MATERIALS AND METHODS

Clinical materials

EBV carrying GCIS

Samples from 11 EBV carrying GCIS patients were collected at the department of pathology, Vrije Universiteit medical centre Amsterdam, The Netherlands. These carcinomas were tested for the presence of EBV by EBER1/2 RISH, and data on 10 of these carcinomas have previously been described.¹⁰ All paraffin wax embedded material available from these cases—that is, all parts of the tumour and non-neoplastic tissue from the resection specimen and the preoperative biopsy (n  =  7)—was submitted to EBER1/2 RISH. In total, 29 normal gastric epithelia, 10 with chronic atrophic gastritis, 18 with intestinal metaplasia, two low grades, and one moderate dysplasia were evaluated. The diagnoses were confirmed by an experienced gastrointestinal pathologist (GJAO). Table 1 summarises the clinicopathological data of these 11 gastric carcinomas. All tumours were adenocarcinomas not otherwise specified. No lymphoepithelial-like carcinomas were studied.

EBV carrying GSC

Samples from eight EBV carrying GSC patients, as determined by EBER1/2 RISH, were obtained from the department of pathology, Academic medical centre, University of Amsterdam, The Netherlands. Preceding biopsies (n  =  8), containing normal gastric epithelium, IM, dysplasia, and carcinoma, collected in the context of a longstanding follow up were also submitted to EBER1/2 RISH. In addition, gastric tissues from the resection specimens with IM and dysplasia adjacent to EBV positive gastric carcinomas were also submitted to EBER1/2 RISH. In total, 24 normal gastric epithelia, five with CAG, 14 with IM, and three low grade dysplasias were evaluated. The diagnoses were confirmed by an experienced gastrointestinal pathologist (GJAO). Table 2 summarises the clinicopathological data of the patients with GSC. No lymphoepithelial-like carcinomas were studied.

EBER1/2 and β actin RISH

RISH was performed for the EBER 1/2 transcripts and the β actin target as described previously.²⁰ Paraffin wax embedded tissues from preoperative biopsies and the gastrectomy of the tumour and adjacent non-neoplastic mucosa of EBV carrying GCIS and GSC samples were subjected to the non-radioactive EBER1/2 RISH using digoxygenin labelled antisense and sense EBER1/2 probes as described previously.²⁰

Immunohistochemistry

To detect EBV specific proteins, which are known to be expressed in EBV associated epithelial malignancies,¹⁹ monoclonal antibodies against LMP-1 (CS1-4; Dako, Glostrup, Denmark) and EBNA-1 (2B4-1; a kind gift from Dr Grässer, Homburg, Germany) were used. Detection of the EBV receptor was performed with a monoclonal anti-CD21 antibody (Dako; clone 1F8). All antibodies were visualised with an avidin–biotin–horseradish peroxidase complex and diaminobenzidine/H₂O₂ staining method, as described previously by our laboratory.^20–22

RESULTS

EBV carrying GCIS

EBER1/2 positivity was found in all tumour cells of the 11 EBV carrying GCISs, both in the resection specimens and in the preoperative biopsies (table 1; fig 1A).¹⁰ EBER1/2 transcripts were never found in the mucosa adjacent to these EBV carrying GCISs, comprising normal gastric tissue (n  =  29), CAG (n  =  10), IM (n  =  18), and dysplasia (n  =  3) (table 1). RISH for β actin performed on corresponding tissue sections revealed sufficient RNA quality. IHC for EBNA-1 and LMP-1 expression on representative tissue of 10 of the cases showed the presence of EBNA-1 protein in the carcinoma cells in four of 10 cases (fig 1B),¹⁰ whereas no LMP-1 staining was found in the neoplastic cells or non-neoplastic epithelial cells (fig 1C). EBNA-1 expression was not detected in the adjacent gastric mucosa of the four EBNA-1 positive gastric carcinomas, neither in the normal epithelium (n  =  13) nor in areas of IM (n  =  3).

Moreover, CD21 IHC was performed on EBV carrying GCIS and its precursor lesions. No positive staining was found in non-neoplastic or neoplastic epithelial cells (fig 1 D), whereas positive staining was seen in dendritic cells and B cells.

Gastric stump carcinomas

Specific nuclear EBER1/2 RISH transcripts were restricted to the tumour cells of GSCs in the gastrectomy specimens and the preceding biopsies (n  =  17) (table 2, fig 2). No EBER1/2 transcripts were detected in normal gastric epithelium (n  =  24), CAG (n  =  5), or IM (n  =  14) in the mucosa adjacent to the carcinoma tissue of both the gastrectomy tissue (fig 2A and B) and the preoperative biopsies (table 2). In three cases (patients 6, 7, and 8), biopsy material was available 4, 18, and 18 years before the diagnosis gastric carcinoma, respectively. These biopsies all contained low grade dysplasia. In all of these cases, EBER1/2 RISH was negative in the preceding biopsy, both in the normal and the dysplastic epithelium (table 2)

The resection cases were tested for LMP-1 expression by immunohistochemistry (n  =  8). No positive staining for LMP-1 was detected in the carcinoma cells, the preneoplastic lesions, or non-neoplastic gastric epithelium.

