Epstein-Barr virus (EBV) is a human oncogenic virus that has been identified in a wide variety of malignancies. Representative examples are nasopharyngeal carcinoma (80,043 new cases/year), about half of all Hodgkin's lymphoma (62,329 new cases/year), and posttransplant lymphoproliferative disease (<2% of recipients of solid organ transplants).1, 2 EBV-associated gastric carcinoma (GC) comprises about 10% of all GC throughout the world.3 Since the worldwide incidence of GC is reportedly 934,000 new cases/year, EBV-associated GC is estimated to occur in nearly 93,400 new cases/year.1
A causal role of EBV in GC has been suggested on the basis of the clonal nature of EBV in neoplastic cells, the presence of EBV in almost all cancer cells and the absence of EBV in noncancerous mucosa.3 EBV-associated GC not only has several distinct clinicopathological features, such as male predominance, localization of gastric corpus, and the accompaniment of lymphocytic infiltration, but it exhibits global and nonrandom DNA methylation of the promoter regions of various cancer-associated genes.4, 5, 6, 7 This promoter methylation correlated well with p16 and E-cadherin expression abnormalities in EBV-associated GC, whereas such correlation was indistinct in EBV-negative GC.5, 7 Regarding methylation density, all CpG sites of p14 and p16 promoter regions were uniformly methylated in EBV-associated GC, while it was sporadic or variable or both in EBV-negative GC.8 These findings suggest that the mechanisms of de novo or maintenance methylation differ markedly between gastric carcinomas with and without EBV infection.
In this study, we investigated the gene expression of p73, a candidate tumor suppressor. p73 is mapped to the human chromosome 1p36.2-3, a region which is frequently lost in a wide variety of human tumors including neuroblastoma.9 The sequence-specific DNA-binding domain, the amino-terminal activation domain, and the carboxy-terminal oligomerization domains of p73 are similar to those of p53. Accompanying these structural similarities, p73 can act as transcription factors and regulate the expression of similar groups of genes by means of direct binding to what were originally identified as p53-binding sites within promoters. Transcriptional activation of these target genes leads to the induction of cell-cycle arrest and apoptosis.9, 10, 11 This evidence suggests that p73 may act as a tumor suppressor with the overlapping function of p53.
We first evaluated the expression of the p73 gene by immunohistochemistry in EBV-associated and EBV-negative GC, and compared with the analysis of p53 expression. Interestingly, we observed that loss of p73 expression, independent of p53 abnormality, specifically occurred in EBV-associated GC. This result led us to further evaluate p73 methylation to clarify the significance of p73 and p73 methylation in the development of EBV-associated GC. The study included analysis of the methylation status of p73, p14 and p16 in early stage GC and nonneoplastic mucosa, and evaluated a possible sequence of EBV infection and CpG methylation in epithelial cells of the stomach. Toyota et al.,12 who first defined the CpG island methylator phenotype (CIMP) in GC, pointed out that some normal mucosal tissues adjacent to CIMP carcinoma also showed aberrant methylation of multiple genes. Aberrant DNA methylation has also been thought to play an important role in the development of carcinoma in association with chronic inflammation, such as ulcerative colitis, Barrett's esophagus, and chronic hepatitis.13, 14, 15, 16 Thus, in spite of limitations using formalin-fixed material, the evaluation of early stage EBV-associated GC and normal mucosa is expected to provide an insight to clarify the development of this specific type of GC associated with a human oncogenic virus.
Material and methods
Thirteen EBV-associated and 38 EBV-negative GCs, treated at Jichi Medical School from 1996 to 1998, were examined. Fresh tissue was obtained and stored frozen until use. In clinicopathological features, the histological type, according to the Japanese classification,17 was the only significant difference between GC cases with and without EBV infection. Moderately differentiated tubular type and poorly differentiated solid type were significantly more prevalent in EBV-associated GC (13/13) than in EBV-negative GC (12/38). Age, male/female ratio and depth did not differ significantly between the 2 groups. EBV infection was identified by EBV-encoded small RNA 1 in situ hybridization, using previously reported procedures.18
Regarding the evaluation of early GC and its adjacent nonneoplastic mucosa, 40 early carcinoma cases, 20 cases each of EBV-associated and EBV-negative GC, were retrieved from the files of the Pathology Division of Tokyo University Hospital from 1990 to 2000. These carcinomas were less than 3 cm in diameter and in the early stage, i.e., carcinoma localized within the mucosa or showing invasion only to the submucosa. Formalin-fixed and paraffin-embedded sections were examined.
