Subcellular Localization of p27^kip1 Expression Predicts Poor Prognosis in Human Ovarian Cancer Free

Human ovarian cancer is the fifth leading cause of death among women in the United States. The wide variability in response to treatment reflects the heterogeneity in tumor histotype and grade as well as the nonspecific nature of the symptoms associated with early-stage disease. The most consistently reported significant prognostic indicators for human ovarian cancer are disease stage, tumor grade, histotype, and extent of surgical cytoreduction. Although these factors are not always related to the biological behavior or aggressiveness of the disease, they are used as guidelines for selecting anticancer therapy (1). There is a wide spectrum of clinical behaviors from an excellent prognosis and a high likelihood of cure to those with rapid progression and poor prognosis irrespective of the clinical stage of the disease, most probably reflecting different biological properties of the tumors.

Cellular progression through the cell cycle is governed by cyclin-dependent kinase (cdk) that is regulated by phosphorylation, activated by binding of cyclins, and inhibited by cdk inhibitors. Based on their protein sequence homologies and putative cdk targets, cdk inhibitors belong to one of two families: the CIP/KIP family (p21^Waf1/Cip1, p27^kip1, and p57^kip2), which inhibits a broad range of cyclin/cdk complexes and the INK4 family (p15^Ink4b, p16^Ink4a, p18^Ink4c, and p19^Ink4d), which inhibit mainly cdk4 and cdk6. The coordinated expression of cyclins, cdks, and cdk inhibitors is often deregulated in cancer (2). The cdk inhibitor p27^kip1 regulates cellular progression from G₁ to S phase. p27^kip1 acts primarily by complexing with cyclins D1 and E, thereby inhibiting the function of these cdks. Several studies have shown that loss of p27^kip1 protein expression is associated with disease progression in various malignancies (3–5) and with poor prognosis in prostate and colon cancer (4, 6). In breast cancer, the appearance of p27^kip in the cytoplasm of tumor cells is associated with poor prognosis (3, 7). However, it is still controversial whether expression or loss of expression of p27^kip1 has any prognostic significance in human ovarian cancer (7–11). In addition, the prognostic significance of subcellular localization has not been previously investigated. In this paper, we addressed this question by evaluating the subcellular localization of p27^kip1 and its expression using tissue microarrays from 441 patients with ovarian cancer.

Patients. Samples from women with primary epithelial ovarian cancer who had undergone initial surgery at the University of Texas M.D. Anderson Cancer Center between 1990 and 2001 were included in this study. A total of 441 correlative patients were identified. Follow-up information was updated through June 2003 by reviewing medical records and the U.S. Social Security Index. Demographic and survival data were entered into a comprehensive database created with Microsoft Access (version 97). Histopathologic diagnoses were based on WHO criteria (12–16); tumor grade based on Gynecologic Oncology Group criteria (17) and each case was assigned a disease stage according to the International Federation of Gynecology and Obstetrics system (18). Disease-specific survival rates were calculated as the percentage of subjects who survived with disease for a defined period, reported as time since diagnosis or treatment, and only deaths from the disease were counted. The extent of cytoreduction was defined as optimal if residual disease after surgery was smaller than 1 cm or suboptimal if residual disease was larger than 1 cm. (19, 20). Use of tissue blocks and chart review was approved by appropriate institutional committee.

Nuclear and Cytoplasmic Extraction. Ten cases of primary ovarian carcinoma were selected for Western blot analysis. Frozen sections were prepared and evaluated before processing to ensure correct sampling of the tumor and that at least 200 μg of pure tumor tissue (>80% tumor) had been collected from each case. The tissue samples were then homogenized in 500 μL of homogenizing buffer [10 mmol/L HEPES, 1.5 mmol/L MgCl₂, 10 mmol/L KCl, 0.3 mmol/L sucrose, 0.1 mmol/L EGTA, 0.5 mmol/L DTT, 0.5 mmol/L phenylmethylsulfonyl fluoride, 1 mmol/L sodium orthovanadate, 1 μg/mL pepstain, 5 μg/mL leupeptin, 0.15 unit/mL aprotinin, 1 mmol/L sodium fluoride, and 0.1% NP40 (pH 7.9)] using a Polytron microprobe set at “7” for 30 seconds on ice. The product was centrifuged at 3,000 rpm for 20 minutes at 4°C, after which the supernatant was transferred to a new tube and ultracentrifuged again at 20,000 rpm for 2 hours at 4°C. The supernatant was transferred to another new tube and used for cytoplasmic fraction analyses. The pellet was washed thrice with homogenizing buffer, 50 μL of nuclear extract buffer was added [20 mmol/L HEPES, 420 mmol/L NaCl, 1.5 mmol/L MgCl₂, 0.2 mmol/L EDTA, 0.1 mmol/L EGTA, 25% glycerol, 1 mmol/L DTT, 0.5 mmol/L phenylmethylsulfonyl fluoride, 1 mmol/L sodium orthovanadate, 1 μg/mL pepstain, 5 μg/mL leupeptin, 0.15 unit/mL aprotinin, and 0.5 mmol/L spermidine (pH 7.9)], and the pellet was dispersed by gentle mixing for 1 hour on ice and centrifuged at 5,000 rpm for 10 minutes at 4°C. The supernatant was dialyzed against homogenizing buffer for 2 hours at 4°C. The resultant supernatant was transferred to a new tube and used for nuclear fraction analyses.

