Immunohistochemical analysis of pRb2/p130, VEGF, EZH2, p53, p16INK4A, p27KIP1, p21WAF1, Ki-67 expression patterns in gastric cancer

Authors

  • Eliseo Mattioli,

    1. Sbarro Institute for Cancer Research and Molecular Medicine, Center for Biotechnology, College of Science and Technology, Temple University, Philadelphia, Pennsylvania
    2. Department of Pathology, University of Bari, Ospedale Policlinico Consorziale, Italy
    Search for more papers by this author
  • Paraskevi Vogiatzi,

    1. Sbarro Institute for Cancer Research and Molecular Medicine, Center for Biotechnology, College of Science and Technology, Temple University, Philadelphia, Pennsylvania
    2. Department of Molecular Biology, Medical Genetics Unit, University of Siena, Siena, Italy
    Search for more papers by this author
  • Ang Sun,

    1. Sbarro Institute for Cancer Research and Molecular Medicine, Center for Biotechnology, College of Science and Technology, Temple University, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Giovanni Abbadessa,

    1. Sbarro Institute for Cancer Research and Molecular Medicine, Center for Biotechnology, College of Science and Technology, Temple University, Philadelphia, Pennsylvania
    2. Department of Oncology and Hematology, Istituto Clinico Humanitas, Rozzano (MI), Italy
    Search for more papers by this author
  • Giulia Angeloni,

    1. Department of Cardiovascular Medicine, Catholic University Medical School, Campobasso, Italy
    Search for more papers by this author
  • Domenico D'Ugo,

    1. Department of Surgery, Catholic University Medical School, Campobasso, Italy
    Search for more papers by this author
  • Daniela Trani,

    1. Sbarro Institute for Cancer Research and Molecular Medicine, Center for Biotechnology, College of Science and Technology, Temple University, Philadelphia, Pennsylvania
    2. Department of Scienze Cardio-Toraciche e Respiratorie, Osp. A. Monaldi, Seconda Universita' degli Studi di Napoli, Italy
    Search for more papers by this author
  • John P. Gaughan,

    1. Biostatistics Consulting Center, Temple University, School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Fabio Maria Vecchio,

    1. Divisione di Anatomia Patologica, Università Cattolica del Sacro Cuore, Roma, Italy
    Search for more papers by this author
  • Gabriele Cevenini,

    1. Department of Chirurgia e Bioingegneria, University of Siena, Siena, Italy
    Search for more papers by this author
  • Roberto Persiani,

    1. Department of Surgery, Catholic University Medical School, Campobasso, Italy
    Search for more papers by this author
  • Antonio Giordano,

    1. Sbarro Institute for Cancer Research and Molecular Medicine, Center for Biotechnology, College of Science and Technology, Temple University, Philadelphia, Pennsylvania
    2. Department of Human Pathology and Oncology, University of Siena, Siena, Italy
    Search for more papers by this author
  • Pier Paolo Claudio

    Corresponding author
    1. Sbarro Institute for Cancer Research and Molecular Medicine, Center for Biotechnology, College of Science and Technology, Temple University, Philadelphia, Pennsylvania
    • College of Science and Technology, Center for Biotechnology, Bio Life Sciences Building, Suite 333, 1900 North 12th Street, Philadelphia, PA 19122-6099.
    Search for more papers by this author

  • Eliseo Mattioli and Paraskevi Vogiatzi contributed equally to this manuscript.

Abstract

Although the considerable progress against gastric cancer, it remains a complex lethal disease defined by peculiar histological and molecular features. The purpose of the present study was to investigate pRb2/p130, VEGF, EZH2, p53, p16INK4A, p27KIP1, p21WAF1, Ki-67 expressions, and analyze their possible correlations with clinicopathological factors. The expression patterns were examined by immunohistochemistry in 47 patients, 27 evaluated of intestinal-type, and 20 of diffuse-type, with a mean follow up of 56 months and by Western blot in AGS, N87, KATO-III, and YCC-2, -3, -16 gastric cell lines. Overall, stomach cancer showed EZH2 correlated with high levels of p53, Ki-67, and cytoplasmic pRb2/p130 (P < 0.05, and P < 0.01, respectively). Increased expression of EZH2 was found in the intestinal-type and correlated with the risk of distant metastasis (P < 0.05 and P < 0.01, respectively), demonstrating that this protein may have a prognostic value in this type of cancer. Interestingly, a strong inverse correlation was observed between p27KIP1 expression levels and the risk of advanced disease and metastasis (P < 0.05), and a positive correlation between the expression levels of p21WAF1 and low-grade (G1) gastric tumors (P < 0.05), confirming the traditionally accepted role for these tumor-suppressor genes in gastric cancer. Finally, a direct correlation was found between the expression levels of nuclear pRb2/p130 and low-grade (G1) gastric tumors that was statistically significant (P < 0.05). Altogether, these data may help shed some additional light on the pathogenetic mechanisms related to the two main gastric cancer histotypes and their invasive potentials. J. Cell. Physiol. 210: 183–191, 2007. © 2006 Wiley-Liss, Inc.

