Article Text

Download PDFPDF

Original article
Synergistic tumour suppressor activity of E-cadherin and p53 in a conditional mouse model for metastatic diffuse-type gastric cancer
  1. Shu Shimada1,
  2. Ayako Mimata1,
  3. Masaki Sekine2,
  4. Kaoru Mogushi3,
  5. Yoshimitsu Akiyama1,
  6. Hiroshi Fukamachi1,
  7. Jos Jonkers4,
  8. Hiroshi Tanaka3,
  9. Yoshinobu Eishi2,
  10. Yasuhito Yuasa1
  1. 1Department of Molecular Oncology, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University
  2. 2Department of Human Pathology, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University, Tokyo, Japan
  3. 3Department of Systems Biology, Graduate School of Biomedical Science, Tokyo Medical and Dental University, Tokyo, Japan
  4. 4Division of Molecular Biology, the Netherlands Cancer Institute, Amsterdam, The Netherlands
  1. Correspondence to Professor Yasuhito Yuasa, Department of Molecular Oncology, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University, 1-5-45 Yushima, Bunkyo-ku, Tokyo 113-8519, Japan; yuasa.monc{at}tmd.ac.jp

Abstract

Background Gastric cancer is the second most frequent cause of death from cancer in the world, diffuse-type gastric cancer (DGC) exhibiting a poor prognosis. Germline mutations of CDH1, encoding E-cadherin, have been reported in hereditary DGC, and genetic and/or epigenetic alterations of CDH1 are frequently detected in sporadic DGC. Genetic alterations of TP53 are also frequently found in DGC. To examine the synergistic effect of the loss of E-cadherin and p53 on gastric carcinogenesis, a mouse line was established in which E-cadherin and p53 are specifically inactivated in the stomach parietal cell lineage.

Methods Atp4b-Cre mice were crossed with Cdh1loxP/loxP and Trp53loxP/loxP mice, and the gastric phenotype of Atp4b-Cre+;Cdh1loxP/loxP;Trp53loxP/loxP double conditional knockout (DCKO) mice was examined.

Results Non-polarised E-cadherin-negative parietal cells and proton pump-negative atypical foci were observed in DCKO mice. Intramucosal cancers and invasive cancers composed of poorly differentiated carcinoma cells and signet ring cells, histologically very similar to those in humans, were found from 6 to 9 months, respectively. Fatal DGC developed at 100% penetrance within a year, frequently metastasised to lymph nodes, and had tumourigenic activity in immunodeficient mice. Gene expression profiles of DGC in DCKO mice also resembled those of human DGC, and mesenchymal markers and epithelial-mesenchymal transition-related genes were highly expressed in mouse DGC as in human DGC.

Conclusion This mouse line is the first genetically engineered mouse model of DGC and is very useful for clarifying the mechanism underlying gastric carcinogenesis, and provides a new approach to the treatment and prevention of DGC.

  • Cancer
  • carcinogenesis
  • differentiation
  • E-cadherin
  • epithelial cell growth
  • epithelial proliferation
  • gastric cancer
  • gastrointestinal cancer
  • gene expression
  • genetically engineered mouse
  • intestinal development
  • intestinal metaplasia
  • p53
  • stem cells
View Full Text

Statistics from Altmetric.com

Significance of this study

What is already known about this subject?

  • Gastric cancer is the second most frequent cause of death from cancer in the world. DGC in particular carries a poor prognosis.

  • Germline mutations of the CDH1 gene, encoding E-cadherin, have been identified in hereditary DGC, and mutations and hypermethylation of the CDH1 gene have frequently been observed in sporadic DGC, suggesting that the loss of E-cadherin is responsible for diffuse-type gastric carcinogenesis.

  • There is, however, no genetically engineered mouse model of DGC. We previously established E-cadherin CKO mice, in which no cancer was induced after 2 years.

What are the new findings?

  • We established an E-cadherin/p53 DCKO mouse model, because p53 deficiency has also been identified in advanced gastric cancer.

  • In DCKO mice, fatal DGC always developed within a year and frequently metastasised to lymph nodes. Moreover, they were very similar to human DGC morphologically and molecularly.

How might it impact on clinical practice in the foreseeable future?

  • The synergistic effects of the inactivation of E-cadherin and p53 can be necessary for the development of DGC.

  • Our mouse model is very useful to clarify the mechanism of DGC and to develop new treatment for DGC.

Gastric cancer is the second most frequent cause of death from cancer in the world.1 Gastric cancers are histologically classified into two major types, intestinal type (well differentiated) and diffuse type (poorly differentiated).2 3 The intestinal-type gastric cancer (IGC) is mainly associated with Helicobacter pylori infection and resultant sequential lesions, such as gastritis and intestinal metaplasia.4 The diffuse-type gastric cancer (DGC) is often refractory to surgical and pharmaceutical treatments and has a poorer patient prognosis, but the mechanism underlying DGC formation is not yet fully understood.

