203 patients with epithelial FIGO stage II to IV ovarian cancer from OVCAD (FP6 EU-project, Ovarian Cancer: Diagnosis of a silent killer, no. 018698,
http://www.ovcad.eu) were included in the study. Patients have been recruited from 2005 to 2008 in the Department of Gynecology at Charité, Campus Virchow-Klinikum, Medical University Berlin, Germany (64); Department of Obstetrics and Gynecology and Gynecologic Oncology, University Hospital Leuven, Belgium (54); Department of Gynecology, University Medical Center Hamburg-Eppendorf, Germany (38); Department of Obstetrics and Gynecology, Medical University of Vienna, Austria (37); Department of Gynecology and Obstetrics, Innsbruck Medical University, Austria (10). Informed consents were obtained from all patients. All processes were approved by the respective local ethical committee (EK207/2003, ML2524, HEK190504, EK366, EK260). Patients were treated with cytoreductive surgery and platinum-based chemotherapy. Tumor tissues were obtained during primary surgery and before any chemotherapeutic treatment. Residual tumor load was defined as negative if macroscopically absent. Overall survival (OS) was defined as the time from diagnosis to death from any disease-related cause. The survival times of patients alive at their last follow-up visit were treated as censored. Progression-free survival (PFS) was defined as the time from diagnosis until progression of disease, censoring patients who were recurrence-free at their last follow-up visit, and excluding patients with refractory disease, defined as progression of disease while receiving first line platinum-based therapy or within four weeks of the last platinum application [
20]. Progression was defined by radiological diagnosis according to the RECIST criteria or as doubling of the nadir serum CA-125 [
21]. Experienced gynecological oncologists and pathologists performed the clinical and histopathological evaluations.
Immunohistochemistry
Antigen heat retrieval was performed by microwaving the slides for 15 minutes in EDTA (1 mM, pH 8.0) and Dako Target Retrieval Solution (Dako, Denmark) for the CD8 and Ki67 staining, respectively. Slides were then cooled to room temperature and quenched for endogenous peroxidase. Blocking solution (Ultra V Block; TA-015HP, Thermo Fisher Scientific, USA) was applied prior to incubation with monoclonal mouse antibodies against CD8 (1:1000; clone C8/144B, code M 7103, Dako, Denmark) and Ki67 (1:75, clone MIB-1, code M7240, Dako, Denmark) overnight at 4°C. UltraVision detection system (Thermo Fisher Scientific, USA) and the Dako LSAB System (Dako, Denmark) were used for CD8 and Ki67, respectively, according to the manufacturers’ instructions. The CD8 stained sections were incubated with Primary Antibody Enhancer (TL-015-PB, Thermo Fisher Scientific, USA), followed by horseradish peroxidase (HRP) Polymer (HRP Polymer; TL-015-PH, Thermo Fisher Scientific, USA); for Ki67 staining, Dako Biotinylated Link (K0675, Dako, Denmark), followed by Dako Streptavidin-HRP (K0675, Dako, Denmark) was applied. The slides were stained with diamino-benzidine (DAB, 1:50 in DAB Substrate Buffer, K0673, Dako, Denmark) and counterstained with hematoxylin. Lymph node and normal colon tissue specimens were used as positive controls for CD8 and Ki67, respectively. Mouse IgG1 (1:75; Negative Control Mouse IgG1, code X0931, Dako, Denmark) was used as isotype control.
Slides were digitally photographed with the TissueFAXS system (version 2.0.4.0147, TissueGnostics, Austria) using an x20 objective lens. HistoQuest software (version 3.0.3.0161, TissueGnostics, Austria) was used for the detection and quantification of CD8+ cells. The TissueFAXS detection and quantification method was recently applied in other human cancer entities [
23] and is based on techniques described and validated by Steiner et al. [
24]. The CD8+ cell density (cells/mm
2) was determined only in epithelial tumor tissue. To avoid false positive cell counting, a specific gate according to cell size and intensity of CD8 staining was set within which the cells were considered positive. These cells were randomly controlled by applying a function permitting the visualization of the corresponding cells on the digital picture.
Scoring of the Ki67 proliferation index was evaluated manually by two independent observers determining the percentage of Ki67+ cells in total epithelial tumor tissue.
Tumor tissues from 158 out of the 203 patients were available. About 30 mg fresh frozen tumor tissue was homogenized using a Mikro-Dismembrator U (B. Braun, Biotech International, Germany) and lysed in one ml Nucleic Acid Purification Lysis Solution (Applied Biosystems, Life Technologies, USA). Total RNA from the lysates was isolated with the ABI PRISM 6100 Nucleic Acid PrepStation (Tissue RNA isolation, Applied Biosystems, Life Technologies, USA) and quantified spectrophotometrically. The quality of RNA was assessed with an Agilent 2100 Bioanalyzer. RNA with an RNA Integrity Number >5 was used.
The cDNA synthesis was performed with the Omniscript Reverse Transcription Kit (QIAGEN, Netherlands) with 500 ng RNA according to manufacturer’s instructions. cDNA was diluted 1:2 with TE buffer and deposited at −80°C for further analysis. The qPCR was performed with the CD8A TaqMan Gene Expression Assay (Hs00233520_m1, Applied Biosystems, Life Technologies, USA) according to the manufacturer’s instructions. As a reference, gene expression of house-keeping gene GAPDH (Hs99999905_m1, Applied Biosystems, Life Technologies, USA) was measured. 2 μl cDNA, 0.4 μl TaqMan Gene Expression Assay, 4 μl 2x TaqMan Gene Expression MasterMix (Applied Biosystems, Life Technologies, USA) and 1.6 μl H2O were used. The reaction mixture was pre-incubated at 50°C for two minutes and at 95°C for ten minutes, followed by 40 cycles of two step incubation at 95°C for 15 seconds and at 60°C for one minute. Each PCR was performed in duplicates.
Statistical analyses
Raw CD8+ cell density values were log
2-transformed to achieve normal distribution. For each patient, the mean value of the two cores was calculated. To evaluate the RT-qPCR data, the mean value of the duplicate expression values (Ct values) was calculated and normalized with the mean Ct value of the reference gene GAPDH. Differences between plates were corrected with a calibrator. Finally the normalized values were multiplied by −1 to be interpretable as log
2-expression (relative expression values). Statistical analyses were performed with SPSS (version 19, Chicago, USA) and SAS (version 9.3, 2011 SAS Institute Inc., Cary, NC, USA). Correlation of continuous variables (CD8+ cell density and relative expression values, Ki67 proliferation index and age) was assessed by Pearson’s correlation coefficients. Continuous variables were compared between groups by ANOVA or T-test. The association of categorical variables was evaluated by Fisher’s exact tests. Cumulative survival probabilities were calculated by the Kaplan-Meier method. Univariate and multivariable Cox proportional hazards regression analysis was used to evaluate the marginal and adjusted association of CD8+ cell density, percentage of Ki67 proliferation index, CD8 relative expression values and the clinicopathological factors age, FIGO stage and residual tumor with survival [
25]. For multivariable regression, the multivariable fractional polynomial approach was used, which evaluates possible non-linear effects of continuous variables such as CD8+ cell density or Ki67 by a set of parsimonious transformations [
26]. Because of the relatively low number of patients with FIGO II (9), we modeled FIGO stage as ordinal rather than categorical variable, assuming the same hazard ratio between FIGO IV and III as between FIGO III and II. This strategy provides more stable results than with categorical modeling of FIGO stage. Pairwise interactions between CD8+ cell density and other variables were tested by assessing significance of corresponding product terms. Two-sided p values <0.05 were considered statistically significant in all the analyses.