Western Blotting. Equal amounts of proteins (about 50 μg) were analyzed using standard methods for protein electrophoresis and transfer. The primary antibodies were to p27^kip1 (clone 57, BD PharMingen, San Diego, CA), c-myc (Zymed Biotech, San Francisco, CA), β-actin (Sigma Chemicals, St Louis, MO). The secondary antibodies were against mouse or rabbit IgGs (Amersham Biosciences, Piscataway, NJ). The detection reagents were from an electrochemiluminescence kit (Amersham Biosciences). The adequacy of nuclear and cytoplasmic extracts was confirmed by using antibodies against c-myc (for nuclear samples) and β-actin (for cytoplasmic samples).

Construction of the Tissue Microarrays. Tissue blocks were stored under ambient conditions, at ∼24°C. H&E-stained sections were reviewed by a pathologist to select representative areas of tumor from which to acquire cores for microarray analysis. Tissue microarray blocks were constructed by taking core samples from morphologically representative areas of paraffin-embedded tumor tissues and assembling them on a recipient paraffin block. This was done with a precision instrument (Beecher Instruments, Silver Spring, MD) that uses two separate core needles for punching the donor and recipient blocks and a micrometer-precise coordinate system for assembling tissue samples on a block. For each case, two replicate 1-mm core diameter samples were collected and each was placed on a separate recipient block. The final tissue microarray consisted of six blocks, the first pair (blocks 1a and b) containing duplicates of 158 spots, the second pair (2a and b) containing duplicates of 164 spots, and the third pair containing duplicates of 119 spots. All samples were spaced 0.5 mm apart. Five-micrometer sections were obtained from the microarray and stained with H&E to confirm the presence of tumor and to assess the tumor histology. Tumor samples were randomly arranged on the blocks.

Sample tracking was based on coordinate positions for each tissue spot in the tissue microarray block; the spots were transferred onto tissue microarray slides for staining. This sample tracking system was linked to a Microsoft Access database containing demographic, clinicopathologic, and survival data on each patient, thereby allowing rapid links between histologic data and clinical features. The array was read according to the given tissue microarray map, each core was scored individually, and the results were presented as the mean of the two replicate core samples. Cases in which no tumor was found or no cores were available were excluded from the final data analysis.

Immunohistochemical Analysis. The tissue microarray slides were subjected to immunohistochemical staining as follows. After initial deparaffinization, endogenous peroxidase activity was blocked by using 0.3% H₂O₂. Deparaffinized sections were microwaved in 10 mmol/L citrate buffer (pH 6.0) to unmask the epitopes. The slides were then incubated for 1 hour at room temperature using the same antibody against p27^kip1 used for western blot analysis (1:100, clone 57, BD PharMingen), next with biotin-labeled secondary antibody for 20 minutes, and finally with a 1:40 solution of streptavidin-peroxidase for 20 minutes. Tissues were then stained for 5 minutes with 0.05% 3′,3-diaminobenzidine tetrahydrochloride that had been freshly prepared in 0.05 mol/L Tris buffer at pH 7.6 containing 0.024% H₂O₂ and then counterstained with hematoxylin, dehydrated, and mounted. All of the dilutions of antibody, biotin-labeled secondary antibody, and streptavidin-peroxidase were made in PBS (pH 7.4) containing 1% bovine serum albumin. Colon carcinoma was used as a positive control. Negative controls were made by replacing the primary antibody with PBS. All controls gave satisfactory results.

The immunostained slides were reviewed by two pathologists, who followed the tissue microarray map to record a score for each sample. Each reviewer was blinded to the other's assessment and to clinicopathologic information. Scoring discrepancies (11 cases, 3%), were resolved by a third pathologist. p27^kip1 expression was graded semiquantitatively by the reviewers. The scoring system was based on the subcellular localization of the p27^kip1 negative, nuclear, or cytoplasmic (21). The cytoplasmic and nuclear stains were scored separately, not additively. Negative staining was defined as absence of cytoplasmic stain and <5% of positive nuclei. Cytoplasmic staining was scored on a three-point system based on intensity: negative (no stain), weakly positive (1+), and strongly positive (2+). When staining was present in the nucleus but not in the cytoplasm, the sample was scored as “nuclear staining only.” Nuclear staining was judged to be positive if >5% nuclei in the sample were stained and negative if <5% of the nuclei stained. Ten high-power fields were examined. Normal ovarian epithelial cells were used as a comparison for intensity and pattern of staining. The mean of the results from the two replicate core samples from each tumor specimen was considered for each case.