In human DGC, mutations and loss of the CDH1 gene encoding E-cadherin, have frequently been detected,5 and the CDH1 promoter also often undergoes hypermethylation.6 Moreover, germline mutations of the CDH1 gene have been observed in hereditary DGC.7 These data indicate that CDH1 is an important tumour suppressor gene of DGC.

E-cadherin is the founder member of the cadherin superfamily of calcium-dependent cell adhesion molecules and plays an important role in cell polarity.8 E-cadherin has five extracellular domains and a conserved intracellular domain with motif binding catenins, α-catenin and β-catenin, which are linked to the actin cytoskeleton. It has been demonstrated that the loss of E-cadherin promotes tumour infiltration and metastatic dissemination,9 and the significance of the loss of E-cadherin for invasiveness and metastasis has been shown in various cancers, such as gastric10 and breast11 cancers.

To confirm whether CDH1 is a tumour suppressor gene of DGC, and to develop a new approach to the treatment and prevention of DGC, it is essential to establish a mouse DGC model with a modified Cdh1 gene. However, a trophectoderm epithelium is not formed in Cdh1 null mutant mouse embryos, which leads to embryonic lethality.12 We thus used the Cre-loxP system to establish a Atp4b-Cre+;Cdh1loxP/loxP conditional knockout (CKO) mouse line in which E-cadherin is specifically lost in the gastric parietal cell lineage. However, gastric carcinomas were not induced even after 2 years in CKO mice.13 Conditional targeting of Cdh1 in the mammary gland14 or skin15 does not induce tumours either, suggesting that the loss of E-cadherin is not sufficient for tumour formation.

TP53 is another candidate tumour suppressor gene in DGC, because TP53 mutations have frequently been found in gastric cancers including DGC.16 17 Germline mutations in TP53 have been found in Li-Fraumeni syndrome, a familial syndrome of sarcomas, breast cancer and other neoplasms.18 19 Interestingly, TP53 germline mutations have occasionally been found in hereditary DGC families including patients with other types of tumours.20 21 There have been many types of Trp53-deficient mice, but gastric cancers have not been induced in these mice.22 On the other hand, metastatic lobular mammary carcinomas, which strongly resemble human invasive lobular breast carcinoma, develop in conditional knockout mice with tissue-specific inactivation of E-cadherin and p53 in mammary glands,23 indicating that the loss of E-cadherin and p53 is a cooperating mutation in tumour development.

Therefore, to determine the effect of E-cadherin/p53 double knockout in the stomach mucosa we established a double conditional knockout mouse line in which E-cadherin and p53 are specifically inactivated by Cre-mediated recombination under the control of the Atp4b gene promoter, which is known to act exclusively in the parietal cell lineage of the stomach.24

Materials and methods

Animals

We carried out Cre-mediated deletion of floxed alleles in the germline by crossing Cdh114 and Trp5325 conditional mutants with Atp4b-Cre transgenic mice.24 To minimise any genetic background differences, we used Atp4b-Cre;Cdh1loxP/loxP;Trp53loxP/loxP littermate mice as negative controls. At least six mice were killed at 3, 6 and 9 months, and all other mice were killed at 12 months and their stomachs were histologically analysed. All experiments involving mice were conducted using protocols approved by the Institutional Animal Care and Use Committee of Tokyo Medical and Dental University.

DNA analysis

We performed genotyping by PCR using tail-tip DNA as described previously.14 24 25 The PCR primer sequences and conditions are given in supplementary table 1 (available online only). Other DNA was extracted from the fomalin-fixed and paraffin-embedded tissue sections as reported previously.26 The truncated alleles of Cdh1 and Trp53 genes were detected by PCR amplification of the tissue DNA as described previously.14 25 Atp4b-Cre allele was used as an internal control.24

Histological analysis

Tissues were isolated from mice and fixed in 20% formalin in phosphate buffer for 12–24 h. They were dehydrated, embedded in paraffin and serially sectioned at 5 μm thickness. Sections were stained with haematoxylin and eosin, and Alcian blue (AB) (pH 2.5) periodic acid Schiff (PAS). Pathological classification was performed according to the general rules established by the Japanese Gastric Cancer Association3 and Laurén's classification.2 For immunohistochemical analyses, Tris-EDTA (pH 9.0) or sodium citrate (pH 6.0) buffer was used for antigen retrieval. Primary antibodies are shown in supplementary table 2 (available online only). Histofine Simple Stain MAX-PO (Nichirei Bioscience, Tokyo, Japan) and alkaline phosphatase-conjugated goat anti-mouse IgM (1:100; Vector Laboratories, Burlingame, California, USA) were used as secondary antibodies. Diaminobenzidine was used as a chromogen, followed by counterstaining with haematoxylin. For Tff2 staining, Vectastain ABC-AP standard kit (Vector Laboratories) and Vector Red Alkaline Phosphatase Substrate Kit (Vector Laboratories) were used.

