Masanobu Oshima, DVM, PhD, Division of Genetics, Cancer Research Institute, Kanazawa University, Kakuma-machi, Kanazawa 920-1192, Japan. Email: firstname.lastname@example.org
Accumulating evidence has suggested that cooperation of oncogenic activation and the host responses is important for cancer development. In gastric cancer, activation of Wnt signaling appears to be a major oncogenic pathway that causes tumorigenesis. In the chronic gastritis caused by Helicobacter pylori infection, cyclooxigenase-2 induces prostaglandin E2 (PGE2) biosythesis, which plays an important role in tumorigenesis. We constructed a series of mouse models and investigated the role of each pathway in the gastric tumorigenesis. Wnt activation in gastric epithelial cells suppresses differentiation, and induces development of preneoplastic lesions. On the other hand, induction of the PGE2 pathway in gastric mucosa induces development of spasmolytic polypeptide-expressing metaplasia (SPEM), which is a possible preneoplastic metaplasia. Importantly, simultaneous activation of Wnt and PGE2 pathways leads to dysplastic gastric tumor development. Moreover, induction of the PGE2 pathway also promotes gastric hamartoma development when bone morphogenetic protein (BMP) signaling is suppressed. These results indicate that alteration in the Wnt or BMP signaling impairs epithelial differentiation, and the PGE2 pathway accelerates tumor formation regardless of the types of oncogenic pathways. We review the phenotypes and gene expression profiles of the respective models, and discuss the cooperation of oncogenic pathways and host responses in gastric tumorigenesis.
Gastric cancer is the second most common cause of cancer-related death worldwide.1 Infection with Helicobacter pylori is associated with gastric cancer development, and the International Agency for Research on Cancer (IARC) classified Helicobacter pylori as a class I carcinogen.2 Accumulating evidence has indicated that chronic inflammatory response associated with infectious disease is a critical component of tumor development.3 Moreover, it has been shown that infections are responsible for more than 15% of all malignant cancers worldwide, including the association between H. pylori infection and gastric cancer.4 Notably, host genetic variants in cytokine genes are related to responsiveness to H. pylori infection and the susceptibility to gastric cancer development.5–7 Specific polymorphisms of interleukin (IL)-1β, an important inflammatory cytokine and a potent inhibitor of gastric acid secretion, contribute to intestinal-type gastric cancer progression.8 Polymorphisms in tumor necrosis factor (TNF)-α, IL-1 receptor antagonist, and IL-10 also influence gastric cancer development,8–10 while polymorphisms in the IL-8 promoter have been linked to diffuse-type gastric cancer.11 These results suggest that the response of the host cytokine network to H. pylori infection is an important factor for gastric cancer development.
On the other hand, several somatic alterations that activate oncogenic pathways have been identified in human gastric cancer. For example, allelic loss or mutations in p53 are detected in 60% or 30–50% of gastric cancers, respectively,12 while mutations in the β-catenin gene is detected in 30% of the Wnt-activated subgroup of gastric cancer.13 TGF-β type II receptor gene is recognized as a tumor suppressor, and mutations have been found in gastric cancer associated with microsatellite instability (MSI).14,15 Moreover, about 15% of gastric cancers show expression of both epidermal growth factor (EGF) and EGF receptor (EGFR), suggesting activation of the EGFR signaling pathway.16 Taken together, these results indicate that both infection-associated inflammatory responses and oncogenic activation by genetic alterations are required for gastric cancer development.
To date, many genetic mouse models have recapitulated some of the developmental stages associated with intestinal-type gastric cancer, such as gastritis, atrophy, mucous cell metaplasia, dysplasia, and invasion.17 These models are useful for examining the phenotypic changes caused by individual genetic alterations. In addition to these models, we constructed a series of transgenic mouse models to investigate the role of oncogenic and inflammatory pathways, such as Wnt, BMP, and PGE2 signaling, in gastric epithelial differentiation, inflammation, and tumor development.