Statistical Analysis. Differences in proportions were evaluated by the χ² or Fisher's exact test as appropriate. Kruskal-Wallis test was used to compare multiple independent samples on the tissue microarray block containing normal ovary and the different ovarian tumors. Also, differences in expression levels between normal ovarian epithelial cells and the different ovarian tumors was calculated using Mann-Whitney U test. These results were adjusted for multiple comparisons and considered statistically significant at the P < 0.01 level. Disease-specific survival rates were calculated using the method of Kaplan and Meier and compared by the log-rank test. Cox proportional hazards regression models were used for multivariate analysis of survival. Statistica software was used for the statistical analysis (SAS Institute, SAS Language Reference, version 8, SAS Institute, Inc., Cary, NC, 1999). Results were considered statistically significant at the P < 0.05 level.

Patient Characteristics. The mean age of the 441 patients was 58.2 years (range, 20-96 years). With regard to surgical disease stage, 36 patients (8.1%) had stage I disease, 32 (7.2%) had stage II disease, 291 (65.9%) had stage III disease, and 82 (18.5%) had stage IV disease. The tumor histotype was serous carcinoma in 336 patients (76.2%), endometrioid in 41 (9.3%), clear cell in 18 (4.1%), mixed malignant mullerian tumor in 17 (3.9%), undifferentiated carcinoma in 12 (2.7), mucinous in 10 (2.3%), and transitional cell carcinoma in 7 (1.6%). The mean follow-up interval was 64 months (range, 1-120 months), and the overall survival rate at 5 years was 38%.

Validation of p27^kip Antibody. To determine the prognostic significance of p27^kip1 according to its subcellular localization, we first validated the antibody with Western blotting in nuclear and cytoplasmic extracts from 13 randomly selected cases of ovarian carcinoma (Fig. 1). Using the same antibody, we immunostain whole sections of the same 13 cases and compared the results with the Western blot. The Western blots showed high expression of p27^kip1 in the nucleus in three cases (cases 2, 9, and 13) and in both the nucleus and cytoplasm in the remaining 10 cases. Immunohistochemical analysis of whole sections from the same 13 cases showed similar results, the example photomicrographs of immunohistochemical staining for case 2 (nuclear only) and case 8 in nuclear and cytoplasmic extracts are shown in Fig. 1. In the five control cases (normal ovarian surface epithelium), only nuclear staining for p27^kip1 was observed (Fig. 1B , 3). Antibodies to c-myc (which is present only in the nucleus) and β-actin (which is present only in the cytoplasm) confirmed the proper preparation of the nuclear and cytoplasmic extracts.

P27^kip1 Subcellular Localization and Its Association with Disease Stage, Tumor Grade, Patient's Age, and Level of Cytoreduction. Of all 441 cases identified, 421 (95.4%) could be scored for p27^kip1 staining; the remaining cases were either lost during the sectioning procedure or did not meet the criteria for inclusion. Results from immunostaining of the microarrays, presented according to clinicopathologic characteristics of the patients, are shown in Table 1. p27^kip1 protein was observed in the nucleus only in 122 cases (29%), in both the nucleus and the cytoplasm in 240 cases (57%) and negative expression in 59 of the cases (14%). With regard to the cytoplasmic intensity staining, 96 cases showed weak intensity and 144 cases strong intensity. No difference was found in the association with clinical variables between the weakly and strongly positive cytoplasmic staining groups (data not shown). Therefore, additional analysis was done between negative, cytoplasmic and nuclear localization.

Subcellular localization of p27^kip1 was increasingly more common in later-stage disease (P < 0.03). Concomitantly, nuclear-only expression was more common in early-stage disease (Table 1). No difference was observed in tumor grade, histotype or patient's age at time of diagnosis between negative, cytoplasmic, or nuclear localization. Expression of p27^kip1 only in the nucleus correlated with optimal levels of cytoreduction (P = 0.03). In contrast, cases showing either negative expression or cytoplasmic localization of p27^kip1 had suboptimal cytoreduction.

p27^kip1 Subcellular Expression and Its Association with Disease-Specific Survival. Overall ovarian cancer–specific survival times and rates at 2 and 5 years are shown in relation to the subcellular localization of p27^kip1 protein in Table 2. Among all patients, the subcellular localization of p27^kip1 significantly influenced disease-specific survival (P < 0.0001; Fig. 2A). Patients with tumors that expressed p27^kip1 only in the nucleus had better survival, in terms of both rate and duration, than did patients with cytoplasmic or no p27^kip1 expression. Moreover, this difference was also evident within certain well-defined clinical groups belonging to the same clinical stage. Among 355 late-stage disease patients, those with nuclear only expression of p27^kip1 had a better overall survival than those with negative or cytoplasmic localization of the marker (P = 0.0002). A trend to better survival in cases with nuclear expression was also observed on early-stage disease group (P = 0.06).

