MICA/B and ULBP1 NKG2D ligands are independent predictors of good prognosis in cervical cancer

The study subjects were comprised of 200 cervical cancer and 426 cervical intraepithelial neoplasias (CINs) patients who underwent surgical resection at Gangnam Severance Hospital, Yonsei University College of Medicine between March 1996 and March 2010. Additional paraffin blocks were provided by the Korea Gynecologic Cancer Bank through the Bio & Medical Technology Development Program of the Ministry of Education, Science and Technology, Korea. All patients had a histological diagnosis of cervical carcinoma or CIN, and the cervical cancer patients were clinically staged according to the International Federation of Gynecology and Obstetrics (FIGO) staging system. Patients with cervical cancer underwent type 3 radical hysterectomy with pelvic lymph node dissection, and, in cases of increased risk of relapse (assessed from spread to lymph node, parametrial invasion, and cancer close to resection margins), platinum-based concurrent chemoradiation was added. Medical records were reviewed to obtain data including age, Hybrid Capture^® 2 results, surgical procedure, survival time, and survival status. Response to therapy was assessed by either computed tomography or magnetic resonance imaging in accordance with the Response Evaluation Criteria in Solid Tumors (RECIST, version 1.0) [17]. Data on tumor size, cell type, tumor grade, and lymph node metastases were obtained from pathology reports. Tissue samples were collected from patients who had signed informed consent forms, which was approved by the Institutional Review Boards of Gangnam Severance Hospital. This study was additionally approved by the Office of Human Subjects Research at the National Institute of Health.

Tissue cores, from formalin-fixed, paraffin-embedded tissue blocks obtained from 626 patients with primary invasive cervical cancer or CIN and 541 matched nonadjacent normal cervical epithelia were arrayed into a recipient paraffin block with a manual tissue arrayer MTA-1 (Beecher Instruments Inc., Silver Spring, MD). These normal cervical tissue cores were obtained from the same block at locations distant from the cancer cells or from different block that contained enough normal epithelial cells for further IHC analyses. For each case, a representative tumor area was carefully selected from a hematoxylin and eosin (H&E)-stained section of the donor block. Four 1.0-mm-diameter cores consisting of matched tumor specimen and normal epithelial samples were retrieved from selected regions of a patient’s donor block. The presence of tumor tissues on the tissue microarray (TMA) was verified with H&E staining. At every 50th section, multiple 5-μm-thick sections were cut with a microtome and H&E staining of TMA slides were examined for the presence of tumor cells.

Human cervical cancer cell lines were obtained from two sources: C-33A, CaSki, HeLa, ME-180 and SiHa from American Type Culture Collection (ATCC, Manassas, VA) and SUN-17 from Korean Cell Line Bank (KCLB, Seoul, Korea). CaSki, HeLa, ME-180, and SUN-17 cells were cultured in vitro in RPMI 1640 while C-33A and SiHa cells were cultured in DMEM (Dulbecco’s modified Eagle’s medium). Both media were supplemented with 10% fetal bovine serum, 50 units/ml of penicillin/streptomycin, 2 mM L-glutamine, 1 mM sodium pyruvate, and 2 mM non-essential amino acids, and cells were grown at 37°C with 5% CO₂.

For in vitro flow cytometry analysis, 2 × l0⁵ tumor cells were incubated with 0.5 μg recombinant human NKG2D Fc chimera (R & D systems, Minneapolis, MN) and then PE-conjugated anti-human Fc secondary antibody was used as a detection antibody (BD Bioscience, San Jose, CA). CELLQuest software (Becton Dickinson Immunocytometry System, Mountain View, CA) was used for FACScan analysis.

