Ovarian cancer is the 4th commonest cause of cancer deaths in women in western countries.1 The mainstay of treatment includes cytoreductive surgery followed by adjuvant chemotherapy. Despite optimal primary treatment, recurrences are common, and the overall prognosis is poor. Various second-line chemotherapy regimes have been developed, achieving an overall response rate of about 20–30%, while producing significant side effects. Hormonal therapy would be an attractive treatment option because of its minimal side effects and relative ease of administration. However, despite up to 60% of ovarian cancers expressing estrogen receptors (ERs),2 tamoxifen, an antiestrogen well established for treatment of breast cancer, produces only a modest response in ovarian cancer.3 Studies on tamoxifen so far did not specifically examine the response with regard to receptors status. The role of ER in ovarian carcinogenesis and prediction of response to treatment or prognosis has not been clearly established. A number of earlier studies investigated the relationship between the presence of ER and survival, but the findings were controversial.2,4,5 Furthermore, since the publication of these studies, a second ER, ERβ, was discovered in 1996, which was genetically distinct from the classical ERα. These two subtypes vary in tissue distribution and ligand binding specificity and affinity. In this study, we aimed to quantify ERα and ERβ expressions in human normal, borderline, and malignant ovarian tissue samples and to correlate the expression with clinical factors, including stage, histology, grade, and progression-free and overall survival, to further understand the role of ERα and ERβ in ovarian cancers.
MATERIALS AND METHODS
This project was approved by institutional review board of the University of Hong Kong /Hospital Authority Hong Kong West Cluster (IRB# UW 06–126 T/1151). Ovarian tissues were obtained when women underwent surgical removal of the ovaries or ovarian tumor for clinical indications at the Department of Obstetrics and Gynaecology, Queen Mary Hospital, The University of Hong Kong. Immediately after collection during surgery, the tissue was dissected, freshly frozen in liquid nitrogen, and stored in our tissue bank. The histologic examinations were carried out independently by pathologists for clinical diagnosis. The histologic types and disease staging were classified according to International Federation of Gynecology and Obstetrics (FIGO) classification. Fifty-eight normal, 25 borderline, and 161 malignant ovarian tissue samples obtained between 1990 and 2006 were randomly selected. All 244 samples used in this study were obtained from different women, with each woman contributing one sample each. All 58 normal samples were obtained from women undergoing hysterectomy with bilateral salpingo-oophorectomy for benign diseases, including 38 for leiomyomata or adenomyosis, 10 for endometrial hyperplasia or polyp, eight for cervical intraepithelial hyperplasia, and one for uterine prolapse. The samples were recorded without disclosing the identities of the patients. The presence of at least 70% tumor content was confirmed histologically for all tumor samples. The mean age±standard deviation of patients with normal ovaries and borderline and malignant tumors were 51.4±13.2, 49.3±18.0, and 50.2±13.4, respectively. Patients with ovarian carcinoma were treated according to our unit’s protocol, which consisted of total abdominal hysterectomy with bilateral salpingo-oophorectomy and staging procedure for early disease and surgical attempt at optimal debulking for advanced disease, followed by adjuvant chemotherapy if the FIGO stage was stage Ic or above. Staging procedure included pelvic lymphadenectomy, omentectomy, peritoneal washings and biopsies, and appendicectomy. Optimal debulking was defined as removal of all tumors with residual disease of less than 2 cm in diameter, and this was achieved in 85% of the women. In early disease, a systematic pelvic lymphadenectomy was done, but in advanced disease, only enlarged pelvic or paraaortic lymph nodes were debulked. First-line chemotherapy had changed over the years, with cisplatin and cyclophosphamide or melphalan being used in the earlier years, and this was changed to carboplatin with Taxol (Bristol-Myers Squibb, Princeton, NJ) in recent years. Second-line chemotherapy included topotecan, gemcitabine, and liposomal doxorubicin. Tamoxifen was not routinely used. Neoadjuvant chemotherapy would be given if the tumor was not clinically operable at first presentation. Clinical data were collected from patients’ records, and the median follow-up period was 80 months (range 12–197 months). The background characteristics for these patients are summarized in Table 1.