Gastric cancer is a common malignancy and still remains a major public health issue. The highest incidence rates are reported in Korea, Japan, and Eastern Asia. A high incidence is also observed in Eastern Europe and parts of Latin America, while in Western Europe and USA the disease is in constant decline. Despite the advance in therapeutic options, less than 20% of patients survive 5 years after diagnosis (Smith et al., 2006). Gastric cancer is usually sporadic, but familial aggregation of the disease may be seen in approximately 10% of the cases (Oliveira et al., 2006). Over 95% of gastric malignancies are adenocarcinomas. According to the widely used Lauren's 1965 classification, there are two types of gastric cancer: the intestinal-type of adenocarcinoma, which follows the pathologic sequential steps of atrophic gastritis, intestinal metaplasia, dysplasia, carcinoma; and the less common diffuse-type, with worse prognosis and correlated with chronic gastritis (Tahara, 2004; Correa and Schneider, 2005; Smith et al., 2006). Distal gastric cancer (non-cardial) is often of the intestinal-type and predominates in developing countries, among blacks, and in lower socio-economic groups, whereas proximal tumors (many of which show diffuse-type histology) are more common in developed countries, among whites, and in higher socio-economic classes. While the incidence of the former is declining, that of the latter is not; in particular, the signet-ring subtype has been increasing. The main risk factors for distal gastric cancer include Helicobacter pylori infection and dietary factors, whereas gastroesophageal reflux disease and obesity play important roles in the development of proximal stomach cancer (Crew and Neugut, 2006).

At the molecular level gastric tumors arise from multiple genetic and epigenetic alterations that involve oncogenes, tumor-suppressor genes, cell-cycle regulators, cell adhesion molecules, DNA repair genes and from genetic instability, and its pathogenesis is still unknown (Tahara, 2004).

In this study we evaluated immunohistochemically the expression of various proteins, crucial to cell-cycle control (p53, p16INK4A, p27KIP1, p21WAF1, and pRb2/p130), tissue development and differentiation (EZH2), and angiogenesis (VEGF), and Ki-67 in gastric cancer patients, in order to further elucidate the molecular mechanisms involved in gastric transformation and its malignant progression.

The p53, p16INK4A, p27KIP1, p21WAF1, RB2/p130 tumor-suppressor genes act by modulating cell proliferation via control of G1 arrest checkpoint of cell-cycle (Ford et al., 2004).

Abnormalities of the p53 gene have been identified in many malignancies, including gastric carcinomas (Martin et al., 1992).

The production of p53 is increased in response to cellular insults or DNA damage, and p53 then induces cell-cycle arrest at the G1/S-junction (Fenoglio-Preiser et al., 2003). The p21WAF1 and p27KIP1 genes produce proteins that are activated by p53 and induce cell-cycle arrest by inhibition of kinase activity of cyclin/cyclin-dependent kinase complexes regulating cell-cycle progression (Michieli et al., 1994; Wiksten et al., 2002). Several authors have reported that overexpression of p21WAF1 and p27KIP1 in gastric cancer results in improved outcome, although a few studies reported opposite results (Feakins et al., 2000; Kaye et al., 2000; Migaldi et al., 2001; Wiksten et al., 2002). More recently, Liu et al. (2001) suggested that combined examination of p21WAF1, p27KIP1, and p53 expression allows precise estimation of prognosis in patients with gastric cancer.

The p16INK4A was originally identified as a tumor-suppressor gene because frequently mutated in melanomas, has been shown to be involved in a broad range of tumors. Alterations of p16INK4A are reported in various human malignancies, and are exceeded in frequency only by the p53 tumor-suppressor gene. In particular, p16INK4A inactivation has been identified as a possible event in malignant transformation of gastric mucosa (Myung et al., 2000). Recently, hypermethylation of the p16INK4A promoter was found more frequently in microsatellite instable gastric carcinomas (Kim et al., 2003).

VEGF is a homodimeric glycoprotein that functions as a mitogenic factor for endothelial cells. Secreted by many different cell types, its expression is upregulated by hypoxia. Many recent experimental data suggest a major involvement of this protein in tumor angiogenesis, a complex process of primary importance for neoplastic progression and metastatic spread. VEGF expression has been found recently highly expressed in intestinal-type gastric cancer with respect to that of diffuse-type gastric cancer (P = 0.017) (Chen et al., 2004). Recently, VEGF expression has been shown to be downregulated, at the transcriptional and translational levels, by Rb2/p130 and p53 expression, both in vitro and in vivo (Riccioni et al., 1998; Claudio et al., 2001).

pRb2/p130 is a member of the Retinoblastoma family of proteins which also includes pRB/p105 and p107. These nuclear proteins, also known as “pocket proteins” for their unique structure, negatively regulate the G1-S cell-cycle transition by interacting with Cyclin–CDK complexes, modulating the activity of several transcription factors such as the E2Fs. Genetic and functional inactivation of pRb2/p130 allows cells to bypass the G0/G1 checkpoint, enabling them to undergo mitosis (Claudio et al., 2002). pRb2/p130 acts by repressing transcription via binding to E2F4 and E2F5 members of E2F transcription factors (Gaubatz et al., 2000), which have binding sites for promoters of genes important for progression of cells from G1 to S phase. Recent investigations have suggested the importance of chromatin remodeling for the suppression of cellular proliferation mediated by the pocket proteins (Ferreira et al., 1998; Kuo and Allis, 1998; Stiegler et al., 1998; Ito and Adcock, 2002). In a recent study, we have also demonstrated that EZH2 interacts with pRb2/p130 both in vitro and in vivo and that the functional role of this protein–protein interaction is to interfere with the repressive activity of pRb2/p130 on cell-cycle-promoting genes, such as Cyclin A, underlying a new mechanism of inactivation of pRb2/p130 function depending on EZH2 expression (Tonini et al., 2004).

