Prognostic significance of Hypoxia-Inducible Factor 1 alpha(HIF-1alpha) expression in serous ovarian cancer: an immunohistochemical study

One hundred (n = 100) neoplastic and 20 benign (controls) pathological samples from paraffin-embedded tissue were included. The explorative laparotomies, chemotherapies and the treatment of patients included in this trial was carried out in the University Hospital of Larissa-University of Thessalia Greece. All the necessary informed consents as requested by the Institutional Review Board (IRB) of the University Hospital were given. They were classified after surgery as stage I (n = 23) and stage III (n = 55), G3 and 22 borderline serous adenocarcinoma patients and 20 benign controls. The mean follow up was 3 years. At the time of the analysis, 13/55 were deceased due to their disease. Due to the short follow up all stage I and borderline patients were alive at the time of the analysis. Fifty five stage III pathological samples from paraffin-embedded tissue were included in the survival analysis. Only patients with the diagnosis of serous carcinoma of stage III, G3 who received 6 cycles of postoperative TC (175–180 mg/m2 paclitaxel and carboplatin after calculating the area under the concentration curve (AUC) with complete medical records were selected for survival analysis. All above-mentioned samples were obtained one per patient. Clinicopathologic information was obtained from medical records.

Cancer patients were classified after a staging laparotomy was performed (the most common initial surgical procedure consisted of abdominal hysterectomy, bilateral salpingo-oophorectomy, omentectomy, and lymph node sampling). The surgery was classified as complete when < 1 cm of tumor was left behind.

All slides were reviewed by two pathologists (G.K, M.I). The surgical procedures were carried out in the Department of Obstetrics and Gynaecology University of Thessalia.

In order for a sample to be given a "positive" final score convincing and easily detectable nuclear staining should be seen with at least 3 out of the 5 HIF-1α antibodies tested. Otherwise it was classified as negative.

The stage, grading, histology, age, family history, Ca 125 and level of cytoreduction (complete versus incomplete) were reported.

Different regimens were given as second- and third-line therapy, and the different groups (after further classifying according to second-line chemotherapy) had small numbers to be analyzed separately. Therefore, although both the progress-free interval (PFI) and the overall survival were calculated, we report PFI after first-line chemotherapy (TC) as indicative of the response to the first-line chemotherapy. The survival analysis of the samples compared two groups after the patients were dichotomized by HIF-1α final score to positive and negative.

Immunostaining was performed with the antibodies listed in Table 1. The tissue samples had all been fixed in 10% buffered formalin, processed and embedded in paraffin routinely. Sections were cut at 3 μm using a Leica TP1020 microtome and dried overnight at 60°C. After deparaffinization in xylene, the sections were rehydrated in decreasing ethanol solutions and incubated in 0.3% hydrogen peroxide for 10 min, to block endogenous peroxidase.

Different methods of antigen retrieval were tested in pilot experiments (data not shown). Under the conditions of the study, optimal antigen retrieval was achieved by microwaving tissue sections in 0.01 M citrate buffer solution (pH 6) for 20 min, (LG WAVEDOM, 850 Watt). This antigen retrieval remained optimal irrespectively of the primary antibody type applied. After the antigen retrieval, the sections cooled and washed in phosphate-buffered saline (PBS) for three times. Tissue sections were incubated overnight at 4°C with each antibody. The optimal dilutions were determined with pilot experiments (data not shown) see Table 1. Then, the slides were washed in PBS and Envision fluid (polymer-peroxidase method, EnVision+/HRP, DAKO, Denmark) was added, followed by incubation for 30 min. Bound antibodies were visualized by using 0.05% 3,3'-diaminobenzidine solution (DAB solution, DAKO). Finally, sections were counterstained with hematoxylin and mounted in Entellan (Merck, Germany). A detailed description of the University of Thessaly antibody has been included in a recently published HIF-1α separate study [11].

Group comparisons were based on Spearman's test for nominal variables (age, Ca125) and on the linear x²-test for ordinal variables (stage, grading, complete/incomplete cytoreduction). The HIF-1α scores of the different groups of patients were compared using the non-parametric test for multiple comparison (Kruscal Wallis-Mann Whitney) followed by Dunn's test, which generalizes the Bonferroni adjustment.

The factors possibly influencing the PFI and survival (age, Ca 125, stage, HIF-1α scores) were determined by binary logistic regression analysis using forward likelihood ratio method.