Histological evaluation of gastritis and H. pylori infection
The grade of gastritis and the presence of H. pylori were evaluated in nonneoplastic tissues adjacent to the early gastric carcinomas described earlier. Two pathologists (TU and MCJ), who were not informed of the methylation-specific polymerase chain reaction (MSP) results, evaluated gastritis in hematoxylin and eosin-stained slides serial to those used for MSP analysis. On the basis of the modified Sydney system,19 the degrees of neutrophil infiltration, mononuclear cell infiltration, atrophy, and intestinal metaplasia were graded with reference to the diagram. Normal to mild pathologic findings were defined as grade 1, moderate findings as grade 2 and severe findings as grade 3. The presence of H. pylori infection was determined by Giemsa staining.
Immunohistochemical analysis of p73 and p53 was performed as described previously.5, 7 The primary antibodies used were a mouse monoclonal anti-p73 antibody (1:50 dilution; clone GC-15, NeoMarkers, Fremont, CA), and a mouse monoclonal anti-p53 protein antibody (1:100 dilution; clone DO-7, Dako Japan, Kyoto, Japan). Antigen retrieval was performed by autoclaving the sections in 10 mM citrate buffer (pH 6.0) for 10 min at 120°C. Endogenous peroxidase was blocked with 3% hydrogen peroxide in methanol for 20 min and washed 3 times with cold 0.01 M phosphate-buffered saline. After a second blocking with 10% normal goat serum, the sections were incubated for 16 hr at 4°C with either monoclonal antibody, washed 3 times with TBS, and incubated with a biotinylated secondary antibody for 20 min at room temperature. The sections were then reacted with a streptavidin–biotin peroxidase reagent (LSAB2 Kit, Dako Japan). The antigen was visualized with a chromogen, diaminobenzidine, in 3% hydrogen peroxide, and the sections were counterstained with hematoxylin, dehydrated and mounted. To obtain negative controls, the primary antibodies were omitted.
For the p73 immunohistochemical study, positivity was defined as cases with more than 5% of p73-positive cells. The results of p53 immunohistochemistry were evaluated as follows: negative, no tumor nuclei positive for p53; sporadic, several p53-positive nuclei scattered in neoplastic cells (less than 10% of neoplastic cells); focal, a cluster of p53-positive neoplastic cells was present (10–50% of neoplastic cells); diffuse, more than 50% of neoplastic cells were positive for p53.
Methylation-specific polymerase chain reaction
The methylation status of p73 exon 1 was determined by MSP using bisulfite-modified DNA. The methylation status of p14 and p16 was evaluated in previous studies.5, 6
DNA was modified by bisulfite reaction using the CpGenome DNA Modification Kit (Intergen, Purchase, NY). After this reaction, all unmethylated cytosines are deaminated and converted to uracil, while methylated cytosines remain unchanged. Following bisulfite conversion, methylated and unmethylated genomic regions could be distinguished by polymerase chain reaction (PCR), using each sequence-specific pair of primers. Primer sequences for methylated and unmethylated alleles of p73 as well as p14 and p16 are presented in Table I. MSP was performed in a final volume of 12.5 μl containing 1μl bisulfite-modified DNA, 250 μmol/l dNTP, 1.5 mmol/l MgCl2, 1×PCR gold buffer, and 0.5 unit AmpliTaq Gold DNA polymerase (Applied Biosystems, Foster City, CA), 1 μmol/l of each primer with the following cycling parameters: 95°C for 5 min, 35 cycles of 95°C for 45 sec, specific annealing temperature, that is, 68°C for the methylation primer and 59°C for the unmethylation primer, for 45 sec, 72°C for 1 min, and a final extension step at 72°C for 10 min. PCR products were electrophoresed on 3% agarose gel and visualized by ethidium bromide staining. As a control, bisulfite-modified unmethylated/methylated DNA and no template control (distilled water) were included in each experiment. MSP experiments were performed at least in duplicate.
Table I. Primer Sequences and PCR Conditions for MSP Analysis
Forward primer (5′-3′)
Reverse primer (5′-3′)
Product size (bp)
Annealing temp (°C)
M: primers used for methylated DNA, U: primers used for unmethylated DNA.