EZH2 is a newly identified nuclear protein that shows sequence homology to the “Enhancer of Zeste” protein of Drosophila, and is therefore considered a member of the Polycomb group of proteins (Varambally et al., 2002). Also EZH2 is thought to be involved in gene expression control by taking part in chromatin remodeling. EZH2 has been found widely expressed in developing embryos, and its expression decays upon tissue maturation and differentiation. In human pathology, EZH2 has been shown overexpressed in the most aggressive forms of prostate cancer, exhibiting correlation with poor clinical outcome, and thus acting as a novel marker of aggressiveness and unfavorable prognosis in this type of cancer (Varambally et al., 2002).

The Ki-67 is a commercially available monoclonal antibody that reacts with a nuclear antigen expressed in proliferating but not in quiescent cells. Expression of this antigen occurs preferentially during late G1, S, G2, and M phases of the cell cycle, while in cells in G0 phase the antigen cannot be detected. Consequently, the antibody is used in tumor pathology to detect proliferating cells in neoplastic diseases. Ki-67 labeling index is calculated immunohistochemically evaluating the cell growth-related antigen Ki-67, using the monoclonal antibody MIB-1. Its positivity has been evaluated also in gastric cancer (Schipper et al., 1998; Igarashi et al., 1999).

In this study, we are describing for the first time the correlation between the expression levels of p53, p21WAF1, p27KIP1, p16INK4A, pRb2/p130, VEGF, EZH2, and Ki-67 in primary gastric cancers and their relative expression in gastric cancer cell lines.

MATERIALS AND METHODS

Patients and samples

Forty-seven patients (aged 48–86 years) underwent surgery at the Department of Surgery, Catholic University Medical School, Campobasso, Italy for malignant gastric tumors (34 males and 13 females) were enrolled in this study. Twenty-seven samples were intestinal-type and 20 were diffuse-type. Twenty-four patients were staged as I–II and 23 as III–IV according to the AJCC classification. Overall mean age of patients was 63 years (range 48–86). Mean age in the intestinal-type group was 68 years (range 48–83). Mean age in the diffuse-type group was 63 years (range 49–78).

Forty-one patients were evaluated with a mean follow-up of 56 months and a median of 37 months (range 8–146 months) because 6 patients were either lost at the follow-up or were deceased for causes other than gastric cancer. Follow-up was updated to June 2006. Patients signed an informed consent for the study that was reviewed by the Institutional Review Board.

Sample processing and histological diagnosis

All bioptic samples were formalin-fixed (for at least 24 h) and paraffin-embedded. Several 4-µm thick sections were cut from each specimen, mounted on glass, and dried at 37°C. Two sections of each sample were stained with Hematoxylin and Eosin and evaluated by a pathologist to confirm the diagnosis.

Immunohistochemistry

Several sections of each sample, cut from the same blocks, were used to perform immuhistochemical reactions, according to the following protocol. All sections were dewaxed in xylene and rehydrated through a sequence of decreasing concentration of alcoholic solutions; endogenous peroxydase activity was quenched by 0.5% hydrogen peroxide incubation for 30 min at room temperature. Sections were microwave-pretreated in 10 mM citrate buffer (pH 6.0) for antigen retrieval (three cycles of 5 min each at 650 W). After blocking with PBS-diluted normal serum, sections were incubated with primary antibodies according to the following conditions:

  • anti p53 (Dako, Hamburg, Germany, clone DO7, mouse monoclonal antibody): incubated overnight at 4°C + 2 h at room temperature; dilution 1:50;

  • anti p21WAF1 (BD Biosciences Pharmingen, Franklin Lake, NJ, clone AB11, mouse monoclonal antibody): incubated overnight at 4°C + 2 h at room temperature; dilution 1:100;

  • anti p27KIP1 (Santa Cruz Biotechnologies, Inc., Santa Cruz, CA, mouse polyclonal antibody): incubated overnight at 4°C + 2 h at room temperature; dilution 1:10;

  • anti p16INK4A (Santa Cruz Biotechnologies, Inc., rabbit polyclonal antibody): incubated 2 h at room temperature; dilution 1:200;

  • anti EZH2 (Upstate, Lake Placid, NY, rabbit polyclonal antibody) incubated overnight at 4°C + 2 h at room temperature; dilution 1:50;

  • anti VEGF (Upstate, clone AB-3, mouse monoclonal antibody): incubated overnight at 4°C + 2 h at room temperature; dilution 1:50;

  • anti Rb2/p130 (Neomarkers, Inc., Union City, CA, mouse monoclonal antibody): incubated overnight at 4°C + 2 h at room temperature; prediluted (ready to use);

  • anti Ki-67 (Dako, clone MIB-1, mouse monoclonal antibody): incubated 2 h at room temperature; dilution 1:100.

After two washes in PBS, sections were incubated with goat anti-mouse or anti-rabbit biotinylated secondary antibody for 30 min at room temperature. After two washes in PBS, sections were incubated with avidin–biotin–peroxydase complex (Vectastain Elite ABC kit, Vector Laboratories, Burlingame, CA) and then washed two more times in PBS. The immunoreactivity was revealed using diaminobenzidine (DAB) as the final chromogen. Finally, sections were counterstained with Meyer's Hematoxylin, dehydrated through a sequence of increasing concentration alcoholic solutions, cleared in xylene, and mounted with epoxydic medium.

During each immunohistochemical assay proof slides were coupled with negative control slides on which the primary antibody was omitted.

Slides evaluation

Slides were evaluated by two different pathologists (EM and FMV), who assessed both percentage of positive neoplastic cells and staining intensity (rated in a three-step scale, low-, medium-, and high-intensity). Discrepancies in the evaluation were resolved by conjoined reobservation of the cases through a multi-headed microscope.