Total RNA was extracted from the collected tissues using TRIZOL Reagent (Invitrogen Corporation, Carlsbad, CA). We then synthesized cDNA from 1 mcg of total RNA using Superscript III reverse transcriptase (Invitrogen). Real time quantitative polymerase chain reaction (RT-PCR) Taqman assay using specific primers (ERα forward: AGGTGCCCT ACTACCTGGA GAAC, ERα reverse: GGTGGCTGGACACATATAGTCGTT, ERβ forward: AAGAGCTGCCAGGCCTGCC, ERβ reverse: GCGCACTGGGGCGGCTGATCA) and probes (ERα: CGCCGGCATTCTACAGGCCAAA, ERβ: CTCACCCTCCTGGAGGCTGAGCCGC) was used to measure the ERα and ERβ mRNA expressions (ABI Prism 7500, Applied Biosystems, Foster, CA). An endogenous control, TATA box binding protein (TBP), was used to normalize the ERα and ERβ expressions to eliminate loading variations. The standard curve was constructed from CT values using five serial 10-fold dilutions of linearized plasmids of TBP to determine the absolute copy numbers of ER α and ERβ mRNA. All measurements were conducted in duplicates. Results for the quantity of mRNA were expressed as the number of copies of ER mRNA per copy of TBP mRNA (ER over TBP).
In a pilot study, we first examined whether the mRNA expressions correlated with the protein expressions of ERα and ERβ. Proteins were extracted from 32 tumor samples and 19 normal ovarian tissues using the conventional method. Estrogen receptor-α and ERβ protein expressions were then analyzed by Western blotting. The antibodies of mouse anti-ERα (Dako, Cytomation, Glostrup, Denmark) and mouse anti-ERβ (Upstate, Lake Placid, NY) were diluted in 1:1000 for immunoblotting. The sheep anti-mouse secondary antibody and donkey anti-rabbit antibody labeled with horseradish peroxidase (GE Healthcare, Giles, UK) were subsequently applied. The signal was visualized with ECL plus chemiluminescent detection kit (GE Healthcare, Chalfont St. Giles, UK). Ovarian cancer cell line C13 expressing both ERα and ERβ was used as the positive control.
Immunohistochemical staining was also performed in pilot sample of 16 normal ovarian tissues and 39 malignant tumor tissues. Formalin-fixed and paraffin-embedded clinical specimens were retrieved from the Department of Pathology, Queen Mary Hospital, The University of Hong Kong. They were sectioned at 5 μm thick and mounted on aminopropyltriethoxysilane–coated (Sigma, Saint Louis, MO) slides. All specimens were stained with hematoxylin and eosin (H&E) for histopathologic evaluation. The consecutive section was used in immunohistochemical staining, which was performed using UltraVision detection system (Thermo Scientific, Laboratory Vision Corporation, Fremont, CA). Briefly, antigen retrieval was performed by microwave pretreatment in 0.01 mol/L citrate buffer 15 minutes for ER-α or 27 minutes for ER-β. Mouse monoclonal antibodies of ER-α (Dako Cytomation, Glostrup, Denmark) and ER-β (Novo Castra, Newcastle, UK) were diluted 1:100 and 1:30, respectively. After overnight incubation with primary antibodies at 4°C, slides were incubation with primary antibody enhancer followed by horseradish peroxidase polymer, then visualized with chromogen diaminobenzidine (Amresco Inc., Salon, OH). Two endometrial cancer cases known to be immunoreactive for these two antibodies were used as positive controls. Negative control, where primary antibody was omitted, was also included in each experiment. Sections were examined at high power (×400), and 10 fields were chosen randomly. Cells were judged as positive for ERα and ERβ expressions when the nuclear was stained. The immunoreactivity was estimated and graded by scoring the percentage of positive cancer cells as follow: 0, negative; 1,less than 5%; 2, 6–25%; 3, 26–50%; 4, 51–75%; and 5, more than 75%.
Mann-Whitney U test and Wilcoxon signed rank test were used for comparison of ERα and β mRNA levels between two groups. For comparison among multiple groups, the Kruskal-Wallis test was used. Dunn’s test was used for pair-wise comparison after Kruskal-Wallis testing, where equal variances could be assumed, whereas the Tamhane test was used when equal variances could not be assumed. Box plots were used to describe the mRNA levels of the ERα and ERβ expressions where the box stretched from the 25th percentile to the 75th percentile. The median value was shown as a line across the box. Error margins indicates the smallest and largest “nonoutliers.” An outlier was defined as a value 1.5 times the interquartile range below the first quartile or 1.5 times the interquartile range above the third quartile. Survival analysis was carried out with Kaplan-Meier method and compared by log rank test. Multivariable analysis using Cox’s proportion hazards regression model was used to adjust for confounding factors that might affect the outcome. Discrete variables were analyzed by χ2 test. Statistical Package for Social Science (SPSS, SPSS Inc., Chicago, IL) 14 was used to perform statistical analysis. Statistical significance was assumed if P<.05.