Immunofluorescent staining

For a detailed description, see the supplementary materials and methods (available online only).

BrdU staining

For a detailed description, see the supplementary materials and methods (available online only).

Transplantation

For a detailed description, see the supplementary materials and methods (available online only).

RNA extraction

Seven normal gastric mucosae, from which stromas were removed mechanically in 30 mM EDTA-Hanks' solution, and five primary gastric cancers were obtained from Atp4b-Cre;Cdh1loxP/loxP;Trp53loxP/loxP and Atp4b-Cre+;Cdh1loxP/loxP;Trp53loxP/loxP mice at 12 months, respectively. Three transplanted tumours were obtained from nude mice. Normal stromal tissues were excised from cancer tissues under a stereoscopic microscope. Total RNA from the tissues was isolated with TRIzol Reagent (Invitrogen, Carlsbad, California, USA). Contaminated DNA was removed by digestion with RNase-free DNase using a DNA-free kit (Applied Biosystems, Carlsbad, California, USA).

DNA microarray analysis

The integrity of obtained RNA was assessed using the Agilent 2100 BioAnalyzer (Agilent Technologies, Palo Alto, California, USA). All samples were confirmed to have an RNA integrity number greater than 7. Using 250 ng of total RNA, cRNA was prepared using the low RNA input fluorescent linear amplification kit PLUS, two-colour (Agilent), and labelled with Cy5 for three independent mouse DGC samples and Cy3 for a pool of five normal gastric mucosal samples. Hybridisation and signal detection of Mouse GE 4x44K v2 Microarray (G4846A; Agilent) was performed following the manufacturer's instructions. Gene expression data were acquired by Feature Extraction Software (Agilent) and were then transformed into a log2 ratio of Cy5/Cy3 for each gene.

Public microarray data of human gastric cancer

A microarray dataset for human gastric cancer data27 was downloaded from the authors' website (http://genome-www.stanford.edu/Gastric_Cancer2/). The dataset contained 68 IGC, 13 DGC and 15 normal gastric samples. Gene expression levels of IGC and DGC were normalised by mean expression levels of normal gastric samples for each gene. To examine gene expression similarity among IGC, DGC and tumours obtained from mice, we used ‘intestinal type versus diffuse type signature genes’28 containing 122 cDNA. Among them, we excluded the gene expression pattern of Cdh1 because the gene was knocked out in our mouse model. Homologous genes between human and mouse were determined using HomoloGene database release 64 (http://www.ncbi.nlm.nih.gov/homologene). When multiple probes were found for a gene, the average gene expression levels were calculated. R statistical software (R Foundation for Statistical Computing, Vienna, Austria) was used for microarray data analysis and hierarchical clustering.

Gene expression analysis

Single-stranded cDNA synthesis was performed using SuperScript III reverse transcriptase (Invitrogen), and reverse transcriptase PCR was carried out using the primer sets as shown in supplementary table 3 (available online only). Gapdh transcripts were amplified as internal controls.

Statistical analysis

The statistical significance of experimental results was determined with a Student's t test using Microsoft Excel 2003. Kaplan–Meier analysis was performed for the survival rate of mice and the difference was evaluated with the log rank test using R statistical software.

Results

Generation of Atp4b-Cre+;Cdh1loxP/loxP;Trp53loxP/loxP mice

To investigate the potential synergistic effects of stomach-specific inactivation of E-cadherin and p53 on tumour formation, we mated Atp4b-Cre transgenic mice24 with Cdh1loxP/loxP 14 and Trp53loxP/loxP 25 mice. The resultant Atp4b-Cre+;Cdh1loxP/loxP;Trp53loxP/loxP double conditional knockout (DCKO) mice were genetically confirmed using tail-tip DNA analysis (see supplementary figure 1, available online only). Atp4b encodes the β-subunit of H+, K+-ATPase and is exclusively expressed in the parietal cell lineage, that is, parietal cells and preparietal cells/parietal cell progenitors, and thus E-cadherin and p53 are predicted to be specifically inactivated in the parietal cell lineage in DCKO mice.