Each case was scored according to the formula:

equation image

Where i = staining intensity (ranging 1–3) and Pi = percentage of positive cells.

Statistical analysis

Correlation analysis (Pearson and Spearman correlation coefficients) was used to evaluate statistical relationships among the candidate proteins and with clinicopathological parameters. Univariate and multivariate Cox proportional hazards analysis was used to relate protein levels to survival. Variables found to be significant by univariate analysis were entered in a multivariate model to evaluate their independent association with survival.

Cell culture

AGS, NCI-N87, KATO-III, and NL-20 cells were obtained by the American Type Culture Collection. The AGS (ATCC CRL-1739) cell line is from fragments of a primary gastric tumor, moderately differentiated, resected from a Caucasian, 54 years old female patient who had received no prior therapy and it is tumorigenic in athymic BALB/c mice. The NCI-N87 [N87] (ATCC CRL-5822) line is derived from a liver metastasis of a well-differentiated carcinoma (intestinal type) of a Japanese male prior to cytotoxic therapy, which is tumorigenic in athymic nude mice. The non-tumorigenic KATO-III (ATCC HTB-103) cell line is derived from pleural effusion, lymph nodes, and Douglas cul-de-sac from a signet ring carcinoma, poorly differentiated or diffuse type in a Japanese male of 55 years. YCC-2, YCC-3, YCC-16 were a kind gift of Dr. Sun Young Rha, Yonsei Cancer Metastasis Research Center (CMRC, Seoul, Korea). YCC-2 and -3 were derived from the ascite fluid, while the YCC-16 cells were from peripheral blood, of three different gastric cancer patients. NL-20 cells (normal lung epithelium) (CRL-2503) were utilized as normal control in Western blot analysis because they express wild-type p53. AGS cells were grown in Ham's F12 medium, the NCI-N87 and KATO-III cells were grown in RPMI 1640, the YCC lines in DMEM at 37°C in a water-saturated atmosphere of 95% air and 5% CO2. All these mediums were supplemented with 2 mM L-glutamine and 10% fetal bovine serum. The NL-20 cells were grown in Ham's F12 medium with 1.5 g/L sodium bicarbonate, 2.7 g/L glucose, 2.0 mM L-glutamine, 0.1 mM non-essential amino acids, 0.005 mg/ml insulin, 10 ng/ml epidermal growth factor, 0.001 mg/ml transferrin, 500 ng/ml hydrocortisone, and 4% fetal bovine serum at 37°C in a water-saturated atmosphere of 95% air and 5% CO2.

Western Blot analysis

Western blot analysis was performed on 50 µg of total protein lysate extracted from six gastric cell lines (AGS, NCI-N87, KATO-III, YCC-2, YCC-3, YCC-16) as described previously (Kim et al., 2005) using antibodies against Rb2/p130 diluted 1:500 (BD Biosciences Pharmingen), VEGF diluted 1:200 (Santa Cruz, CA), EZH2 diluted 1:250 (Abcam, Inc., Cambridge, MA), p16INK4A diluted 1: 200 (Santa Cruz), p21WAF1 diluted 1: 500 (Santa Cruz), p27KIP1 diluted 1: 200 (Santa Cruz), p53 clone D-11 diluted 1: 200 (Santa Cruz), and HSP-72/74 diluted 1:5,000 (Calbiochem, San Diego, CA).

RESULTS

Immunohistochemical analysis

The expression levels of p16INK4A, p21WAF1, p27KIP1, pRb2/p130, VEGF, p53, EZH2, and Ki-67 were determined by immunohistochemistry in 27 cases of intestinal-type and in 20 cases of diffuse type gastric cancer (Fig. 1). The immunohistochemical assay revealed several significant differences in the expression of the investigated markers between our patients and are summarized in Table 1.

Figure 1.

Representative panel of immunohistochemical analysis of EZH2, VEGF, pRb2/p130, p27KIP1, p16INK4A, p21WAF1, p53, and Ki-67 in gastric cancer. A: High expression levels of EZH2 in a gastric carcinoma, (B) Case showing extremely high expression of VEGF in the cytoplasm, (C) Low expression levels of pRb2/p130 in the nuclei of a primary gastric cancer, (D) High expression levels of p27kip1, (E) High expression levels of p16INK4A, (F) Representative case expressing high levels of p21WAF1, (G) Case showing extremely high expression of p53, and (H) High levels of Ki-67 labeling index.

Table 1. Correlation between overall series of protein expressions in gastric cancer
  Correlations with Pearson rP-value
LowModerateHigh
  1. pRb2/p130 n: nuclear pRb2/p130.

  2. pRb2/p130 c: cytoplasmic pRb2/p130.

pRb2/p130 nVEGF0.296  <0.05
 EZH20.295  <0.05
 p210.380  <0.01
pRb2/p130 cVEGF 0.520 <0.01
 EZH2 0.515 <0.01
EZH2pRb2/p130 n0.295  <0.05
 pRb2/p130 c 0.515 <0.01
 p530.361  <0.05
 Ki-670.357  <0.05
VEGFpRb2/p130 n0.296  <0.05
 pRb2/p130 c 0.520 <0.01
p27p210.330  <0.05
p16Ki-670.435  <0.01
 p530.464  <0.01
p53EZH20.361  <0.05
 p160.464  <0.01
 Ki-67  0.990<0.01
Ki-67EZH20.357  <0.05
 p160.435  <0.01
 p53  0.990<0.01

The immunostaining for each protein was also determined as positive or negative by a cutoff value determined as follows: p53, p16INK4A, p27KIP1, and p21WAF1 and EZH2 staining was interpreted as positive when >10% of the tumor cells showed distinct nuclear staining and Ki67 when >25% showed distinct nuclear staining as previously reported (Al-Moundhri et al., 2005). In general, the expression of the studied genes in our cases was as follows: p16INK4A (46.8%), p21WAF1 (53.2%), p27KIP1 (72.3%), p53 (55%), EZH2 (78.8%), and Ki-67 (38.3%).