To confirm the completion of bisulfite modification and to rule out false amplification, MSP products were sequenced in methylated (n = 8) and unmethylated (n = 3) samples. The specific band was retrieved from the gel and purified using the Qiaex II gel extraction kit (Qiagen, Hilden, Germany). The purified products were cloned into a TA cloning vector, and then sequenced using a DYEnamicTM ET terminator cycle sequencing kit (Amersham Pharmacia Biotech, Uppsala, Sweden) with a DNA sequencer (Model 373A, Applied Biosystems).
To further analyze the methylation status of the p73 gene, we employed bisulfite sequencing to evaluate the frequency of CpG methylation. We examined 4 EBV-associated GCs, all were identified as methylated by MSP, and 4 EBV-negative GCs, 3 of which were identified as unmethylated and the other as methylated by MSP. Briefly, using the same DNAs as for MSP analysis, we amplified the region shown in Figure 3A, which contains 45 CpGs, including the sequences analyzed by MSP. PCR reaction and sequencing were performed as described earlier. Primers for bisulfite sequences are listed in Table I.
Tissue microdissection of formalin-fixed tissues and MSP analysis
DNA was extracted from microdissected tissue from formalin-fixed, paraffin-embedded sections. Nonneoplastic mucosa was microdissected from 2 regions adjacent to the tumor. Two serial sections (8 μm in thickness) were deparaffinized in xylene and rehydrated in alcohol. After the sections were briefly stained with hematoxylin, a 5 × 5 mm2 region was dissected with an 18G needle tip. The tissue samples were treated with proteinase K (1 mg/ml) at 37°C overnight. After phenol–chloroform–isoamyl alcohol extraction and ethanol precipitation, each 1 μg of DNA was subjected to MSP analysis, as described earlier.
Statistical analyses were performed using the chi-square test, Fisher's exact test, or Mann-Whitney's U test. Differences were considered significant at a two-tailed p < 0.05.
Immunohistochemistry of p73 and p53
The immunohistochemistry results of p73 and p53 are presented in Table II.
Table II. Immunohistochemistry of p53 and p73, and Promoter Methylation of p73 in Gastric Carcinoma with and without EBV Association
In nonneoplastic mucosa, a monoclonal antibody against p73 labeled the nuclei of gastric epithelial cells of the gastric pit (Fig. 1a), pyloric glands, and intestinal metaplasia, as well as infiltrating inflammatory cells, which served as an internal control. In gastric carcinomas, there were varying numbers of p73-positive cells in 35 of 38 EBV-negative GC (Fig. 1b). On the other hand, p73-positive cells were hardly observed in 11 of 13 EBV-associated GC (Fig. 1c), and there were only occasionally positive cells in 2 cases.
For p53 immunohistochemistry, diffuse, focal and sporadic positivity was observed in 1, none and 4 of 13 cases of EBV-associated GC, respectively (Fig. 1d). On the other hand, diffuse, focal and sporadic positivity was found in 7, 4, and 1 of 38 cases of EBV-negative GC, respectively. All p53-positive cases were negative for p73 in EBV-associated GC, while all positive cases were positive for p73 in EBV-negative GC, except for 1 case showing a diffuse pattern of p53 positivity.
p73 methylation status
When CpG island methylation of the p73 gene was evaluated by MSP (Fig. 2), it was observed in 12 of 13 cases of EBV-associated GCs, and in 2 of 38 cases of EBV-negative GCs. Sequencing of PCR products confirmed the expected patterns of bisulfite-induced changes in all examined cases.
In the correlation between p73 methylation and immunohistochemical expression, all but 1 case of p73 methylation showed a negative expression of p73 in EBV-associated GC, but 2 cases with p73 methylation showed a positive expression in EBV-negative GC.
Bisulfite sequencing of the CpG island of the p73 gene
The methylation status of 45 CpG sites of p73 gene is shown in Figure 3B. In EBV-associated GCs, all clones showed global and high-density methylation in all 4 cases examined, and the mean numbers of methylated CpG sites were 44.2, 43.5, 43.7 and 39, respectively. In particular, the first 3 cases had some clones totally methylated in 45 CG sites. One of four EBV-negative GCs, identified as p73 methylated by MSP, showed global and high-density methylation with an average of 43.6 CpG sites of methylation, which was comparable to EBV-associated GCs. On the other hand, 3 of 4 EBV-negative GCs, identified as unmethylated by MSP, showed only low to moderate methylation variability at an average of 7.2, 0.8 and 20.8 CpG sites of methylation. In regard to the region covered by MSP analysis, the sequences of all clones were consistent with the results of MSP.
