Presence of Epstein-Barr virus viral interleukin-10 in the serum of patients with non-human-immunodeficiency-virus-related diffuse large-cell non-Hodgkin's lymphomas.

Three series have reported increased serum interleukin-l 0 (ILIO) in patients with non-Hodgkin's lymphoma (NHL).'~4 In our initial report, total (ie, human and Epstein-Barr virus [EBV]) ILI O (h-vILI O ) was detectable in the serum of 47 of 101 (46%) patients with intermediateor high-grade NHL and correlated to overall and progression-free survival.' In the series by Stasi et al,'.' 42 of 105 (40%) patients with aggressive lymphoma were found to have detectable serum human IL-10. In a recent publication, serum human IL-IO (hlL-IO) was found increased in 30 of 52 (58%) patients with diffuse large-cell lymphomas (DLCL)." There are several differences in terms of serum IL-IO levels and their prognostic significance in these three series. ( I ) In our initial series' and in the series of Stasi et A,'.' respectively, 12 of 24 (50%) and 15 of 37 (40%) patients with DLCL were found to have serum IL-IO levels greater than 100 pg/mL, compared with 5 of 52 (10%) in the report by Cortes et al.' (2) Serum L 1 0 correlated to progression-free and overall survival in univariate and multivariate analysis in our initial report.' In the report by Stasi et al;,' serum IL-IO was found to be an independent prognostic factor to the risk of failure of induction treatment in a multivariate analysis in a series of IO5 patients treated according to the same regimen. In the report by Cortes et al; serum IL-IO was not correlated to progression-free or overall survival in a cohort of 52 patients with DLCL treated with the same regimen. We have analyzed the prognostic value of serum IL10 in a larger series of patients with DLCL and obtained results that may help us to understand the discrepancies between these results. Serum IL-IO levels were determined in a series of 73 previously untreated human immunodeficiency virus (HIV)-negative patients with DLCL who received regimens combining cyclophosphamide, doxorubicin. vincristinhindesine, bleomycine, and methotrexate between 1984 and 1994.' Using the Shering-Plough (SP) enzyme-linked immunosorbent assay (ELISA),?.' which detects both hIL-10 and EBV viral (vIL-IO), serum h-vIL-l0 was detectable (ie, > 100 pg/mL) in 23 of the 73 (32%) patients of this series. Serum hIL-IO levels were then measured with the immunoassay used in the report by Cortes et ala (Cytoscreen; Biosource International, Camarillo, CA) that is specific for hIL-IO in 40 patients of this series. With this assay, only I (2.5%) patient had serum IL-IO levels greater than 100 pglmL, compared with I O (25%) with the SP assay in the same group (Fisher's exact test, P = .006). Using a regression analysis with a linear model, the values of IL-10 measured with the two assays were not found to be significantly correlated ( I = .29, P = . l I ) . The serum levels of 40 patients were then tested with a third immunoassay (level of sensitivity, 3 pg/mL) that recognizes both human and viral IL-10 (IL-10 EIA; Dianova, Immunotech, Marseille, France). With this assay, 9 (23%) patients had serum IL-IO levels greater than 100 pg/mL compared with 10 (25%) with the SP assay (Fisher's exact test, P > .9). Using a regression analysis with a linear model, the values of serum IL-10 measured with the Dianova and the SP assays were found to he highly correlated ( r = .69, P = ,00001). Serum IL-IO levels measured with the Dianova and Cytoscreen assay were not found to he correlated ( r = .19, P = 26) . The correlation observed between the two assays recognizing both hIL-IO and vIL-IO (Shering-Plough and Dianova) as well as the discrepancies with the assay specific for hIL-IO (Cytoscreen) suggested that vIL-l0 could be present in the serum of these patients. vIL-l0 is a cytokine and the product of an EBV gene termed BCRFI." vlL-l0 is produced by EBV lymphoblastoid cell lines in vitro" and expresses biologic properties that are largely similar to those of hIL-IO, in particular concerning its immunosuppressive properties and B-cell growth factor Serum vlL-l0 levels were measured using a specific ELISA that was previously reported.' The limit of sensitivity of this ELISA is 50 pg/mL. hIL-IO at I O pg/mL does not react with the vIL-IO ELISA.' Nonspecific positive serum samples were excluded as in the initial reports.'.' Nine of the 73 (12%) patients with DLCL ( 1 T-NHL, 6 B-NHL, and 2 undetermined) were found to have detectable vIL-l0 in serum (Fig I ) , ie. 39% of the 23 patients with detectable h-vIL-l0 with the SP assay. In contrast, vIL-l0 was undetectable in the serum of the 42 normal controls tested (Fisher's exact test, P = .02) and of 44 patients with nonhematologic tumors (metastatic melanoma In = 271 or renal cell carcinoma [n = 171. Fisher's exact test, P = .02). Viral L 1 0 was detectable in none of the 27 patients with DLCL in complete remission (CR) nor in the 14 patients in partial remission (PR). vIL-IO was detectable in 3 of 9 (33%) patients with untreated DLCL in relapse, indicating that serum vILI O is detectable only at an active phase of the disease with this assay (Fig 1). All patients with detectable vIL-l0 had a positive EBV serology. In this series, the percentage of patients with detectable h-vIL-l0 or vlL-l0 was not significantly different according to stage, lactate dehydrogenase (LDH) levels, age, performance status, 8 2 microglobulin levels, or the presence of B symptoms ( P > .2 by x' analysis). These results indicate that vlL-IO. the product of an EBV gene, is produced in vivo in a subset of patients with DLCL at an active phase of their disease.