Consistently with the results of a very recent publication (Matsukawa et al., 2006), the non-cancerous gastric mucosa showed faint or no EZH2 immunoreactivity restricted to the nuclei of glandular epithelial cells (data not shown). In the majority of the gastric cancers examined, high EZH2-specific nuclear immunostaining was found (78.8%). Positive statistically significant correlations were found between the expression levels of EZH2 and Ki-67, p53 and nuclear Rb2/p130 expression (P < 0.05). Stronger positive correlation (P < 0.01) was found between the expression levels of EZH2 and cytoplasmic pRb2/p130, even though a variable pRb2/p130 nuclear staining was also noted in all the samples analyzed.

Positive correlations were found between the expression levels of nuclear pRb2/p130 and p21WAF1 or VEGF (P < 0.01 and P < 0.05, respectively). Stronger positive correlations were found between the expression levels of VEGF and cytoplasmic pRb2/p130 that were statistically significant (P < 0.01).

It has been reported by Al-Moundhri et al. and others that the growth inhibitory activity of p21WAF1 and p27KIP1, may be modulated in more advanced colorectal and gastric tumors through inactivation (Cheng et al., 1999; Al-Moundhri et al., 2005). Consistently with these previously published data we found a positive correlation between p21WAF1 and p27KIP1 in our gastric tumor samples (P < 0.05).

Additionally, we found positive significant correlations between the expression levels of Ki-67 and EZH2 or p16INK4A (P < 0.05 and P < 0.01, respectively). Interestingly, we found also a very strong positive correlation (P < 0.01) between Ki-67 and p53 expression levels.

Correlation of the biological parameters according to histological gastric tumor types

The specimens we examined were from patients affected by either diffuse (42.55%) or intestinal gastric cancers (57.45%) according to the Lauren's criteria.

Analyzing the data according to tumor type (intestinal vs. diffuse), we found some interesting and statistically significant correlations that were not present when considering the global patients' population and that are summarized in Tables 2 and 3. Considering the intestinal-type, we found the following correlations (Table 2): EZH2 was still correlated with the cytoplasmic expression of pRb2/p130 (P < 0.05). Moreover, some other correlations that were found at the global analysis were also confirmed. Specifically, cytoplasmic pRb2/p130 expression was correlated with VEGF expression (P < 0.01) whereas the expression levels of p53 were highly correlated with that of Ki-67 (P < 0.01).

Table 2. Correlations of protein expressions according to the intestinal-type of gastric cancer
  Correlations with Pearson rP-value
LowModerateHigh 
  1. pRb2/p130 n: nuclear pRb2/p130.

  2. pRb2/p130 c: cytoplasmic pRb2/p130.

pRb2/p130 cVEGF0.499  <0.01
 EZH20.412  <0.05
VEGFpRb2/p130 c0.499  <0.01
EZH2pRb2/p130 c0.412  <0.05
p53Ki-67  0.986<0.01
Ki-67p53  0.986<0.01
Table 3. Correlations of protein expressions according to the diffuse-type of gastric cancer
  Correlations with Pearson rP-value
LowModerateHigh 
  1. pRb2/p130 n: nuclear pRb2/p130.

  2. pRb2/p130 c: cytoplasmic pRb2/p130.

pRb2/p130 np27 0.504 <0.05
pRb2/p130 cVEGF0.466  <0.05
 EZH2 0.644 <0.01
VEGFpRb2/p130 c0.466  <0.05
EZH2pRb2/p130 c 0.644 <0.01
 p53 0.583 <0.01
 Ki-67 0.584 <0.01
p27pRb2/p130 n 0.504 <0.05
 p21 0.583 <0.01
p16p53  0.749<0.01
 Ki-67  0.749<0.01
p53EZH2 0.583 <0.01
 p16  0.749<0.01
 Ki-67  1.000<0.01
Ki-67EZH2 0.584 <0.01
 p16  0.749<0.01
 p53  1.000<0.01

Considering instead the diffuse-type, we found the following correlations (Table 3): cytoplasmic pRb2/p130 expression was still correlated with that of VEGF (P < 0.05) and EZH2 (P < 0.01). Interestingly in the diffuse-type, generally manifesting a more aggressive biological behavior, the correlation between nuclear expression levels of pRb2/p130 and EZH2 was lost. We also found that the EZH2 expression correlated with those of p53 and Ki-67 (P < 0.01). Finally, the correlations between p21WAF1 and p27KIP1 expressions levels and between p53 and p16INK4A or Ki-67 were still maintained (P < 0.01).

Interestingly, in the analysis of diffuse-type gastric cancer specimens, nuclear expression of pRb2/p130 and p27KIP1 revealed a novel statistically significant relationship (P < 0.05), that was not found analyzing either the total patient population or just the intestinal histological type.