ACTIVATION OF WNT SIGNALING IN GASTRIC MUCOSA
Wnt activation in human gastric cancer
Canonical Wnt signaling (Wnt/β-catenin signaling) is a critical pathway in the regulation of development as well as in tumorigenesis.18 When Wnt signaling is in a resting state, cytoplasmic β-catenin is phosphorylated by GSK-3β within a complex containing APC and Axin, resulting in the degradation of β-catenin through the ubiquitin proteasome pathway.19 When the pathway is activated, the binding of Wnt ligands to Frizzled receptors leads to the suppression of the phosphorylation of β-catenin, resulting in the to stabilization and nuclear translocation of β-catenin. β-Catenin then interacts with T-cell factor/lymphocyte enhancer factor (TCF/LEF) to induce transcription of Wnt target genes. In the normal intestine, Wnt signaling is important for maintaining the stem cell characteristics and undifferentiated status of the epithelial cells, whereas Wnt signaling is suppressed in the differentiated epithelia (Fig. 1)19–21 Mutations in the APC or β-catenin genes constitutively activate Wnt signaling, which causes tumor development in the intestine.22,23 In the gastric mucosa, epithelial cells expressing Lgr5, which is a target of Wnt signaling, show stem cell phenotypes, confirming the role of Wnt pathway in normal gastric stem cells.24 Moreover, nuclear accumulation of β-catenin, a hallmark of Wnt signaling activation, is found in 30–50% of gastric cancers (Fig. 2),13,25–27 and mutations in the β-catenin gene have also been detected,13,25,28,29 which suggest that activation of Wnt signaling is a major cause of gastric cancer development. However, APC gene mutations are not common in gastric cancer, and β-catenin mutations are present in fewer than 30% of the Wnt-activated gastric cancers.25 Accordingly, it is possible that other mechanism(s) may also activate Wnt signaling in gastric cancer. For example, it has been reported in gastric cancer cells that downregulation of E-cadherin is associated with β-catenin accumulation,30 somatic mutations in the ubiquitin ligase β-TrCP causes stabilization of β-catenin,31 and the expression of the SFRP1, 2 and 5 genes encoding the secreted endogenous antagonist of the Wnt ligands is silenced by promoter methylation in gastric cancer cells.32 All of these alterations contribute to activation of Wnt signaling.
Gastric preneoplastic lesions in K19-Wnt1 transgenic mice
To examine the role of Wnt signaling in gastric tumorigenesis, we constructed K19-Wnt1 transgenic mice that express Wnt1, one of the canonical Wnt ligands.27 The K19 gene promoter was used to express Wnt1 in gastric epithelial cells, including undifferentiated isthmal cells.33,34 The number of undifferentiated epithelial cells that express trefoil factor 2 (TFF2) increases significantly in the K19-Wnt1 mouse glandular stomach, indicating that Wnt signaling functions to maintain the undifferentiated status of the gastric epithelial cells (Fig. 3a,b). Notably, aberrant cryptic foci were found on the mucosal surface of K19-Wnt1 mice, which consist of dysplastic epithelial cells with irregular branching, and increased cell proliferation and β-catenin accumulation were also detected (Fig. 3c–g). We thus diagnosed these foci as preneoplastic lesions. However, gastric tumors do not develop in K19-Wnt1 mice, indicating that activation of Wnt signaling alone is not sufficient for gastric tumor formation. Importantly, macrophages were infiltrated into the preneoplastic lesions, whereas tissue macrophages were sparsely scattered in the normal mucosa of the same mice. It is possible that local inflammatory responses caused by spontaneous physical insult or infection might promote proliferation of Wnt-activated dysplastic cells, resulting in formation of the preneoplastic lesions.