Loss of cell polarity in parietal cells in DCKO mice

Abnormalities of parietal cells were observed in DCKO mice from 3 months (figure 1A). Some of them were not triangular like normal parietal cells but were round in shape and were pushed out from the fundic glands to the stromata. These non-polarised parietal cells were proton pump positive and eosinophilic (figure 1A), indicating that they had dense mitochndoria, like normal parietal cells. These non-polarised parietal cells did not exhibit E-cadherin on their cell membrane, while other gastric epithelial cells of the DCKO mice and all parietal cells in control mice did. The ratio of non-polarised parietal cells to all parietal cells was similar from 3 to 9 months (figure 1B). The morphology and the ratio of non-polarised parietal cells in DCKO mice resembled those of type 1 cells in CKO mice (figure 1B).13 E-cadherin loss may thus be sufficient to induce the loss of cell polarity in parietal cells.

Figure 1

Loss of cell polarity in E-cadherin-negative parietal cell lineages. (A) Haematoxylin and eosin (HE) and immunofluorescent staining of sections of the stomachs of Atp4b-Cre;Cdh1loxP/loxP;Trp53loxP/loxP (Cre, left) and Atp4b-Cre+;Cdh1loxP/loxP;Trp53loxP/loxP (Cre+, right) mice at 6 months of age with antibodies against E-cadherin (red) and the proton pump (green). Arrows indicate typical non-polarised parietal cells pushed out from the fundic glands. Scale bars, 20 µm. (B) The frequency of appearance of non-polarised parietal cells in double conditional knockout (DCKO) (left) and conditional knockout (CKO) mice (right). The values represent percentages (means±SD) of non-polarised parietal cells in all parietal cells that were counted in eight fields (×20) in two or three sections per mouse.

Formation of proton pump-negative atypical foci

Spheroid clusters of atypical cells not expressing E-cadherin were observed in DCKO mice (figure 2A). They were composed of E-cadherin-negative, proton pump-negative but eosinophilic cells, which appeared to be derived from the parietal cell lineage. In the clusters, while most cells did not contain mucus, some cells contained AB-positive and/or PAS-positive mucus, suggesting that multipotent cells were present in the clusters. Many Ki67-positive cells were also observed (figure 2A). Therefore, it is possible that the E-cadherin/p53-deficient parietal cell lineage might have undergone abnormal differentiation and then formed proton pump-negative atypical foci. They were not observed at 3 months, but their number increased after 6 months (figure 2B).

Figure 2

Formation of proton pump-negative atypical foci. (A) Haematoxylin and eosin (HE), Alcian blue, periodic acid Schiff (AB-PAS) and immunohistochemical staining of serial sections of a 9-month-old double conditional knockout (DCKO) mouse stomach with antibodies against E-cadherin, the proton pump and Ki67. Dotted lines show some proton pump-negative atypical foci. Scale bars, 50 µm. (B) The numbers of proton pump-negative atypical foci in 3-, 6- and 9-month-old DCKO (left), and in 3-, 6- and 12-month-old conditional knockout (CKO) mice (right). The values represent means±SD of the number of them per area of the corpus regions, which were determined in two or three sections per mouse. Note that scales on the y-axis are different between the two mouse models. (C) Ki-67 labelling index of proton pump-negative atypical foci in DCKO and CKO mice. The values represent percentages (means±SD) of Ki67-positive cells in all cells in the lesions. (D) Immunohistochemical staining of sections including proton pump-negative atypical foci in DCKO (left) and CKO (right) mice with antibodies against p53. Dotted lines surround proton pump-negative atypical foci. Arrowheads exhibit non-polarised parietal cells. Scale bars, 50 µm.

The proton pump-negative atypical foci were morphologically similar to E-cadherin/proton pump-negative cell clusters, termed as type 2, in CKO mice.13 However, the number of atypical foci in DCKO mice was much larger than that in CKO mice (figure 2B), and the Ki67 labelling index of them in DCKO mice was significantly higher than that in CKO mice (figure 2C). Cleaved caspase-3-positive cells, indicating apoptosis, were occasionally observed in atypical foci in CKO mice, but not in those in DCKO mice (see supplementary figure 2, available online only). We next examined p53 expression in aberrant lesions. The Expression of p53 was observed in some atypical cells of the foci and a few of non-polarised parietal cells in CKO mice (figure 2D), suggesting that wild-type p53 was partially activated in the lesions. However, p53 expression was never seen in those in DCKO mice (figure 2D). Although the expression of p21/Waf1, one of the p53 downstream targets, was weakly positive in some cells of atypical foci in CKO mice, it was never seen in those in DCKO mice (see supplementary figure 2, available online only). These data indicate that E-cadherin loss may be sufficient to induce abnormal differentiation and that p53 loss may enhance the process against cell growth arrest and apoptosis.