Immunohistochemical staining of MICA/MICB, ULBP1, ULBP2, ULBP3, RAET1E, and RAET1G was performed by using streptavidin-biotin peroxidase method. Prior to applying IHC with TMA section, we tested whole section for immunohistochemial staining condition. Briefly, the TMA sections were deparaffinized by xylene and then rehydrated through a descending alcohol gradient. Endogenous peroxidase activity was blocked by 10 min of incubation in 3% H₂O₂. To retrieve antigenicity, sections were immersed in antigen retrieval buffer, pH 9 (Dako, Carpinteria, CA), and heated for 20 min in a steam pressure cooker (Pascal, Dako). Slides for ULBP1 and ULBP2, however, were heated for 10 min instead of 20 min at high pressure. To block nonspecific staining, sections were treated with Protein Block (Dako) for 20 min. The sections were incubated with anti-MICA/MICB antibody (Novus Biologicals, Littleton, CO; mouse monoclonal antibody, Clone 6D4, 1:50 for 120 min), anti-ULBP1 antibody (Sigma-Aldrich, St. Louis, MO; rabbit polyclonal antibody, Cat.# HPA007547, 1:50 for 120 min), anti-ULBP2 antibody (R&D systems; goat polyclonal antibody, Cat.# AF1298, 1:50 for 30 min), anti-ULBP3 antibody (Abcam, Cambridge, MA; mouse monoclonal antibody, Clone MM0594-6E12, 1:500 for 120 min), anti-RAET1E antibody (Novus Biologicals; mouse polyclonal antibody, Cat.# H00135250-B01, 1:2000 overnight at 4°C), and anti-RAET1G antibody (Novus Biologicals; mouse polyclonal antibody, Cat.# H00353091-B01, 1:1000 overnight at 4°C) in a Dako autostainer plus (Dako). Different systems were used for the detection of primary antibodies: Dako EnVision + Dual Link System-HRP (Dako) for MICA/MICB and ULBP3; Dako EnVision FLEX+ (Dako) for ULBP1, RAET1E, and RAET1G; and Dako LSAB^®2 System-HRP (Dako) for ULBP2. The stain was visualized using DAB⁺ kit (3,3’-Diaminobenzidine; Dako) and then lightly counterstained with hematoxylin. The slides were covered and observed under a light microscope (Axioplot, Carl Zeiss, Jena, Germany). Negative controls were processed by omitting the primary antibodies, and TMAs included colorectal cancer positive control tissues [19].

For the assessment of NKG2DL staining, two scores were assigned to each core: (a) the staining intensity (no evidence of staining, 0; weak staining, 1+; moderate staining, 2+; and strong positive staining in most cells, 3+) and (b) the percentage of positively stained epithelial cells (no cells staining positive, 0; less than 25% of cells staining positive, 1+; 25–50% of cells staining positive, 2+; 50–75% of cells staining positive, 3+; and more than 75% of cells staining positive, 4+). An overall protein expression score was calculated by multiplying the intensity and positivity scores (overall score range, 0–12). The IHC staining score was then dichotomized into low expression (≤ mean score of cancer specimens) and high expression (> mean score of cancer specimens). Slides were scored without any clinical information, and the final immunostaining score reported was the average of two independent pathologists, both with experience in the analysis of tissue microarray.

Table 1 summarizes patient clinicopathologic characteristics. The overall mean age of patients was 40.6 ± 9.8 years for low-grade CIN, 38.9 ± 11.2 years for high-grade CIN, and 49.4 ± 11.7 years for cervical cancer. The distribution of FIGO staging for the 200 cases of cervical cancer is as follows: 138 stage I, 53 stage II, and 9 stage IV. The following cell types were assigned according to World Health Organization (WHO) criteria: 164 squamous cell carcinomas, 30 adenocarcinomas/adenosquamous carcinomas, 4 small cell carcinomas, 1 neuroendocrine, and 1 clear cell carcinomas. HC2-confirmed HPV infection rate was 78.7% (74/94) in low-grade CIN, 92.2% (226/245) in high-grade CIN, and 93.9% (92/98) in cervical cancer.

Expression of NKG2DLs was determined by flow cytometric assay using recombinant human NKG2D-Fc chimera protein in cell cultures prior to IHC analysis for individual NKG2DL expression in cervical cancer tissues. NKG2D-Fc chimera protein binds to NKG2D through ligand-receptor interaction on the cell surface, and it is detected by anti-Fc antibody conjugated with fluorophores. As a result, using NKG2D-Fc chimera protein in flow cytometric assay allows us to confirm the expression and binding ability of NKG2DLs even without a characterization of ligands. As shown in Figure 1, expression of NKG2DLs was determined in six different cervical cancer cell lines (C-33A, CaSki, HeLa, ME-180, SiHa and SUN-17). Five cancer cell lines (all but C-33A) expressed NKG2DLs, which could bind to NKG2D, supporting the hypothesis that cervical cancers express NKG2DLs in vivo.