INDUCTION OF PGE2 PATHWAY IN GASTRIC MUCOSA
Induction of COX-2/PGE2 pathway in human gastric cancer
Epidemiological studies indicate that the regular use of non-steroidal anti-inflammatory drugs (NSAIDs) is associated with a decreased incidence of gastric cancer.35–38 NSAIDs inhibit the enzymatic activity of cyclooxygenases (COXs), which are rate-limiting enzymes for prostaglandin biosynthesis. COX enzymes catalyze synthesis of prostaglandin (PG)H2, which is subsequently converted to various prostanoids, including PGE2, by tissue-specific converting enzymes. There are two COX isozymes, COX-1 and COX-2, which share a high degree of structural and enzymatic homology. COX-1 is constitutively expressed in most tissues and is considered to be responsible for maintaining physiological levels of prostaglandin biosynthesis,39 while COX-2 expression is induced in inflammation and in tumor tissues by cytokines and growth factors.40–42 Induction of COX-2 is found in approximately 70% of gastric cancer, predominantly in intestinal-type gastric cancer, whereas COX-1 expression is not elevated.43–45 Moreover, the level of COX-2 expression in gastric cancer correlates with the tumor size, depth of invasion and lymph-node metastasis.46–48 Microsomal PGE synthase-1 (mPGES-1) is an inducible enzyme that converts PGE2 from PGH2, and is functionally coupled with COX-2.49 Simultaneous induction of COX-2 and mPGES-1 has been observed in a variety of cancers, including gastric cancer, suggesting that the inflammatory PGE2 pathway is induced in these tumors (Fig. 1).50–53 Consistently, the PGE2 level is found to be significantly increased in gastric cancer,47 and the COX-2 and PGE2 level is associated with the H. pylori infection status,54,55 indicating that H. pylori infection causes induction of the PGE2 pathway (Fig. 1). These results suggest that the COX-2/PGE2 pathway plays a key role in H. pylori infection-associated inflammation in gastric cancer development.
Suppression of gastric cancer by COX-2 inhibition in animal models
Suppression of gastric tumorigenesis by treatment with COX-2 selective inhibitors (COXIBs) has been examined in several animal model experiments. Growth of gastric cancer cell xenografts was inhibited by treatment with COXIBs in immunodeficient mice.56,57 Rat gastric cancer induced by carcinogen N-methyl-N′-nitro-N-nitrosoguanidine (MNNG) and mouse gastric tumors induced by H. pylori infection or a combination of H. pylori with N-methyl-N-nitrosourea (MNU) were suppressed by treatment with NSAIDs or COXIBs.58–60 A significant decrease in the PGE2 level and substantial induction of apoptosis were also found in the tumors of mice treated with NSAIDs or COXIBs. H. pylori infection of Mongolian gerbils is an established model to study gastric tumorigenesis by H. pylori. Treatment of the H. pylori-infected Mongolian gerbils with COXIBs suppressed the development of gastric cancer, as well as intestinal metaplasia.,61,62 These animal studies indicate that the COX-2 pathway plays a key role in gastric tumorigenesis.
Gastric metaplasia in K19-C2mE transgenic mice
To examine the effect of PGE2 pathway activation in the gastric mucosa, we constructed another transgenic mouse model, K19-C2mE, which expresses both COX-2 and mPGES-1 simultaneously in gastric epithelial cells.34K19-C2mE mice developed hyperplasia in the gastric mucosa, which was suppressed by treatment of the mice with a COX-2 selective inhibitor, NS-398 or meloxicam (Fig. 4a,b).34,63 Histologically, the major cell type involved in the hyperplasia were mucous cells, which was similar to the findings in the spasmolytic polypeptide/TFF2-expressing metaplasia or SPEM (Fig. 4c,d).34,64 The SPEM is characterized by the presence of TFF2 immunoreactive cells, which is morphologically similar to those of Brunner's gland. Since SPEM was associated with greater than 90% of the resected gastric cancers analyzed in three studies,64–66 it is suggested that SPEM is a putative preneoplastic metaplasia in the stomach.67 In other animal models, SPEM development was also found in the stomach of Helicobacter-infected mice, gastrin gene knockout mice, and in Stat3-activating gp130 mutant mice.68–70 Notably, SPEM in these mouse models was accompanied by inflammatory responses. Importantly, disruption of the TNF-α gene in K19-C2mE mice results in the suppression of inflammatory responses and SPEM development, suggesting that TNF-α-associated inflammation plays an essential role in SPEM formation (Fig. 4e–g).63 In contrast, Rag2 gene disruption did not suppress SPEM formation in the K19-C2mE mice, indicating that acquired immune responses are not involved in SPEM development.
Treatment with MNU to the H. pylori-infected mice led to the development of gastric tumors. Notably, multiplicity of gastric tumors induced by H. pylori infection and MNU treatment was significantly higher in K19-C2mE mice compared with wild-type mice.71 These results suggest that PGE2-induced SPEM is a precursor for chemical carcinogen-induced gastric tumors.