We further performed immunohistochemical analyses to characterise the proton pump-negative atypical foci in DCKO mice. They were composed of cytokeratin-positive epithelial cells and surrounded by myofibroblasts expressing α-smooth muscle actin (see supplementary figure 3, available online only). Double cortin-like kinase 1 and trefoil factor family 2 (Tff2) have been reported as a gastric stem cell marker29 and a marker of progenitor cells for mucus neck, zymogenic and parietal cell lineage,30 respectively. However, the atypical foci were negative for both the gastric stem/progenitor markers (see supplementary figure 3, available online only).

Intramucosal cancer formation

Intramucosal carcinomas were observed in two of six (33%) of the DCKO mice at 6 months and in seven of eight (88%) at 9 months (figure 3A), whereas no tumours were found in the stomachs of control mice, that is, Atp4b-Cre;Cdh1loxP/loxP;Trp53loxP/loxP littermate mice, Atp4b-Cre+;Cdh1loxP/loxP;Trp53loxP/+ mice or Atp4b-Cre+;Cdh1loxP/+;Trp53loxP/loxP mice at 12 months. Unlike proton pump-negative atypical foci, intramucosal carcinomas exhibited architectural distortion and an accumulation of malignant cells with nuclear atypia and a high nucleus/cytoplasm ratio (figure 3B). They were mainly composed of poorly differentiated malignant cells, not expressing E-cadherin or the proton pump, or not eosinophilic, and were partially composed of signet ring cells containing AB- and/or PAS-positive mucus. Most, if not all, cancer cells were Ki67 positive (figure 3B) and frequently formed depressed, non-polypoid lesions, which were very similar to superficial depressed type early gastric cancers in humans.3

Figure 3

Characterisation of diffuse-type gastric cancer in double conditional knockout (DCKO) mice. (A) Proportions of the depth of cancers in 6-, 9- and 10–12-month-old mice. –, No cancer; m, mucosa; sm, submucosa; mp/ss, muscularis propria and subserosa; se/si, serosa exposed and serosa infiltrating. (B) Intramucosal cancers. Haematoxylin and eosin (HE), Alcian blue, periodic acid Schiff (AB-PAS) and immunohistochemical staining of serial sections of a 9-month-old DCKO mouse stomach with antibodies against E-cadherin, the proton pump and Ki67. The right images are magnifications of boxed regions in the left ones, respectively. Scale bars, 100 µm (left) and 20 µm (right).

To characterise intramucosal gastric cancers, we carried out immunohistochemical analyses on the same markers described for the characterisation of proton pump-negative atypical foci. As shown in supplementary figure 3 (available online only), the staining patterns of the markers were very similar between intramucosal gastric cancers and proton pump-negative atypical foci. Moreover, intramucosal gastric cancer cells were often located next to proton pump-negative atypical foci (see supplementary figure 4, available online only), suggesting that proton pump-negative cells in the atypical foci may transform into malignant cells.

Invasive cancer formation

Invasive cancers were found in two of eight (25%) of the DCKO mice at 9 months and in 33 of 48 (69%) of them at 12 months (figure 3A). The cancer-related mortality of the DCKO mice was 50% at 12 months, whereas control mice never died within 12 months (p<0.001, log rank test) (figure 4A). The causes of death were usually digestive haemorrhage and/or pyloric stenosis. Macroscopically, irregular but not polypoid epithelia with cancer-associated ulcers and a white thickened wall indicating a dense fibrous reaction were observed in the corpus region (figure 4B). Histologically, the cancers were composed of predominantly poorly differentiated carcinoma cells and partially signet ring cells (figure 4C). Infiltrating malignant cells, which were E-cadherin negative and β-catenin negative, were distributed from the intramucosal layer to the submucosal and subserosal layers (figure 4C). On BrdU uptake analysis, the cancer cells exhibited low mitogenic activity in the mucosal layer but high activity in the muscle and subserosal layers (figure 4D). These pathological findings were very similar to those of scirrhous-type gastric cancer in humans.

Figure 4

Invasive cancers. (A) Kaplan–Meier survival curves for Atp4b-Cre+;Cdh1loxP/loxP;Trp53loxP/+, Atp4b-Cre+;Cdh1loxP/+;Trp53loxP/loxP, Atp4b-Cre;Cdh1loxP/loxP;Trp53loxP/loxP and double conditional knockout (DCKO) mice. (B) Macroscopic images of the glandular stomachs of Atp4b-Cre;Cdh1loxP/loxP;Trp53loxP/loxP (Cre, left) and Atp4b-Cre+;Cdh1loxP/loxP;Trp53loxP/loxP (Cre+, right) mice at 12 months of age. A black arrow indicates a cancer-associated ulcer and a white arrow indicates a white thickened wall. Scale bars, 10 mm. (C) Haematoxylin and eosin (HE) and immunohistochemical staining of serial sections of a 12-month-old DCKO mouse stomach with antibodies against E-cadherin and β-catenin. The right images are magnifications from boxed regions in the left ones, respectively. Scale bars, 200 µm (left) and 50 µm (right). (D) Immunohistochemical staining of serial sections of the same mice as in section (C) with antibodies against BrdU, followed by methyl green counterstaining. Almost all the cancer cells in the muscle layer (middle panel) and in the subserosal layer (lower panel) are BrdU positive. Scale bars, 200 µm (left) and 50 µm (right).