We used a multivariate regression analysis based on the Cox proportional hazard model to test the independent value of each variable predicting overall survival among all patients and on those with late-stage disease. The estimated prognostic value of each variable in relation to overall survival is expressed as a P value. The variables used in Cox regression analysis are shown in Table 3. Subcellular localization of p27^kip1 was an independent prognostic factor for poor survival among all patients (P = 0.0007; hazard ratio, 1.7; 95% confidence interval, 1.2-2.3) and also in those with late-stage disease (P = 0.001; hazard ratio, 1.6; 95% confidence interval, 1.2-2.2). Other independent prognostic factors associated with poor prognosis were International Federation of Gynecology and Obstetrics stage, age at diagnosis >60 years, and extent of cytoreduction.

In this large study of 421 patients with primary ovarian cancer with long-term follow-up available, we analyzed the prognostic significance of p27^kip1 subcellular localization in terms of survival and its association with clinicopathologic variables. p27^kip1 has been considered as a tumor suppressor gene and loss of its function has been associated with development of many types of human cancer. The tumor suppressor function of p27^kip1 was first implicated in the context of cell cycle regulation (2). Recent studies showed that oncogenically activated kinase Akt/PKB can also phosphorylate p27^kip1 at T157 inducing its relocalization to the cytoplasm (22, 23). Akt-mediated cytosolic accumulation of p27^kip1 is critical for Akt mitogenic signaling. Akt-mediated exclusion of wild-type p27^kip1 from the nuclear compartment results in activation of nuclear cdk2 and cell cycle progression, whereas a mutation at T157 confers resistance to Akt-mediated p27^kip1 nuclear exclusion and impairs Akt-dependent rescue of p27^kip1 induced cell cycle arrest (7). Hence, sufficient evidence has accumulated to suggest that cytoplasmic relocalization of p27^kip1 might facilitate the tumor development. The presence of cytoplasmic p27^kip1 (induced by phosphorylation at T157) has been shown to predict poor prognosis in breast cancer (22, 23).

Other studies have also analyzed the relationship of p27^kip1 in ovarian carcinoma with survival using different scoring systems to quantify its expression (8–11, 24, 25). In two studies of 66 and 99 cases and scoring all immunoreactive cells regardless the subcellular localization, Masciullo et al. (8, 9) described that p27^kip1 is an independent prognostic factor of disease progression and survival; however, no correlation with other clinicopathologic variables was found. Scoring the frequency of only nuclear immunopositive cells in a series of 54 patients (30 long-term and 24 short-term survivors), Newcomb et al. (10) also reached to similar conclusions. Hurteau et al. (21) found that decreased nuclear staining of p27^kip1 was associated with poor survival in some epithelial ovarian cancers. In another study including 79 cases of ovarian carcinoma, Shigemasa et al. did not find an effect on prognosis. However, they did found a prognostic significance when only the serous carcinomas cases were studied (24). The only other study that showed lack of prognostic significance is the one from Baekelandt et al. (11). These diverse results probably reflect differences in the number of subjects, study design, and protein quantification methods among these studies, in particular specificity of antibodies.

In this study, we first validate the specificity of the antibodies used to detect the expression of subcellular localization on immunohistochemistry. Using this validated antibody, we have shown that the nuclear only expression of p27^kip1 is a favorable prognostic marker (21). However, the most interesting results from our study, and one that differs from all previous ones, was the association between cytoplasmic localization of p27^kip1 and prognosis in human ovarian cancer. The presence of p27^kip1 in the cytoplasm and regardless of the nuclear expression correlated with higher International Federation of Gynecology and Obstetrics disease stage (P < 0.03), with extent of cytoreduction (P < 0.03), and with shorter disease survival (P < 0.0001). By analyzing the subcellular localization of this cdk inhibitor, we were able to identify a subset of patients with particularly poor outcome. Using the Cox proportional hazard model of factors influencing survival, we could also show the prognostic independence of p27^kip1 in both groups, one including all the patients and in the other only those with late-stage disease. However, the underlying mechanism of why patients with negative and cytoplasmic relocalization of p27^kip1 have poor prognostic significance is not clear and will require further investigation.

In conclusion, p27^kip1 subcellular location in the cytoplasm was independently associated with poorer survival among women with ovarian carcinoma, particularly for those with late-stage disease and regardless of tumor histotype.