ONCOGENIC ACTIVATION AND PGE2 PATHWAY IN GASTRIC TUMORIGENESIS
Gastric dysplastic tumors in K19-Wnt1/C2mE transgenic mice
Compound K19-Wnt1/C2mE transgenic mice were generated by crossing K19-Wnt1 mice and K19-C2mE mice, in which both Wnt and PGE2 pathways are activated simultaneously in the gastric mucosa.27K19-Wnt1/C2mE mice develop large gastric tumors in the glandular stomach (Fig. 5a,b), while no such tumors are found in either simple K19-Wnt1 or K19-C2mE transgenic mice. Histologically, gastric tumors in the K19-Wnt1/C2mE mice (hereafter Gan mice for gastric neoplasia) consist of dysplastic epithelial cells with nuclear stratification and irregularly branched tubules (Fig. 5c). Tumor epithelial cells showed increased nuclear accumulation of β-catenin and increased Ki-67 labeling, indicating promotion of Wnt activity and an increased proliferation of tumor cells, respectively (Fig. 5e). Increased capillary vessels were also found in the tumor stroma, which was caused by enhanced angiogenesis through the tumor-stroma interaction (Fig. 5f,g).72 Interestingly, part of the activated stromal fibroblasts in Gan mouse tumors were derived from bone marrow, thus suggesting that the bone marrow cells contribute to the formation of the gastric cancer microenvironment.72
Notably, TFF2-expressing SPEM was found adjacent to the dysplastic tumor tissue,27 showing a similar histology to human gastric cancer (Fig. 5h). K19-C2mE mice start to develop SPEM lesions at 5 weeks of age, and the number of TFF-positive metaplastic cells increases with age. In the Gan mouse stomach, the same SPEM phenotype is found at 5 weeks of age, however, dysplastic tumor cells are also found beginning at 10 weeks of age. The number of dysplastic tumor cells then increases with age, leading to formation of gastric tumors around 20–30 weeks of age, with SPEM being found adjacent to the tumors. These results, taken together, indicate that the simultaneous activation of the Wnt and PGE2 pathways causes gastric tumor development through the metaplasia (SPEM)-carcinoma sequence.
Gastric hamartomas in K19-Nog/C2mE transgenic mice
Juvenile polyposis syndrome (JPS) is characterized by hereditary gastrointestinal hamartomatous polyposis,73 and a subset of JPS is caused by germline mutations in the BMP receptor type IA gene (BMPR1A).74 BMP signaling through its type I and II receptors leads to the phosphorylation of Smad 1,5, and 8, resulting in formation of a complex with Smad4, which induces transcription of target genes.75 BMP signaling inhibits epithelial cell proliferation and promotes differentiation (Fig. 1), and suppression of BMP signaling in the mouse intestine results in JPS-type hamartomatous polyp development,76–78 elongated villi and crypt fission.79 Accordingly, it is possible that BMP suppression results in hamartoma formation by impairment of epithelial differentiation. Since the cancer risk in JPS patients increases in the gastrointestinal tract,80,81 BMP suppression may also contribute to gastric cancer development.
To examine the effect of BMP suppression in gastric epithelial cells, we next constructed K19-Nog mice that express noggin, an endogenous BMP antagonist, in the gastric epithelial cells.82 Although BMP signaling was suppressed in the stomach, K19-Nog mice do not develop any gastric lesions, and the histology of the gastric mucosa was normal. To examine the effect of cooperation of BMP suppression and PGE2 induction, K19-Nog mice and K19-C2mE mice were crossed to construct compound K19-Nog/C2mE mice, in which BMP signaling is suppressed and the PGE2 pathway is induced in the gastric mucosa. Importantly, the K19-Nog/C2mE mice developed large tumors in the glandular stomach (Fig. 6a), suggesting that induction of the PGE2 pathway is required for tumor formation in the BMP-suppressed gastric mucosa. Histologically, K19-Nog/C2mE mouse gastric tumors are not dysplastic, but consist of irregular branching of the epithelial cell layers, combined with the formation of dilated cysts filled with mucin (Fig. 6b,c). Such histological characteristics are distinct from the dysplastic gastric tumors of Gan mice (Fig. 5c), but are typical of the hamartomas of JPS patients.80,81,83 These results indicate that the suppression of BMP signaling associated with PGE2 induction causes gastric hamartoma development.