From 55 poorly differentiated serous ovarian carcinomas, 33 tumour samples were classified as HIF-1α protein "positive" (60%%) and 22 as HIF-1α protein "negative" as they did not stain at all (40%). After the patients were dichotomized by HIF-1α expression final score as positive or negative (used as gold standard) Cohen's kappa κ statistics were used to evaluate the measure of agreement kappa (κ value) for HIF-1α protein detection between the different antibodies.

Uni-variable and multivariable analyses were performed. The covariates age, Ca 125, HIF-1α scores were used for the multivariable analysis (in the 55 stage III, G3, serous adenocarcinoma patients). Progression-free survival and overall survival were estimated by Kaplan and Meier's method. The log-rank test was used to compare differences between survival curves. The Cox regression was used to calculate hazard ratios [12].

These were materialized using GraphPad Prism version 5 and SPSS version 15 software. All P values calculated are two sided. P < 0.05 was considered to be significant.

In the 55 patients with stage III, G3 serous carcinomas, the overall survival of patients with tumours that stained strongly for HIF-1α differed significantly than that of patients with tumours that stained weakly or were negative for HIF-1α (p < 0.05). HIF-1α positive patients displayed a median survival of 28 months (18–43 months) versus 39 months (range 15–73 months) for HIF-1α negative patients (Figure 2). Additionally, Kaplan-Meier survival curves confirmed that HIF-1α positive stage III, G3 patients had decreased overall survival compared to HIF-1α negative patients (p < 0.01). Cox regression analysis demonstrated that HIF-1α protein had a hazard ratio (HR) of 3.853 (1.544 to 9.614) (Figure 2). Additionally increased HIF-1α protein expression was an independent adverse prognostic factor for survival (see multivariable analysis in Table 2, p < 0.01).

HIF-1α positive patients displayed a median PFI of 19 months (range 6–67 months) versus 20 months (8–42 months) in HIF-1α negative patients (not statistically significant, p > 0.05). Additionally, Kaplan-Meier survival curves showed that HIF-1α protein positive stage III, G3 cases had a decreased overall PFI compared to HIF-1α negative patients but this was not statistically significant (p > 0.05). Cox regression analysis demonstrated that HIF-1α had an HR of 1.185 (0.5617 to 2.499). Increased HIF-1α protein expression was not an independent adverse prognostic factor for PFI (see multivariable analysis in Table 2, p > 0.05).

We performed a separate analysis in patients (n = 21) that have undergone suboptimal cytoreduction at primary surgery. The overall PFI of this subgroup of patients with tumours that stained strongly for HIF-1α differed significantly than that of patients with tumours that stained weakly or were negative for HIF-1α (p < 0.05). HIF-1α positive patients displayed a median PFI of 12 months (range 6–24) versus 19.5 months (12–42) in HIF-1α negative patients. Additionally, Kaplan-Meier survival curves showed that HIF-1α positive stage III, G3 patients had a decreased overall PFI compared to HIF-1α negative patients, and this was statistically significant (p < 0.05) (Figure 3). Cox regression analysis demonstrated an HR of 2.88 (1.007 to 8.37) for HIF-1α. Increased HIF-1α expression was an independent adverse prognostic factor for PFI for these group of patients (see multivariable analysis in Table 2, p = 0.05).

The group of HIF-1α positive samples (n = 33) were further subdivided according to the presence of nuclear or cytoplasmic immunostaining. The patients with samples showing combined nuclear and cytoplasmic HIF-1α staining (n = 14) displayed a median survival of 38 months (24–73) and a median PFI of 18 months (6–28). The ones without cytoplasmic staining and with nuclear only staining (n = 19) displayed a median survival of 39 months (15–69) and a median PFI of 24.5 months (6–67). When these groups were compared, they did not differ significantly for PFI (p > 0.05) or survival (p > 0.05).

We used Cohen's κ statistics to compare the value of each of the antibodies as probes capable of categorizing each case as expressing (i.e. positive) or not expressing (negative) nuclear HIF-1α. As reported in the Methods section, a "positive" designation required easily-detectable immunoreactivity with three out of five antibodies. A sample showing nuclear immunoreactivity using one or two antibodies would be designated as "negative". A high κ value compared to HIF-1α positive patients would indicate that a given antibody performed better in detecting positive cases when compared with lower κ value antibodies. The κ values (95% confidence intervals) for the different antibodies are presented in ranking order in Table 3. The antibody with the lowest kappa value was H1a67-Abcam. Birner et al. used this antibody in his IHC study and reported that HIF-1α overexpression had no impact on the prognosis [6]. To verify this, we performed a separate survival analysis using only the findings from this antibody (Figure 4) and the survival analysis was not significant, as Birner et al. had reported [6].