Comparing the data of the two main histological groups, we found some protein expression hallmarks that could underline the fact that the diffuse more than the intestinal type of gastric cancer has been linked to genetic alterations. Intriguingly, we found that in patients with diffuse-type gastric cancer p16INK4A was expressed in 45% and p53 in 40% of the cases, and that the two protein expressions were also highly correlated (P < 0.01). Additionally, the percentage of diffuse-type tumor samples expressing high levels of Ki-67 was low (30%), but also in this case there was a highly statistical correlation with p16INK4A (P < 0.01).

Correlation of EZH2, p16INK4A, p21WAF1, p27KIP1, pRb2/p130, VEGF, p53, and Ki-67 expression levels with clinicopathological parameters

Immunohistochemical expression levels and their associations with clinicopathological features in the 47 gastric cancers tissues samples were analyzed and are summarized in Table 4. Thirty-seven cases (78.8%) belonged to the high EZH2 expression group. Conversely, none of the corresponding normal mucosa expressed EZH2. Intestinal-type gastric cancers showed higher levels of expression of EZH2, VEGF, p53 (P < 0.05), and cytoplasmic pRb2/p130 or Ki-67 (P < 0.01). We also analyzed the clinicopathological features in relation to the proteins' expression using an ANOVA test considering the various tumor locations. We found that the expression of p27KIP1 (P < 0.001), p53 (P = 0.002), and Ki-67 (P = 0.009) correlated with intestinal-type tumors independently of the various gastric tumor locations: antrum, cardias medium, medium antrum, cardias medium antrum, and medium. No associations were found between the diffuse-type and any of the considered biological parameters considering the various tumor locations using the ANOVA test.

Table 4. Correlations between biological and clinicopathological parameters in gastric cancer
  Correlations with Spearman rP-value
LowModerateHigh 
  • pRb2/p130 n: nuclear pRb2/p130.

  • pRb2/p130 c: cytoplasmic pRb2/p130.

  • G1: G1 grade differentiated disease.

  • T > 1: T2, T3, T4.

  • a

    Inverse correlation.

pRb2/p130 nG10.291  <0.05
 T > 10.323  <0.05
pRb2/p130 cG10.340  <0.05
 Intestinal type0.399  <0.01
VEGFIntestinal type0.334  <0.05
EZH2Intestinal type0.362  <0.05
 Metastasis0.422  <0.01
p53Intestinal type0.370  <0.05
p27Metastasis−0.393  <0.05a
p21G10.322  <0.05
Ki-67Intestinal type0.377  <0.01

Importantly, while we were submitting our manuscript we found that in accordance with a very recent publication (Matsukawa et al., 2006), tumors expressing high levels of EZH2 correlated with more aggressive biological behavior. In fact, high EZH2 expressing tumors were those presenting with distant metastasis including those along the hepatic ilum (P < 0.01).

Notably, high p27KIP1 expression levels were correlated with lower risk of distant metastasis confirming a protective role for p27KIP1 in gastric carcinomas (P < 0.05). Additionally, high p21WAF1 and nuclear and cytoplasmic pRb2/p130 expressing tumors were classified as low-grade tumors (G1) (P < 0.05), confirming a protective role for these tumor-suppressor proteins in gastric tumors. Surprisingly, invasive tumors (T2-T4) showed higher expression levels of nuclear Rb2/p130 (P < 0.05).

Expression of p16INK4A, p21WAF1, p27KIP1, pRb2/p130, VEGF, p53, and EZH2 in gastric cancer cell lines

In order to verify the immunohistochemical results, we decided to study the expression pattern of the same proteins considered in the immunohistochemical experiments in a series of Western blot analysis (Fig. 2). We studied the expression levels of p16INK4A, p21WAF1, p27KIP1, pRb2/p130, VEGF, p53, and EZH2 in various gastric cell lines of Caucasian (AGS) or Japanese (N87, KATO-III), and Korean origin (YCC-2, YCC-3, and YCC-16). We found that p16INK4A was not expressed in AGS, N87, and KATO-III cells (Fig. 2A), while it was expressed abundantly in the cell lines of Korean origin (YCC-2, YCC-3, and YCC-16) (Fig. 2B). The NL-20 cells (human normal bronchial epithelium) were used as a control and showed high levels of p16INK4A. On the other hand, we found that p21WAF1 was not expressed in the cell lines of Korean origin (YCC-2, YCC-3, and YCC-16) even though these cells expressed p53 (Fig. 2B). The p21WAF1 protein was instead expressed abundantly in the AGS cells, which express low levels of p53. The p21WAF1 gene product was expressed less abundantly in N87 cells (Fig. 2A), which instead expressed a faster migrating form of p53 (*p53) when reacted with an antibody against p53 (clone D-11, which was raised using the entire p53 molecule as an antigen). The KATO-III cellular lysate did not reveal a band when reacted with an antibody against p53 because of a genomic deletion (Yokozaki, 2000).

Figure 2.

Western blot analysis of various gastric cells lines. A: Western blot analysis of total lysates extracted from AGS, N-87, and KATO-III cells. On the left are indicated the antibodies used. *p53 indicates a faster migrating form of p53. Total lysates extracts from NL-20 cells were used as a control for the p16INK4A antibody reaction. HSP72/73 was used as a loading control. B: Western blot analysis of total lysates extracted from YCC-2, YCC-3, and YCC-16 cells. On the left are indicated the antibodies used. Total lysates extracts from AGS cells were used as a control for the p21WAF1 antibody reaction. HSP72/73 was used as a loading control.

