Tissue microarrays were constructed as described previously 3940. The TMAs contained a total of 302 tissue samples from tumours, which had been classified prior to this study as diffuse large B-cell lymphomas according to the WHO classification 1. All samples had been obtained at the time of diagnosis, before any treatment had been given. Four different TMAs were constructed, each containing tumour samples from different histopathologic institutions (Basel, Bologna, Innsbruck, Zurich). Clinical data of patients with DLBCL at time of primary diagnosis and follow-up had been obtained by reviewing the charts. Clinico-pathologic data of patients at time of diagnosis are detailed in table 1. Retrieval of tissue and clinical data was performed according to the regulations of the local institutional review board and data safety laws.

Four-micrometre sections of the TMA blocks were cut to adhesive-coated slides (Instrumedics Inc, Hackensack, NJ, USA) and stained using standard procedures. Briefly, immunohistochemical studies utilized an avidin-biotin peroxidase method with diaminobenzidine chromatogen. After heat induced antigen retrieval (microwave oven for 30 min at 250 W) immunohistochemistry was carried out in a NEXES immunostainer (Ventana, Tucson, AZ). The following primary antibody was used: BM28 (mouse monoclonal, clone 46, BD Biosciences, San Jose, US, final dilution 1:3000). The dilution had been established using adequate controls. Negative controls were obtained by omitting the primary antibody. The slides were evaluated without knowledge of clinical data. At least 10% of cases were re-evaluated by a second observer. Only cases containing clearly recognisable tumour tissue were analysed. One-hundred cells were assessed in each tumour core and the percentage of positive cells (i.e. cells with a distinct staining) was calculated. If a biopsy core contained less than 100 cells as many neoplastic cells as possible were evaluated. If more than one core from the same tumour was available the results were averaged. Cases were defined positive for Mcm2 if ≥40 % of tumours cells showed distinct nuclear staining. The determination of cut-off levels was based on the analysis of the area under the receiver operating characteristic curve (AUROC) as described below and as detailed in table 2.

Statistical analyses were performed using SPSS version 10.0 (SPSS, Chicago, IL). Differences were considered statistically significant if P values were <0.05. Receiver operating characteristic (ROC) curves by plotting sensitivity versus (1-specificity) were used to evaluate the diagnostic performance of parameters at various cut-off points. An AUROC closer to 1 indicates greater discriminatory power, whereas an AUROC of 0.5 denotes no diagnostic potential. The optimum cut-off was calculated as the maximum value of sensitivity multiplied by specificity. Sensitivity and specificity were calculated according to standard formulas for the cut-off that represented the best discrimination derived from the ROC curves. The cut-off value representing the best discrimination derived from the ROC analysis was chosen for categorization of metric variables (LDH (lactate dehydrogenase) and expression of Mcm2). A statistical association between clinico-pathologic parameters and Mcm2 expression was tested using a two-sided Fisher's exact test. Disease specific survival (DSS) curves were calculated using the Kaplan-Meier method with significance evaluated by two-sided log-rank statistics. For DSS analysis, patients were censored at the time of their last clinical follow-up appointment or at their date of death not related to the tumour. A stepwise multivariable Cox regression model was adjusted, testing the independent prognostic relevance of Mcm2 positivity. The limit for reverse selection procedures was P = 0.1. The proportionality assumption for all variables was assessed with log-negative-log survival distribution functions.

Clinical follow-up data were available for 123 of 302 (40.7 %) patients with a median follow-up period of 23.5 months (range 1 to 177 months). The median follow-up for censored patients was 28.5 months. Clinico-pathologic parameters associated with poor disease specific survival were Ann Arbor stage (p = 0.0026), IPI (p < 0.0001), and serum LDH ≥300 U/l (p = 0.0133) according to the univariate analysis. The optimum cut-off level for LDH had been established at 306 U/l (AUROC method) and a cut-off of 300 U/l was used for further statistical analysis. Age at diagnosis and gender of patients did not make an impact on DSS. Neither the presence of bone marrow involvement, extranodal involvement nor bulky disease at presentation was statistically relevant for DSS.

The results of the univariate analysis are presented in table 3.

Investigation of Mcm2 expression was informative in 69.9 % (211/302) of cases. Mcm2 expression in ≥40% of cells was detected in 46.4 % (98/211) of DLBCL. Mcm2 was clearly evident in the nuclei of tumour cells with low background staining. In non-neoplastic lymphoid tissue, which was used as a positive control and for establishing the appropriate staining protocols, expression of Mcm2 was mainly found in the germinal centres harbouring proliferating cells; expression in the mantle zone and parafollicular zone was negligible (Figure 1A). Identical results have previously been described 41.

The optimum cut-off for Mcm2 expression was calculated as 39.5 % of cells using the AUROC method. For practical reasons, a cut-off level of 40 % was chosen and patients were divided into two subgroups accordingly (positive: protein expression ≥40 % of cells; negative: protein expression <40 % of cells). For descriptive data analysis, all relevant variables were compared with Mcm2 expression (table 4). A representative image of immunohistochemical expression of Mcm2 in DLBCL is shown in figure 1B.

Disease specific survival was compared between tumours positive and negative for Mcm2 by univariate log-rank statistics. Positivity for Mcm2 was associated with shorter DSS (p = 0.0424) (table 3; figure 2). However, when Mcm2 positivity was tested in the multivariate analysis in combination with LDH, IPI, and Ann Arbor stage it did not retain its prognostic value (p = 0.546). LDH was not found to be of prognostic relevance in the multivariate analysis as well (p = 0.063). Both IPI (p = 0.010) and Ann Arbor stage (p = 0.025) were significant prognostic markers in the multivariate analysis with respect to disease specific survival.