We then performed IHC analysis of MICA/MICB, ULBP1, ULBP2, ULBP3, RAET1E, and RAET1G in 200 cervical cancer specimens, 327 high-grade CINs, 99 low-grade CINs, and 541 matched nonadjacent normal cervical epithelial tissue samples. Representative immunohistochemical expression of individual NKG2DLs are presented in Figure 2. ULBP3 was expressed exclusively in the nucleus in both tumor and normal epithelial cells, while the other markers were expressed primarily in the cytoplasm, with some cases also demonstrating weak nucleus staining (Figure 2). Scoring results from the IHC analyses are summarized in Table 2. The TMA contains 200 cases of cervical cancer, however due to the complexity of sectioning and staining, between 180 and 195 samples could be interpreted for the individual marker. Invasive cervical cancer tissues had higher MICA/B, ULBP1, and RAET1E expression than CIN or normal cervical epithelial tissues (all p < 0.001). This trend of progressively increasing MICA/B, ULBP1, and RAET1E expression corresponded to the phases of cervical cancer progression and was significant according to Spearman’s rank correlation (ρ-value of 0.313 [p < 0.001], 0.285 [p < 0.001], or 0.136 [p < 0.001], respectively). ULBP2 and ULBP3 expression, on the other hand, was higher in low-grade CIN than in normal epithelium but gradually decreased in high-grade CIN and cervical cancer (p = 0.001 and p = 0.017, respectively).

To determine the association between expression of MICA/MICB, ULBP1, ULBP2, ULBP3, RAET1E, and RAET1G, Spearman’s rank correlation analysis was performed (Table 3). Cervical cancer lesions were evaluated for co-expression between individual NKG2DLs. MICA/B expression was significantly correlated with the expression of ULBP1 (Spearman’s rho = 0.677, p < 0.001) or ULBP2 (Spearman’s rho = 0.257, p < 0.001), whereas MICA/B expression was not significantly correlated with those of ULBP3 (Spearman’s rho = 0.119, p = 0.112), RAET1E (Spearman’s rho = 0.044, p = 0.555), or RAET1G (Spearman’s rho = -0.042, p = 0.574).

Finally, to investigate the prognostic significance of the expression of individual NKG2DLs in cervical cancer, we studied the correlation of NKG2DL expression with overall and disease-free survival. Clinicopathologic and outcome information were available for all 200 cervical cancer patients who were monitored for survival and recurrence. Kaplan-Meier plots demonstrated that patients with high MICA/B (Log-rank p = 0.027) or ULBP1 (Log-rank p = 0.009) expression and low RAET1E (Log-rank p = 0.018) or RAET1G (Log-rank p = 0.029) expression had significantly longer disease-free survival time (Table 4, Figure 3A and B). When analyzed individually for effect on overall survival, high ULBP1 expression predicted significantly longer survival time (Log-rank p = 0.007) (Table 4, Figure 3D and E). In particular, when survival of patients with expression of high MICA/B/high ULBP1 was compared with those of other patients, Kaplan-Meier analysis revealed a significant difference in both disease-free (p < 0.001, Figure 3C) and overall survival (p = 0.001, Figure 3F). A Cox multivariate proportional hazards analysis showed that advanced stage (hazard ratio = 3.60 [95% CI, 1.36–9.55], P = 0.010) and lymph node metastasis (hazard ratio = 2.71 [95% CI, 1.08–6.79], p = 0.032) were related to poor disease-free survival, whereas high ULBP1 (hazard ratio = 0.31 [95% CI, 0.11–0.86], p = 0.024) and high MICA/B/high ULBP1 (hazard ratio = 0.16 [95% CI, 0.13–0.70], p = 0.015) expression was related to good disease-free survival (Table 2). When analyzed for effect on overall survival, advanced stage (hazard ratio = 2.77 [95% CI, 1.13–7.76], p = 0.025) and high ULBP1 (hazard ratio = 0.27 [95% CI, 0.07–0.97], p = 0.044) expression were significant independent prognostic factors in multivariate analysis.