In our pilot sample of 16 normal and 39 malignant ovarian tissues, immunohistochemical staining for ER α and ERβ showed that ERα immunohistochemical staining positivity correlated well (P<.001) with mRNA expression by RT-PCR. For samples with 0–0.1 copies of ERα mRNA per copy of TBP mRNA (n=11), only 9% stained positive, whereas 35.5%, 68%, and 100% stained positive for mRNA levels of 0.1–1.0 copies (n=17), 1.0–10 copies (n=22) and more than 10 copies (n=5), respectively. For ERβ, 16 (29%) samples were stained positive, but we could not find any significant correlation between mRNA levels and immunohistochemical staining results. However, we showed that ER protein expressions, measured by western blotting, corresponded with mRNA expressions for both ERα and ERβ (Fig. 1).
Estrogen receptor α and ERβ mRNA expression alone were then determined in 58 normal, 25 borderline, and 161 malignant ovarian tissue samples. There was a falling trend in ERβ expression from normal to malignant, with the expression in malignant tissue (n=161) significantly lower than in normal tissues (n=58) (P<.001) (Fig. 2). However, there was no significant difference in ERα expression between normal and malignant tissues. The ratio of ERα to ERβ was significantly higher in malignant compared with normal tissues (1.4 compared with 0.08, P<.001). Figure 3 shows the median ERα and ERβ expression for the different stages of disease for malignant tumors. Estrogen receptor β expression was significantly higher in early disease (stage I) compared with late disease (stage II–IV)(P<.001). However, ERα expression was not associated with stage of the disease.
For malignant ovarian samples, ERα was significantly higher in epithelial tumors (median 0.52, interquartile range 0.05–3.25) compared with nonepithelial tumors, including germ cell, sex-cord stromal, and carcinosarcomas (median 0.14, interquartile range 0.01–0.19, P=.032) but there was no significant difference for ERβ. Among the epithelial tumors, ERα and ERβ expression had different distributions for the different histologic types (Table 2). Endometrioid adenocarcinomas had significantly higher ERα expression compared with serous and clear adenocarcinomas (P=.003 and P<.001, respectively). In contrast to ERα expression, there were no significant differences among ERβ expression for different histologic subtypes, except for a significantly higher expression in clear cell compared with serous adenocarcinoma (P=.041). The relative distribution of the two receptor subtypes was significantly different in endometrioid, serous, and clear cell adenocarcinomas. For borderline tumors, ERβ was predominant in mucinous tumors, and ERα was more predominant in serous tumors.
For malignant tumors, 138 patients had no evidence of disease after initial treatment, whereas 23 patients had residual disease. Of those with no disease after initial treatment, 48 (34.8%) recurred. Eleven patients were lost to follow-up, and their recurrence status could not be ascertained. For borderline tumors, six of 25 (24%) recurred, of which two had stage 3c disease.
For borderline and malignant tumors (n=186), including patients with residual disease after initial surgery, survival status was known in 175 women. A total of 102 women were alive with no disease at the time of analysis, nine were alive with disease, and 64 were dead. In this cohort of 175 patients, by categorizing ERβ expression into four quartiles (ie, the lowest 25%=first quartile and the highest 25%=fourth quartile), ERβ was found to be significantly associated with disease-free survival (log rank P=.007) as well as overall survival (log rank P=.011) (Fig. 4). The 5-year overall survival for each quartile was 39%, 61%, 66%, and 78 %, respectively. However, there was no significant association between ERα expression and survival. For these 175 patients, the stage of the disease (stages I–IV), optimal debulking and borderline compared with malignant tumors were also significantly associated with both disease-free survival (P<.001, P<.001, and P=.004, respectively) and overall survival (P<.001, P<.001, and P=.013, respectively). Other factors, such as histology (clear cell compared with non–clear cell), cytologic grading, and primary treatment, were not significantly associated with survival. Multivariable analysis using Cox proportional hazards regression analysis with stage of disease, borderline compared with malignant tumors, optimal debulking, and ERβ expression in quartiles as cut off as covariates, ERβ remained a significant predictor for both disease-free and overall survival (P=.012 and P=.040, respectively) in this cohort of 175 women. Apart from ERβ status, optimal debulking was also a significant independent predictor for disease-free survival (P=.027), whereas stage and optimal debulking were found to be significant predictors for overall survival (Table 3; P=.045 and P=.001, respectively).
Previous studies investigating the relationship between the estrogen receptor status and clinical outcome in ovarian tumors concentrated on the classical estrogen receptor, ERα.6 In addition to ERα, a second ER, ERβ was identified in 1996.7 The relative levels of ERα and ERβ had been suggested to be important determinants of biologic response to ER agonists and antiestrogens in specific target tissues.8 Reduced levels of ERβ mRNA expression were found in tumor tissues compared with normal tissues in various estrogen dependent tumors such as breast and prostate cancers,9–11 suggesting that the loss of ERβ expression may be involved in carcinogenesis.