These results of compound mutant mice indicate that the type of genetic alteration, such as Wnt activation or BMP suppression, determines the histological type of tumors, such as adenocarcinoma or hamrtoma.84 On the other hand, induction of the PGE2 pathway promotes gastric tumor formation regardless of the genetic or histological types.
Gene expression profiles of mouse models and human gastric cancer
Gan mice and K19-Nog/C2mE mice develop gastric tumors caused by genetic alterations similar to those found in human gastric cancer and hamartomas, respectively. However, it is still important to compare gene expression profiles of mouse tumors with those of human cancer or hamartomas in order to examine whether these models really recapitulate human gastric tumors. We have measured mRNA expression levels using the Affymetrix GeneChip system, which includes 21 066 Entrez genes and 5324 other sequences.85 Genome-scale overview of the microarray data revealed that expression changes in the three models, K19-C2mE and K19-Wnt1/C2mE (Gan), and K19-Nog/C2mE mice are quite similar, whereas over-expression of Wnt1 or Noggin in K19-Wnt1 or K19-Nog mice, respectively, showed expression changes in a small portion of genes (Fig. 7a). These results suggest that most of the expression changes in Gan gastric tumors and K19-Nog/C2mE hamartomas are caused by induction of PGE2 pathway, rather than by Wnt activation or BMP suppression. In other words, a small number of genes that are upregulated or downregulated by Wnt activation or BMP suppression are important for determining the tumor phenotype.
Gene expression signatures of human gastric cancer86 and breast cancer87 retrieved from the Stanford Microarray Database,88 colon cancer89 from the NCBI GEO (accession GSE5206), and lung tumors90 retrieved from the United States National Cancer Institute website91 can be plotted to distinct areas in a 3D figure based on the calculations of principal component analysis using the selected genes (Fig. 7b). Importantly, expression signatures of Gan, K19-C2mE, and K19-Nog/C2mE mice are clustered in a similar area as that of human gastric cancer, but not to the same area as cancers of other organs. These results indicate that Gan mouse tumors recapitulate human gastric cancer from the molecular etiology to histology and gene expression profiles. It is also possible that most of the changes in gene expression in human gastric cancer are attributable to H. pylori infection-associated inflammatory responses. Taken together, these results indicate that the Gan mouse model is a useful tool for studying the effects of oncogenic activation and inflammatory responses in human gastric cancer development and the evaluation of anti-gastric cancer drugs.
Wnt signaling functions to maintain the undifferentiated status of gastric epithelial cells. On one hand, activation of Wnt signaling by genetic or epigenetic alteration causes development of preneoplastic lesions. On the other hand, H. pylori infection induces expression of COX-2 and mPGES-1, resulting in induction of PGE2 biosynthesis. Induction of the COX-2/PGE2 pathway is responsible for SPEM development, which is a possible preneoplastic metaplasia of gastric cancer. Importantly, simultaneous induction of Wnt and PGE2 pathways causes development of dysplastic gastric tumors (Fig. 1). The results of mouse model studies reported herein suggest that oncogenic activation, such as activation of Wnt signaling by genetic or epigenetic alterations, triggers tumor initiation. However, initiated epithelial cells cannot continue proliferation in the non-inflamed gastric mucosa, thus indicating that Wnt signaling alone is not sufficient for tumor development. In contrast, if the oncogenic pathway is activated in the H. pylori-infected (and thus inflamed) stomach, the initiated cells proliferate to develop gastric cancer. Considering the relatively low frequency of somatic oncogenic activation, this hypothesis is consistent with the epidemiology of gastric cancer, in that only a small minority of the H. pylori-infected population develops gastric cancer, although H. pylori infection is an important risk factor for gastric cancer. It is possible that a similar mechanism underlies the development of hamartoma in the BMP-suppressed gastric epithelial cells (Fig. 1). Further studies using these mouse models will be useful for elucidating the role of oncogenic activation and host responses in gastric tumorigenesis at the molecular level.
We thank Hiraku Itadani and Hidehito Kotani for microarray analyses, and Manami Watanabe for her technical contributions to the gastric tumor mouse model studies.