All the different cell lines expressed p27KIP1 at an abundant level. VEGF expression was also abundant in most of the cell lines. However, the N87 cells that showed high expression levels of hypophosphorylated form (active) of pRb2/p130 showed lower levels of VEGF when compared to the AGS and KATO-III cells, that expressed less abundant hypophosphorylated pRb2/p130 (Fig. 2A). Regarding the cell lines of Korean origin, it needs to be pointed out that the YCC-2 and -3 cells were derived from the ascite fluid, and YCC-16 cells derived from peripheral blood (Kim et al., 2005). Moreover, only the YCC-3 and -16 have been reported to grow in nude mice. In this respect, the YCC-2, -3, and -16 cells represent a progression model of gastric tumor aggressiveness. Analyzing the expression levels of VEGF and EZH2 in these cells, we found that there was an increasing expression level of these markers from the YCC-2 to the YCC-16 cells (Fig. 2B). The YCC-3 and -16 cells which are able to grow in nude mice, showed higher expression levels of VEGF and EZH2 when compared to the YCC-2 cells that were previously shown to bear lower biological aggressiveness. Additionally, the more biologically aggressive YCC-16 cells showed lower levels of hypo-phosphorylated pRb2/p130 and higher levels of VEGF, confirming previous data obtained in vitro (Claudio et al., 2001). HSP72/73 was used as a loading control.

DISCUSSION

Gastric carcinoma is a major cause of morbidity and mortality worldwide. The most reliable prognostic factors are tumor stage and completeness of excision. Tumor grade and histological type may be also useful factors. Although previous reports are conflicting, immunohistochemical studies are important in helping finding potential novel prognostic factors, since they may help predict not only baseline life expectancy, but also tumor response to specific anticancer drugs. This will be especially true with further uncovering of the contribution to disease progression by old and new proteins and their interacting pathways. In an era where many effective but expensive tests are being developed and used by the research community, immunohistochemistry still remains a widely used and affordable technique in clinical settings. The development of human cancers including gastric cancer is a multistep process and phenotypic changes during cancer progression reflect the sequential accumulation of genetic alterations in cells. We have performed a series of protein expression analysis by immunohistochemistry or Western blot of various proteins involved in the cell cycle and in angiogenesis in primary gastric tumor samples and in established gastric cancer cell lines. Cellular proliferation follows an organized and timely regulated progression through the cell cycle, which is controlled by protein complexes composed of cyclins and cyclin-dependent kinases (CDKs) (Tonini et al., 2002). The cell-cycle progression is determined by checkpoints between early and late G1 phases, and between the S- and G2/M-phases. A major contribution for cell-cycle regulation is due to the cdk inhibitors (CKIs) such as p16INK4A, p21WAF1, and p27KIP1. The p21WAF1 protein is transactivated and mainly controlled by p53 and its activation leads to G1-phase arrest of the cell cycle by inhibiting the kinase activity of cyclin-dependent kinase complexes regulating cell-cycle progression. The p16INK4A protein is overexpressed in cells defective in pRb function, and it may participate in a feedback loop wherein repression of p16INK4A expression by pRb may allow CDK4 to phosphorylate and inhibit pRb. The p16INK4A protein inhibits specifically CDK4 and CDK6, and by inhibiting their activity it inactivates pRb. The p27KIP1 protein is a negative regulator implicated in G1 phase arrest by inhibiting cyclin E–CDK2, cyclin A–CDK2, and cyclin D–CDK4 complexes, and abrogating their activity. Therefore, deregulation of any of these molecules may result into an uncontrolled proliferation.

The immunohistochemical assay revealed several significant differences in the expression of the investigated markers among our patients some of which were confirmed at the Western blot analysis. Abnormalities in p16INK4A and pRb are not infrequent in gastric cancer. Lack of p16INK4A expression is often due to hypermethylation of the p16INK4A promoter region. Previous studies have also shown that deletion of the p16INK4A gene is associated with the degree of differentiation and metastasis of gastric cancer. In our cases instead p16INK4A expression was not correlated to any of the available clinicopathological parameters. Additionally, 46.8% of our cases expressed high levels of p16INK4A. Serrano et al. (1993) have proposed that physiological inactivation of pRb during the G1-phase leads to increased p16INK4A expression in order to limit CDK4 activity. Inactivation of pRb would also stimulate cell to increase p16INK4A expression in an attempt to inhibit CDK4. This would establish a negative feedback in which pRb negative tumors would have high levels of p16INK4A, while pRb positive tumors might require decreased amounts of functional p16INK4A in order to achieve a level of CDK4 activity sufficient for pRb inactivation. This could be the scenario of the pRb status, at least half of our patient population.

In accordance with Al-Moundhri et al. (2005) we have also found that most of the tumor samples examined expressed high levels of p21WAF1 and p27KIP1. In our study we have also found that high levels of p21WAF1 significantly correlated with those of p27KIP1, and that high-levels of p27KIP1 inversely correlated with the presence of metastasis confirming a role for p27KIP1 as a tumor-suppressor gene in this type of cancer. Moreover, we also found that high levels of p21WAF1 expression directly correlated with differentiation-grade 1 (G1) gastric cancers. The fact that we found high-levels of p21WAF1 and p27KIP1 expression may be explained as follows. It has been proposed that the growth-inhibitory activity of the p27KIP1 and p21WAF1 may be modulated in advanced tumors through their inactivation (Polyak et al., 1994; Cheng et al., 1999). Western blot analysis of a panel of gastric carcinoma cells lines demonstrated that the biologically more aggressive cell lines AGS and YCC-16 contained more p27KIP1. To note that the AGS and YCC-16 cell lines are able to grow if injected subcutaneously in nude mice. Additionally, the AGS cell line of Caucasian origin showed a linear increase in p21WAF1 and p27KIP1, that is, in agreement with our immunohistochemical results. Surprisingly, we observed that while all the cell lines tested showed low levels of p53 expression, the N87 cells expressed abundantly a faster migrating form of p53 (labeled *p53 on Fig. 2A) when reacted with the antibody against p53 (clone D-11, which was raised using the entire p53 molecule as an antigen). We are currently investigating the significance of the apparent increased stability of p53 in this cell line and its genetic structure and therefore we can only speculate that the faster migrating form of p53 could be the result of a possible genetic alteration not infrequently found in the p53 family of proteins in many types of cancer (Murray-Zmijewski et al., 2006).