Our study of 58 normal ovarian tissue samples agreed with Pujol et al’s12 series of six normal ovaries that ERβ was the predominant ER receptor in normal ovaries. Our findings in 161 tissue samples that ERβ expression was lower in malignant ovarian tumors was in line with the findings in breast and prostate tumors,9–11 supporting the hypothesis that the loss of ERβ is associated with ovarian carcinogenesis. Estrogen receptor expression in ovarian epithelial cancers seemed to be different for different histologic types. We found that ERα expression was minimal in clear cell carcinoma compared with the others, suggesting that this might be a biologic factor that contributes to the overall poorer response to chemotherapy and prognosis. However, we found that there was no clear association between levels of ERβ expression and histologic types that have worse prognoses.
Estrogen receptor expression in borderline tumors was much less reported than in ovarian cancers. We found that ER was common in borderline tumors. Furthermore, we showed that ERβ was the predominant subtype in mucinous borderline tumors, whereas ERα was the predominant subtype in serous tumors, a pattern similar to that seen in malignant tumors. The gradual reduction of ERβ expression from normal to borderline to malignant tumors suggests that the change from normal to malignant tumors may be a continuous process, with borderline tumors representing an intermediate stage in the transformation.
Several earlier studies had investigated the level of ER in relation to survival,6,13 but these did not differentiate the effect of the two subtypes. Fujimoto et al14 followed up 28 women for 48 months, and it was found that patients with a low or high ratio of ERα to ERβ had significantly worse prognosis than patients with a medium ratio. In our study, with a larger sample, we were able to take cutoffs using quartiles for ERβ expression in ovarian tumors. With this classification, we showed that patients with a higher ERβ level had better prognosis, even after regression analysis. However, ERα level, with a similar classification as for ERβ, was not shown to significantly correlate with survival, suggesting a different role for ERβ in ovarian carcinogenesis compared with ERα.
Real-time quantitative PCR (RT-PCR) instead of immunohistochemical staining was used in this study because this was the most sensitive assay developed in recent years. Although immunohistochemical staining is the commonest method for hormone receptor status in clinical laboratories, this method relies on the subjective scoring by the pathologist and there are variations in antibody preparations. Real time quantitative PCR measurement of mRNA has the advantage of a quantitative analysis but lacks the ability for specific tissue localization as offered by immunohistochemical staining. In practice, mRNA expression cannot be directly translated into receptor positivity by immunohistochemical staining, thus limiting the direct application of our results in current clinical practice. Cutoff values for ERα mRNA copies number for receptor positivity by immunohistochemical staining in breast cancer has recently been reported, but data are still awaited for ERβ.15 To confirm that mRNA levels detected in our study can potentially be detectable by immunohistochemical staining in the clinical setting, we assessed receptor positivity by immunohistochemical staining in a pilot sample. For ERβ, although the range of mRNA levels was within that detectable by immunohistochemical staining, we failed to demonstrate any significant correlation. One possible cause was the lack of translation of mRNA into proteins, but we confirmed protein expression of both ERα and ERβ in a pilot sample using Western blotting. The lack of correlation may be secondary to the instability of antigenicity in ERβ rather than lack of protein expression. Loss of antigenicity of steroid receptors has been reported in stored sections16 but the stability of ERβ antigenicity has not been fully evaluated. Although we performed our immunohistochemical staining on freshly cut slides, antigenicity may have already been lost and may not reflect the true ERβ protein levels. This needs to be investigated further in the future and may have implications in the use of immunohistochemical staining for ERβ status assessment in clinical samples.
Currently, ER status is not commonly routinely determined in the histologic examination of the resected tumor specimen. In view of the current findings that ERβ may have a good clinical correlation and prognostic significance, more widespread determination of ERβ expression in surgically resected specimen should be considered, although the best method for ERβ measurement is yet to be determined in the clinical setting. Estrogen receptor β expression may be another factor to take into account when deciding the need for adjuvant chemotherapy in very early (FIGO stage 1a and 1b) disease or when choosing combination chemotherapy instead of single agent in stage 1c disease.
The knowledge of the level of ERβ expression may have implications for the clinical usefulness of hormonal treatment. The low level of ERβ expression in cancer tissues may explain the overall modest effect of tamoxifen in the treatment of ovarian cancer. Preliminary laboratory investigations (current work from our team) showed that hormonal treatment, eg, tamoxifen, has different responses in ovarian cancer cells expressing different ER subtype combinations. With the arrival of new selective estrogen receptor modulators with a range of different ERα-to-ERβ binding affinities, response to endocrine therapy can potentially be improved by tailoring the use of different selective estrogen receptor modulators in women with different ER subtype expression combination.
