.. Table 1 Differential immunoexpression of cyclin D1, p21WAF1, and NF-��B RelA between EBV-positive gastric carcinomas and EBV-negative gastric carcinomas We categorized these 120 tumors into diffuse type (poor glandular formation) and intestinal type (well-developed glandular contain formation) in the Lauren classification. Of 23 EBV-positive gastric carcinomas, 21 were diffuse type (91%), whereas 57 of 97 (59%) of EBV-negative gastric carcinomas were diffuse type. DISCUSSION The present study highlights the important role of the EBV-encoded BARF1 gene in proliferation of EBV-infected gastric carcinoma. This is consistent with a recent report that cell growth was activated in BARF1-transfected HaCaT immortalized human keratinocytes (39). We did not find any effects of BARF1 expression on apoptosis in the present study.

However, others suggested that the intracellular N-terminal fragment of BARF1 contributes to its antiapoptotic function as an oncogene in rodent fibroblasts (21). This conflict might be due to context, as different cell types may show different biological effects from BARF1 expression. Alternatively, additional factors may be required to enhance the antiapoptotic effect of BARF1 in gastric cancer cells. For example, in studies that examined the anticancer drug paclitaxel (originally named taxol), the antiapoptotic role of BARF1 was associated with increased expression of bcl-2 (22). However, we have previously shown that bcl-2 is rarely expressed in surgically resected EBV-positive gastric carcinoma tissues (9).

Secreted BARF1 may contribute to the increased proliferation observed in BARF1-expressing cells. This is consistent with previous suggestions that secreted BARF1 contributes to viral oncogenesis (17�C19, 27�C35). Additionally, in order to verify the autocrine/paracrine effect of BARF1, we put supernatant from SNU601 BARF1 (50-fold concentration) or supernatant from SNU601 (50-fold concentration) into original SNU601 cells, respectively. Then, in comparing cell proliferation rates, cells with BARF1 supernatant appeared to show greater proliferation than did cells with supernatant (minus BARF1), but the finding was not significant statistically (P = 0.06) (data not shown). There might be some obstacles to the proliferation effect of BARF1 supernatant, such as influence of pH and adsorption or deformation of macromolecules like growth-related factors during membrane penetration for concentration, or no persistency of BARF1 secretion, etc.

The present study also showed that BARF1 protein was mainly secreted into culture supernatants and only marginally detectable within the cells (Fig. 1D). This is consistent with recent observations in transfected HaCaT cells (39). Meanwhile, most papers in the past 16 years have described BARF1 as being almost completely secreted. The functions of intracellular Batimastat BARF1 protein were not addressed in the present study.