DISCUSSION

Growing evidence of a role for EBV in gastric carcinogenesis derives from the following observations: EBV carrying gastric carcinomas harbour distinct chromosomal aberrations and are characterised by a unique transcription pattern resembling but not identical to nasopharyngeal carcinomas.^10,11 In addition, EBV carrying gastric carcinomas display specific clinical features compared with EBV negative gastric carcinomas.22a Therefore, the purpose of our present study was to determine the stage of gastric carcinogenesis at which EBV possibly enters the gastric epithelium. Although not firmly established, the present hypothesis on gastric carcinogenesis—of at least the intestinal type of gastric carcinoma—comprises the sequence of gastritis, CAG, IM, dysplasia, and invasive carcinoma.² Therefore, we investigated the presence of EBER1/2 transcripts in these presumed precursor lesions and neoplastic gastric epithelium of EBV carrying GCIS and GSC samples. Because the viral genome is present in limited numbers in most EBV associated malignancies, DNA ISH is not very sensitive. Therefore, we used EBER RISH, a technique that measures EBER1/2 RNAs, which are highly abundant in most EBV carrying tumour cells, and is the most sensitive method to demonstrate the presence of EBV.¹⁹ This is the first study in which two types of gastric carcinoma (GCIS and GCS) were investigated and all available paraffin embedded material from each case was analysed.

Specific nuclear EBER1/2 transcripts were found only in the carcinoma cells of both types of gastric carcinoma, and not in the adjacent normal, metaplastic, or dysplastic epithelia. In addition, EBV was detected in none of the normal or dysplastic epithelia of the preceding biopsies in EBV positive GSC.

In addition to EBER1/2 RISH, IHC for EBNA-1 and LMP-1 was carried out. Expression of EBNA-1 was restricted to gastric carcinoma cells (GCIS). No expression was found in the presumed precursor lesions of gastric carcinoma. EBNA-1 protein expression as measured by IHC was not detected in the cases of EBV positive gastric carcinoma probably because of the low content of EBNA-1 and the lower sensitivity of IHC in comparison with EBER1/2 RISH.²² Interestingly, LMP-1 expression was not found in the EBV positive carcinomas and their precursor lesions, which is in line with previous observations.^23,24

“Our data strongly suggest that Epstein-Barr virus is only present in susceptible cells of gastric carcinomas but not in its supposed precursor lesions”

Our results agree with other studies in which EBER transcripts have not been detected in adjacent precursor lesions, such as IM.^13–15,23 However, some reports describe the presence of EBV in dysplasia,^3,16 and some even in IM and CAG.^17,18 There are several reasons for these discrepancies. First, the problem associated with diagnosing dysplasia and carcinoma is well recognised.² In addition, dysplasias adjacent to carcinomas are difficult to distinguish from local carcinoma spread. The second reason is the use of different techniques and the interpretation of the results. Yanai and co-workers reported EBNA-1 and LMP-1 staining in CAG and IM.¹⁷ Because we used the same antibodies for EBNA-1 and LMP-1 this discrepancy is surprising. However, in our longstanding experience with these antibodies,^10,21,22 we have found that crossreactivity can occur and that the interpretation of positive IHC results should always be done in the context of transcript analysis by reverse transcription polymerase chain reaction.^10,22

It is of interest that EBV carrying GCIS and EBV carrying GSC did not differ with respect to the distribution of EBER1/2 transcripts. Although the carcinogenesis of both of these entities is closely linked to H pylori, in GSC bile reflux is thought to play a major role in the multistep process of gastric carcinogenesis.

The fact that the expression of EBER1/2 transcripts and EBNA-1 protein is restricted to the tumour cells in EBV carrying GCIS and GSC strongly indicates that EBV enters the gastric epithelium at a late stage of the multistep process of gastric carcinogenesis. In addition, we showed that CD21, the EBV receptor, is not present in normal, preneoplastic, and neoplastic gastric cells. This is in line with the observations of others,^25–27 and suggests the existence of an alternative EBV receptor on epithelial cells,^28,29 or an alternative route of entry of EBV into epithelial cells.

Because dysplasias are EBER1/2 RISH negative and all tumour cells carry EBV it is highly likely that EBV infects a neoplastic gastric epithelial cell, providing a growth advantage to the EBV positive tumour cell. Alternatively, EBV might infect a highly dysplastic gastric epithelial cell, transforming it into a carcinoma cell. Another explanation might be that gastric carcinogenesis develops through an as yet unknown route, in which the morphology of the precursor lesion has not yet been identified. However, all hypotheses point to a role for EBV late in the process of gastric carcinogenesis and explain why monoclonal EBV has been reported in all tumour cells of EBV carrying GCISs and GSCs.

In summary, our data strongly suggest that EBV is only present in susceptible cells of gastric carcinomas but not in its supposed precursor lesions. This suggests that EBV enters the gastric epithelium at a late stage of the multistep carcinogenic process and that EBV plays a role late in the process of gastric carcinogenesis. This is strengthened by the demonstration of restricted EBV transcription in the tumour cells of two pathogenetically related but distinct models of gastric carcinogenesis.