Another important step in gastric carcinogenesis is the upregulation of the angiogenic factor VEGF, which usually is associated to more aggressive tumors (Fondevila et al., 2004). We found higher VEGF expression levels in the intestinal-type of gastric carcinomas in accordance with previously published data (Chen et al., 2004). High levels of VEGF expression were also found in the cell lines tested. In particular, the Korean cell lines (YCC-2, -3, and -16) demonstrated a progressive increase of VEGF expression passing from the less to the more aggressive biological phenotype.

Importantly, tumors that expressed high levels of VEGF also expressed higher levels of cytoplasmic pRb2/p130 (P < 0.01) providing a link between the two proteins also in gastric carcinomas. The role of pRb2/p130 in the cytoplasm, if any, is still unknown and the significance of the potential link of the direct correlation between the expression of pRb2/p130 in the cytoplasmic compartment and that of VEGF in this type of cancer, remains to be elucidated.

Cytoplasmic, but also nuclear pRb2/p130 expressions were also found significantly correlated to differentiation-grade 1 (G1) tumors indicating the lack of a role for this protein in less differentiated tumors. Additionally, high cytoplasmic pRb2/p130 expression was found significantly correlated to the intestinal type, which develops following the pathologic sequential steps of atrophic gastritis, intestinal metaplasia, dysplasia, and carcinoma. We did not find cytoplasmic pRb2/p130 correlated to the less common diffuse-type, which presents with worse prognosis, develops after chronic gastritis and for which there are evidence of genetic predisposition (Caldas et al., 1999; Tahara, 2004; Correa and Schneider, 2005; Smith et al., 2006). Distal gastric cancer (non-cardial) is often of the intestinal-type and predominates in developing countries, among blacks, and in lower socio-economic groups, whereas proximal tumors (many of which show diffuse-type histology) are more common in developed countries, among whites, and in higher socio-economic classes. The main risk factors for distal gastric cancer include Helicobacter pylori infection and dietary factors, whereas gastroesophageal reflux disease and obesity play important roles in the development of proximal stomach cancer (Crew and Neugut, 2006). More studies are needed to assess the role of these proteins in the pathogenesis of gastric cancer especially with respect to the presence or not of Helicobacter pylori infection and dietary factors in the patients screened. Unfortunately, we did not have available these clinicopathological data at the time of our study and therefore we can not discuss the possible implications and modulations of molecular events due to the presence of Helicobacter pylori.

Additionally, cytoplasmic, but also nuclear pRb2/p130 expressions were also found significantly correlated to that of the newly identified human nuclear protein (EZH2) that shows sequence homology to the “Enhancer of Zeste” protein of Drosophila, and is therefore considered a member of the Polycomb group, which maintains homeotic gene repression and is thought to control gene expression by regulating chromatin (Varambally et al., 2002). A very recent publication reported that EZH2 expression is restricted to gastric cancer cells and that non-cancerous gastric mucosa showed faint or no EZH2 immunoreactivity restricted to the nuclei of glandular epithelial cells (Matsukawa et al., 2006). In the present study, we found that 78.8% of the cases expressed high levels of EZH2 and that the expression levels of EZH2 protein determined by Western blot analyses were in good agreement with those of the immunohistochemical analysis. Additionally, we found that EZH2 correlated with higher levels of p53 and Ki-67. Interestingly, these three proteins as well as cytoplasmic pRb2/p130 and VEGF were significantly associated with intestinal-type gastric cancer. In particular, high levels of EZH2 expression in gastric cancer tissues were significantly associated with the presence of metastasis indicating a possible role of this protein in gastric tumor spread. These results are in agreement with other reports indicating the re-expression of this protein in different human cancers (Varambally et al., 2002; Kleer et al., 2003; Dukers et al., 2004; Arisan et al., 2005; Gil et al., 2005; Weikert et al., 2005; Matsukawa et al., 2006), but pointing out that in the specific case of intestinal-type of gastric malignancy its re-expression occurs in tight correlation with the presence of metastasis, a factor that influence its prognosis. This study represents another evidence that the immunohistochemical investigation of various genes' expression could be of aid in understanding and predicting the aggressiveness of gastric malignancies. The fact that EZH2 could be a marker of aggressive subgroups in several cancers as well as in gastric neoplasias may be of significant practical interest, since the polycomb proteins have been proposed recently as candidates for targeted therapy in breast cancers (Takeshita et al., 2005). Additionally, the fact that EZH2 and Ki-67, pRb2/p130, and p53 showed tight correlations in their expressions, suggests that their association should be further studied as possible predictive factors in gastric malignancies.

Acknowledgements

This study was supported by a grant from the W.W. Smith Charitable Trust to P.P.C., and by NIH grants to A.G. Paraskevi Vogiatzi acknowledges the Ph.D. program: “Oncological Genetics” of the University of Siena, Italy.

Ancillary