However, this was not a phenotypic trait common to all viruses associated with gastric cancer, as it was not found in cells infected with GP202 (Supplementary Figure 4) and it did not correlate with cell growth rate (Figure ?(Figure1).1). cells, gastric carcinoma cells and gastric spheroids more efficiently than Akata Indeglitazar or B95-8. Reciprocally, Akata and B95-8 had a stronger tropism for B cells than YCCEL1 or M81. These data suggest that different EBV strains will induce the development of lymphoid tumors with variable efficacy in immunocompromised patients and that there is a parallel between the cell tropism of the viral strains and the lineage of the tumors they induce. Thus, EBV strains can be endowed with PI4KB properties that will influence their transforming abilities and the type of tumor they induce. and at unusually high levels and also had a high propensity to infect epithelial cells [13]. EBV lytic replication has been identified as a cancer risk factor as populations at risk for NPC evince high level of antibodies against viral lytic proteins [4, 14, 15]. These phenotypic traits are not shared by B95-8, a virus strain that has extensively been studied and that is genetically close to viruses found in Western countries where the incidence of NPC is low [12]. These observations demonstrate the existence of distinct EBV subtypes and suggest that the unusual properties evinced by M81-type viruses are likely to explain their tight association with NPC. Whilst the contribution of a subtype of EBV to NPC has been extensively studied, its implication in the development of gastric carcinoma (EBVaGC) has been comparatively neglected. The percentage of EBV-positive cases of gastric carcinomas is on average 10%, but can vary from 4 to 18% in different geographic areas and populations [16, Indeglitazar 17]. The risk factors for the development of this tumor have not been clearly identified [18, 19]. In this paper, we report a comparative analysis of multiple EBV strains including three strains isolated from gastric carcinomas, with regard to their transformation abilities and cell tropism. RESULTS Generation of a panel of EBV strains, construction of a recombinant YCCEL1 virus and isolation of GP202 We collected a panel of EBV strains involved in different diseases and that infected individuals from different regions of the world (Supplementary Table 1). This panel included the recombinant viruses B95-8, Akata and M81. We also cloned the genome of the YCCEL1 virus from a gastric carcinoma cell line (Supplementary Figure 1A and 1B). The recombinant virus was stably transfected in 293 cells to generate Indeglitazar a producer cell line that delivers high virus titers (Supplementary Figure 1C). In this recombinant virus, the F-plasmid is flanked by terminal repeats and is excised with high efficacy upon infection of B cells (Supplementary Figure 1D) [13]. Furthermore, we infected marmoset peripheral blood B cells with viruses rescued from SNU719 and GP202, 2 gastric carcinoma cell lines, to generate virus producer cell lines. GP202 was established from a gastric carcinoma that arose in a Portuguese patient and we performed an EBER staining to show that it is EBV-positive (Supplementary Figure 2A). Thus, it allows comparison with other gastric carcinoma viruses such as SNU719 and YCCEL1 that were isolated in Korean patients. Sequencing of the EBNA2 gene showed that GP202 is also a type A EBV strain (Supplementary Figure 2B and 2C). Different type A viruses differ in their ability to infect and transform B cells We first compared the transforming potential of the virus panel. To this end, we infected primary B cells from 5 independent peripheral blood samples and performed transformation assays by seeding 3 or 30 EBNA2-positive B cells per well 3 days after Indeglitazar infection and counted the number.