Metastasis and tumourigenicity in immunodeficient mice

Lymph node metastases of cancers, particularly in the hepatic hilar region, were observed in two of seven (28.6%) and 19 of 48 (39.6%) cancer-positive cases at 9 and 12 months, respectively (figure 5), while haematogenous metastases (lung, liver or spleen) were not observed, suggesting early onset and a high frequency of lymphatic metastases of DGC in DCKO mice. These data are consistent with the high frequency of lymphatic metastases in human DGC, and suggest important roles of the E-cadherin deficiency in metastasis.9 When we inoculated aliquots of invasive cancers into nude mice, tumours grew rapidly and exhibited the original histological pattern of DGC (figure 6). These data indicate that DGC in DCKO mice were highly malignant.

Figure 5

Lymph node metastasis of diffuse-type gastric cancer. Haematoxylin and eosin (HE) staining of a section of a lymph node from a 12-month-old double conditional knockout mouse. Poorly differentiated gastric cancer cells can be observed in marginal sinus areas (surrounded by dotted lines) of the hepatic hilar lymph node. Scale bars, 200 µm (left) and 50 µm (right).

Figure 6

Transplanted tumour of diffuse-type gastric cancer. Upper panel, tumour formation in a nude mouse injected with invasive gastric cancer cells. An arrow indicates a transplanted tumour. A scale bar, 10 mm. Lower panels, haematoxylin and eosin (HE) and immunohistochemical staining of sections of the primary tumour and the transplanted tumour with antibodies against E-cadherin and Ki67. Scale bars, 20 µm.

DNA analyses of DGC

To detect the loss of the conditional Cdh1 and Trp53 alleles in cancer cells, we performed PCR analyses. We detected bi-allelic deletion of both Cdh1 and Trp53 in DGC located in the submucosal region, lymph node metastases and transplanted tumours, while the loss of the conditional Cdh1 or Trp53 alleles was not seen in other normal tissues (figure 7). These data were consistent with the findings that the expression of E-cadherin or p53 was not observed in intramucosal and invasive DGC (figures 3B and 4C, and supplementary figure 5, available online only). It is, therefore, conceivable that DGC induced in DCKO mice were originated by the loss of E-cadherin and p53 in the parietal cell lineage.

Figure 7

DNA analysis to detect the truncated Cdh1 and Trp53 alleles. All DNA were extracted from tissues of double conditional knockout mice and amplified using the primers shown in supplementary table 1 (available online only). Atp4b-Cre was used as an internal control. Lane 1, a stomach tissue including parietal cells; lanes 2–4, submucosal, muscle and subserosal layers into which cancers had infiltrated, respectively; lanes 5–7, three independent lymph node metastases; lanes 8–10, three independent transplanted tumours; lanes 11–13, normal liver, spleen and tail, respectively. DGC, diffuse-type gastric cancer.

Gene expression analyses of tumour tissues

To analyse the molecular characteristics of tumour tissues obtained from DCKO mice, we compared gene expression profiles of murine (GEO accession no GSE25302) and human gastric cancers. Out of 122 cDNA listed in ‘intestinal type versus diffuse type signature genes’,28 79 genes that were designed on Mouse GE 4x44K v2 Microarray and were marked as ‘present’ by feature extraction software were selected for further analysis. The hierarchical clustering analysis using these genes (figure 8A) showed that the left cluster was composed of 61 IGC and two DGC, whereas the right cluster included seven IGC and 11 DGC (p<0.001, Fisher's exact test). Notably, three DGC tissues from the mice were all classified into the right cluster. These results suggest that the molecular profiles of E-cadherin/p53-deficient DGC are much more similar to human DGC than human IGC.