Methylation status in early carcinoma and adjacent nonneoplastic mucosa
CpG island methylation was evaluated in formalin-fixed, paraffin-embedded cancer tissues and adjacent nonneoplastic tissues of early gastric carcinomas (Fig. 4).
The methylation frequencies of p14, p16 and p73 in cancer tissues (Fig. 4A) were 95%, 80% and 60% in 20 EBV-associated GCs, and 45%, 58% and 5% in 20 EBV-negative GCs, respectively. Each frequency in p14, p16 and p73 was significantly higher in EBV-associated GC than in EBV-negative GC.
We examined nonneoplastic mucosa from 2 regions adjacent to cancer tissue. Forty regions were examined in EBV-associated and -negative GC cases, respectively, but the following results were not equal in total number because some were noninformative by MSP. The methylation frequencies of p14, p16 and p73 (Fig. 4B) were 24% (8/33), 18% (6/34) and 8% (3/38) in cases of EBV-associated GC, and 27% (6/23), 11% (4/35) and 3% (1/35) in cases of EBV-negative GC, respectively. Each frequency was significantly lower than that of the corresponding cancer tissues in EBV-associated GC, while only the frequency of p16 methylation was significantly different between nonneoplastic and cancer tissues in EBV-negative GC.
EBV-negative GC consisted of 6 cases of diffuse-type and 14 cases of intestinal-type GC. We evaluated the methylation frequencies in both types, which showed similar frequencies to those in EBV-negative GC (Fig. 4A and 4B).
Chronic gastritis, H. pylori infection and methylation
The pathologic changes of neutrophil infiltration, mononuclear cell infiltration, atrophy and intestinal metaplasia were graded in nonneoplastic mucosa adjacent to EBV-associated GC and EBV-negative GC (Table III). The nonneoplastic mucosa of EBV-associated GC showed a higher grade of mononuclear cell infiltration and atrophy compared to that of EBV-negative GC (p = 0.0097 and 0.0046, respectively). These differences were still significant when compared to either diffuse or intestinal EBV-negative GC.
Table III. Gastritis in Adjacent Non-Neoplastic Mucosa of Early Gastric Carcinoma Cases with and without EBV Association
EBV-associated GC1 (n = 40)
1 vs. 2
3 vs. 4
Diffuse3 (n = 12)
Intestinal4 (n = 28)
The data show the number of grade 1/2/3 cases. NS: not significant.
p = 0.0097
p = 0.0046
H. pylori (+/−)
As for the correlation of methylation with chronic gastritis, p14 methylation correlated with a moderate degree of atrophy in EBV-negative GC (p = 0.005), but not with atrophy in EBV-associated GC (Fig. 5). The latter seemed due to marked atrophy in older patients, since p14 methylation occurred more frequently in younger patients (the age of positive vs. negative cases, 53 ± 6 years vs. 62 ± 9 years) in EBV-associated GC (p = 0.015). There was no correlation of p16 or p73 methylation with age or pathologic factors by statistical analysis; however, p73 methylation was identified in mucosa without H. pylori infection in all 4 samples.
p73 has been identified as a transcription factor with structural and functional homology to a tumor suppressor, p53, the abnormality of which has been intensively studied in GC. The p53 immunohistochemistry results in this study are compatible with observations in previous studies, confirming a limited role in EBV-associated GC.20, 21, 22, 23 On the other hand, the significance of p73 has not yet been clarified in GC with or without EBV association. p73 mutation is extremely rare.24 Frequent 1p36 deletion may eliminate functional p73 in a large number of tumors.9 EBV involvement is rare in these tumors, although EBV-positive nasopharyngeal carcinoma and a small number of posttransplant lymphoproliferative diseases suffer such alteration.25, 26 The LOH of p73, without the accompaniment of p73 mutation, occurs preferentially in a subset of well-differentiated GC with foveolar epithelial phenotype.27 Tomkova et al.28 recently demonstrated that p73 and ΔNp73 were frequently overexpressed in >60% of GC. ΔNp73 not only behaves as an antagonist of p73, but induces β-catenin upregulation and T-cell factor/lymphocyte enhancement factor-dependent transcription. In this study, we demonstrated another mechanism of p73 involvement in GC, the suppression of p73 through promoter methylation, which occurs specifically in EBV-associated GC. Kang et al.29 observed that p73 expression in nonneoplastic stomach mucosa is derived from either allele, suggesting that the p73 gene is imprinted. If promoter methylation occurs additionally in the nonimprinting allele, p73 methylation might directly result in the loss of p73 expression. Transcriptional inactivation of the p73 gene by promoter methylation has been reported in other EBV-associated lymphoid malignancies, such as NK cell lymphoma30, 31 and Burkitt's lymphoma.32 Therefore, there might be a mechanism much more specific to EBV in p73 methylation than the global methylation observed in EBV-associated GC.