Estrogen receptor expression in borderline tumors may suggest a role for endocrine therapy in women with this group of tumors. Since borderline tumors are generally less responsive to chemotherapy, endocrine therapy would be an attractive alternative. With the knowledge that ER expression in borderline tumors is relatively high compared with malignant tumors, endocrine therapy may have an even bigger role in the treatment of these tumors. The maximal benefit may be achieved in selected patients with known ER expression. This potential treatment option will need to be tested by further clinical trials. Overall, understanding the distribution of ER may help to explore further the potential of hormonal therapy in ovarian tumors. Future studies would be needed to determine fully the clinical implications of ER subtypes in ovarian tumors.
1. U.S. Cancer Statistics Working Group. United States Cancer Statistics: 2003 Incidence and Mortality. Atlanta (GA): U.S. Department of Health and Human Services, Centers for Disease Control and Prevention and National Cancer Institute; 2006.
2. Anderl P, Fuith LC, Daxenbichler G, Marth C, Dapunt O. Correlation between steroid hormone receptors, histological and clinical parameters in ovarian carcinoma. Gynecol Obstet Invest 1988;25:135–40.
3. Perez-Gracia JL, Carrasco EM. Tamoxifen therapy for ovarian cancer in the adjuvant and advanced settings: systematic review of the literature and implications for future research. Gynecol Oncol 2002;84:201–9.
4. Masood S, Heitmann J, Nuss RC, Benrubi GI. Clinical correlation of hormone receptor status in epithelial ovarian cancer. Gynecol Oncol 1989;34:57–60.
5. Rose PG, Reale FR, Longcope C, Hunter RE. Prognostic significance of estrogen and progesterone receptors in epithelial ovarian cancer. Obstet Gynecol 1990;76:258–63.
6. Geisler JP, Wiemann MC, Miller GA, Geisler GE. Estrogen and progesterone receptors status as prognostic indicators in patients with optimally cytoreduced stage IIIc serous cystadenocarcinoma of the ovary. Gynecol Oncol 1996;60:424–7.
7. Mosselman S, Polman J, Dijkema R. ER beta: identification and characterization of a novel human estrogen receptor. FEBS Lett 1996;392:49–53.
8. Hall JM, McDonnell DP. The estrogen receptor beta-isoform (ERbeta) of the human estrogen receptor modulates ERalpha transcriptional activity and is a key regulator of the cellular response to estrogens and antiestrogens. Endocrinology 1999;140:5566–78.
9. Skliris GP, Munot K, Bell SM, Carder PJ, Lane S, Horgan K, et al. Reduced expression of oestrogen receptor beta in invasive breast cancer and its re-expression using DNA methyl transferase inhibitors in a cell line model. J Pathol 2003:201;213–20.
10. Iwao K, Miyoshi Y, Egawa C, Ikeda N, Noguchi S. Quantitative analysis of estrogen receptor-beta mRNA and its variants in human breast cancers. Int J Cancer 2000:88;733–6.
11. Horvath LG, Henshall SM, Lee CS, Head DR, Quinn DI, Makela S, et al. Frequent loss of estrogen receptor-beta expression in prostate cancer. Cancer Res 2001:61;5331–5.
12. Pujol P, Rey JM, Nirde P, Roger P, Gastaldi M, Laffargue F, et al. Differential expression of estrogen receptor-alpha and-beta messenger RNAs as a potential marker of ovarian carcinogenesis. Cancer Res 1998:58;5367–73.
13. Munstedt K, Steen J, Knauf AG, Buch T, von Georgi R, Franke FE. Steroid hormone receptors and long term survival in invasive ovarian cancer. Cancer 2000;89:1783–91.
14. Fujimoto J, Hirose R, Sakaguchi H, Tamaya T. Clinical significance of expression of estrogen receptor alpha and beta mRNAs in ovarian cancers. Oncology 2000;58:334–41.
15. Poola I, Yue Q. Estrogen receptor alpha (ER alpha) mRNA copy numbers in immunohistochemically ER alpha-positive-, and negative breast cancer tissues. BMC Cancer 2007;7:56.
© 2008 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
16. Fergenbaum JH, Garcia-Closas M, Hewitt SM, Lissowska J, Sakoda LC, Sherman ME. Loss of antigenicity in stored sections of breast cancer tissue microarrays. Cancer Epidemiol Biomarkers Prev 2004;13:667–72.