Figure 8

Gene expression analyses of tumour tissues. (A) Heat map of gene expression patterns in human intestinal-type gastric cancer (IGC), human diffuse-type gastric cancer (DGC) and DGC tissues in double conditional knockout mice using 79 homologous genes. Each column represents samples (colour labels on top: red, human IGC; blue, human DGC; sky-blue, murine DGC), and each row represents genes. Pearson's correlation coefficient was used to calculate distance matrix among genes or samples, respectively. The average linkage method was used for agglomeration. Gene expression levels were represented by log2 ratios to normal gastric tissues from human or control mice. Red and green squares indicate overexpression and underexpression, respectively. Grey squares represent missing values. (B) The expression of three mesenchymal markers, Vimentin (Vim), Fibronectin (Fn1) and N-cadherin (Cdh2), and five epithelial–mesenchymal transition regulators, Twist1, Snail (Snai1), Slug (Snai2), Zeb1 and Zeb2, and two parietal cell-specific genes, Atp4a and Atp4b, was analysed. Lanes 1–7, normal gastric mucosae; lanes 8–12, primary gastric cancers; lanes 13–15, transplanted gastric cancers.

Gene expression profiles also revealed that some mesenchymal markers and epithelial–mesenchymal transition (EMT) regulators were upregulated in murine DGC. As the loss of E-cadherin expression is involved in EMT31 and EMT plays an important role in the development of human gastric cancers, in particular DGC,32 33 we validated the expression of well-documented mesenchymal markers and EMT regulators in primary and transplanted DGC. Vimentin, Fibronectin, Twist1 and Snail were highly expressed in all DGC compared with in normal gastric mucosae (figure 8B). The expression of Cdh2, Zeb1 and Zeb2 was also high in all DGC, but their expression was high in some normal mucosae (figure 8B), which is consistent with the previous results.32 33 Immunohistochemical analyses also demonstrated that N-cadherin, encoded by Cdh2, and Twist1 were expressed in highly invasive portions of primary DGC (see supplementary figure 6, available online only). These data indicate that DGC in our model were biochemically very similar to those in humans, and that EMT might be one of the important mechanisms underlying DGC formation.

Discussion

There have been several lines of genetically engineered mice that are predisposed to the development of gastric cancer. IGC have been induced in pS2 trefoil protein-deficient mice,34 mice lacking the SHP2-binding site on the IL-6 family receptor gp130,35 gastrin-overexpressing transgenic mice36 and mice expressing Wnt1, cyclooxygenase-2 (COX-2) and prostaglandin E synthase-1.37 Neuroendocrine-type gastric cancers develop in an Atp4b-SV40 T antigen transgenic mouse line.24 As to a mouse DGC model, although gastric signet ring cell carcinomas have been observed in Cdh1+/− mice after N-methyl N-nitrosourea treatment,38 our DCKO mouse line represents the first genetically engineered mouse model of DGC. It should be noted that DGC developed in all the 48 mice within 12 months (figure 3A). The cancers in DCKO mice were mainly composed of poorly differentiated and partly of signet ring carcinoma cells, which is similar to that of human hereditary DGC.3 39 Moreover, the DGC arising in our model invaded into submucosal regions, metastasised to lymph nodes and had tumourigenic activity in nude mice, indicating a highly malignant phenotype. Together with gene expression patterns, we concluded that DCKO murine DGC were very similar to human DGC.

In CKO and DCKO mice, non-polarised parietal cells and proton pump-negative atypical foci were observed after 3 and 6 months, respectively.13 Clusters of Ki67-negative signet ring-like cells, termed type 3, were also observed after 1 year in CKO mice,13 but not in DCKO mice. However, no cancers were found even after 2 years in CKO mice.13 E-cadherin loss in the parietal cell lineage can thus be associated with the loss of cell polarity and possibly abnormal differentiation. A study of early DGC including hereditary DGC revealed that E-cadherin downregulation does not result in increased proliferation.39 Similarly, conditional targeting of Cdh1 in the mammary gland14 or skin15 does not induce tumours. These data suggest that E-cadherin loss is, in itself, not sufficient for tumour formation.

The loss of E-cadherin expression promotes invasion and metastasis of cancer cells,9 possibly through EMT.31 It was recently reported that the induction of EMT in immortalised human mammary epithelial cells resulted in the acquisition of mesenchymal traits and in the expression of stem cell markers, indicating a direct link between EMT and the gain of stem cell-like properties.40 Several EMT-related genes were highly expressed in DGC of our mouse model, suggesting that the DCKO parietal cell lineage may easily undergo abnormal differentiation through EMT.

It has been reported that E-cadherin suppresses cellular transformation by inhibiting β-catenin signalling in an adhesion-independent manner,41 suggesting that the E-cadherin defect may upregulate the β-catenin level and accordingly promote tumour formation. However, β-catenin was not detected in DGC formed in DCKO mice, even at the invasive front (figure 4C). β-Catenin expression has not been detected in human sporadic DGC,42 in early DGC in a hereditary DGC kindred,39 or in signet ring cell carcinomas of N-methyl N-nitrosourea-treated Cdh1+/− mice.38 It is, therefore, likely that β-catenin signalling is not involved in DGC formation in E-cadherin-deficient mice.