In this study, comparative MSP analysis of p14, p16 and p73 methylation, using 20 cases each of formalin-fixed and paraffin-embedded tissues of early GC carcinomas with and without EBV infection, confirmed 2 types of methylation: global methylation with increased rates (p14 and p16) and specific methylation of p73 in EBV-associated GC. As for the methylation status, the methylation of p14 and p16 promoter regions is distributed uniformly throughout the whole region in EBV-associated GC, while it is sporadic or variable or both in EBV-negative GC.8 In this study, we confirmed high-density methylation of CpG islands in the p73 gene. This suggests that a common mechanism generating high-density methylation induces global and specific methylation in EBV-associated GC. One of the important roles of DNA methylation is to act as a genome defense against foreign DNA.33 In the case of viral infections, the viral genome is suppressed and latent infection is restricted by the methylation of episomal genomes such as EBV, or integrated virus genomes, such as HIV type 1 or human T-cell leukemia-virus type-1.34, 35, 36, 37, 38 Although there are few reports dealing with DNA methylation in the host cell,39, 40 our findings suggest that the overriding machinery causes global and uniform methylation of CpG islands of host genes, leading to the development of EBV-associated GC. The aberrant methylation of cancer-related genes is relatively frequent in mesothelioma, which is infected with SV40.41 So far, to our knowledge, there have been few studies investigating the hypermethylation mechanism associated with viral infection, and none to document the specific combination of the virus and its target gene.
It is of interest to note that the frequency of p73 methylation was relatively lower than p14 and p16 methylation in early carcinomas of EBV-associated GC. This result indicates that selection clearly plays a part in the developmental process: both global and specific methylation can occur concurrently in the early stage of this type of GC, and a certain clone or clones are selected to grow within the mucosa.42 Toyota et al. demonstrated that some normal mucosae adjacent to CIMP GCs showed aberrant methylation of multiple genes, suggesting that CIMP GC originates from gastric epithelial cells through the accumulation of aberrant DNA methylation.12 In our study of nonneoplastic mucosa adjacent to EBV-associated GC, we did not identify mucosal tissue showing the concurrent methylation of multiple genes, although the number of examined genes was limited. Since EBV-associated GC consists of EBV-infected monoclonal cells, this indicates that the process of methylation and selection occurs within EBV-infected cells, that is, after EBV infects nonneoplastic epithelial cells. Further studies are necessary to clarify the interaction or relationship between global and specific methylation processes in EBV-infected epithelial cells.
Chronic inflammation is also an important factor for the development of GC. In this study, we confirmed the observation of our previous studies that higher grades of atrophy and infiltration of lymphocytes are characteristic of nonneoplastic mucosa adjacent to the early carcinoma of EBV-associated GC.43, 44 Kang et al.45 recently investigated the methylation status of multiple genes in a large sample collection of chronic gastritis, intestinal metaplasia and GC, and found that various genes show different methylation behavior in the lesions; for example, (i) genes showed low methylation frequency in chronic gastritis and intestinal metaplasia, but significantly higher frequency in GC (COX-2, hMLH1 and p16), and (ii) genes showed an increasing tendency along the sequence of the lesions (DAP-kinase, THBS1, TIMP-3 and p14). The latter also showed a general progressive increase in methylation frequency as a function of aging.46 The correlation between p14 methylation and atrophy in nonneoplastic mucosa of EBV-negative GC is consistent with these findings. No such correlation was observed in EBV-associated GC, although it is worth noting that p73 methylation was relatively frequent in EBV-associated GC without association with H. pylori infection. Thus, although augmented pressure toward CpG methylation was not demonstrated in nonneoplastic mucosa, EBV-associated GC was accompanied by a specific type of inflammation, which may facilitate p73 methylation in EBV-infected cells independent of H. pylori infection.
In conclusion, loss of p73 expression, independent of p53 abnormality, is primarily caused by aberrant methylation of the p73 promoter in EBV-associated GC. Specific p73 methylation occurred concurrently with global methylation in the early stage of this type of GC. A specific type of gastritis may facilitate p73 methylation and the development of EBV-associated GC.