As for the roles of p53, the appearance of proton pump-negative atypical foci in DCKO mice was chronologically similar to that of type 2 cell clusters in CKO mice, while the number and Ki67 labelling index of the lesions were significantly larger in DCKO mice than those in CKO mice. Moreover, p53, p21 and cleaved caspase-3 were occasionally found in type 2 cells in CKO mice, but never in the atypical foci or DGC in DCKO mice (figure 2D and supplementary figures 2 and 5, available online only), suggesting that p53 loss promotes cell growth and anti-apoptotic events.

There have been many types of Trp53-deficient mice, but gastric cancer does not seem to develop in these mice,22 indicating that p53 inactivation is not associated with the initiation of gastric cancer. TP53 mutations have been more frequent in advanced human DGC compared with early DGC.17 As DGC was not found in Atp4b-Cre+;Cdh1loxP/+;Trp53loxP/loxP mice, the loss of p53 function may be a late event in DGC formation. Synergistic tumour formation caused by inactivated Trp53 and Cdh1,23 and inactivated Trp53 and Brca225 in the mammary gland, has been reported. These results indicate that the synergistic effects of inactivated Cdh1 and Trp53 are necessary for DGC formation in our mouse model.

On the basis of our observations, we propose a model for the development of DGC from the parietal cell lineage in DCKO mice, that is, non-polarised parietal cells, proton pump-negative atypical foci and intramucosal and invasive cancers (figure 9). These lesions seem to occur stepwise for the following reasons. First, they were sequentially found in all DCKO mice. Second, proton pump-negative atypical foci and intramucosal cancers share some common immunohistochemical characteristics. Finally, we observed transitional images that proton pump-negative atypical foci might advance to intramucosal cancer (see supplementary figure 4, available online only). However, proton pump-negative atypical foci and DGC may not originate from parietal cells but from preparietal cells. E-cadherin/p53-deficient preparietal cells are supposed to become proton pump-negative atypical foci easily rather than parietal cells, because preparietal cells are at an earlier stage of differentiation than parietal cells.

Figure 9

The model of gastric carcinogenesis in double conditional knockout mice. At first, E-cadherin/p53-deficient parietal cells become non-polarised and pushed out to the stromata. E-cadherin/p53-deficient preparietal cells may also undergo loss of cell polarity. Second, abnormal differentiation of the non-polarised parietal cell lineage, particularly preparietal cells, may induce proton pump-negative atypical foci. Then, inactivation of the Cdh1 and Trp53 genes further leads to the initiation and progression of cancers. AB-PAS, Alcian blue, periodic acid Schiff.

The origin of human DGC remains unknown. Barker et al43 recently demonstrated that Wnt-driven transformation of adult Lgr5-positive stem cells efficiently induced adenoma formation in the pylorus. On the other hand, E-cadherin and p53 were specifically inactivated in the parietal cell lineage in our DCKO mice, because we used the Atp4b gene promoter for Cre. Analysis of multiple tissues harvested from Atp4b-Cre+;Rosa26R bitransgenic mice confirmed that Escherichia coli β-galactosidase expression was limited to the parietal cell lineage.24 In our study, immunohistochemical analyses demonstrated that the precancerous and cancerous lesions were cytokeratin positive but negative for the markers of gastric stem cells, progenitor cells for mucus neck, zymogenic and parietal cell lineage. It is thus likely that the origin of the atypical foci and DGC may be preparietal cells in DCKO mice, but further studies are necessary to confirm it. Even though human DGC may not necessarily originate from the parietal cell lineage, our data indicate that the DCKO mouse line represents a good model of human DGC.

Our mouse model may be very useful in several ways. First, it would be important to clarify the mechanism underlying DGC formation thoroughly. Second, it is an excellent model for preclinical intervention studies and for testing preventive measures to DGC, because DGC developed in our DCKO mice at 100% penetrance within a year. Finally, gene expression microarray analyses of mouse DGC and corresponding normal stomach tissues would facilitate the identification of important genes associated with DGC formation, including candidates for target therapy for DGC.

Acknowledgments

Atp4b-Cre transgenic and Cdh1loxP/loxP mice were generously provided by Professors Jeffrey I Gordon and Rolf Kemler, respectively. The authors acknowledge S Shimizu and K Uchida for invaluable assistance in pathological procedures.

References

View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Funding This work was supported by a grant-in-aid for scientific research on priority areas cancer 17015013 from the Ministry of Education, Culture, Sports, Science and Technology, Japan, by the JSPS A3 Foresight Program and by a research grant from the Princess Takamatsu Cancer Research Fund.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.