• prophylactic oophorectomies;
  • Disabled-2;
  • ovarian carcinoma;
  • epithelium;
  • premalignant lesion;
  • collagen IV;
  • laminin;
  • Cox-2;
  • basement membrane


  1. Top of page
  2. Abstract
  6. Acknowledgements


The authors suggested that the loss of collagen IV and laminin-containing basement membrane and the loss of Disabled-2 (Dab2) expression were two critical events associated with morphologic dysplastic changes of the ovarian surface epithelium as a step in tumorigenicity. Both the basement membrane and Dab2, a candidate tumor suppressor of ovarian carcinoma, were involved in epithelial cell surface positioning and organization. The authors speculated that the purging of the basement membrane may be similar to the proteolysis during gonadotropin-stimulated ovulation, a cyclooxygenase 2 (Cox-2)-mediated process.


Prophylactic oophorectomy is used to prevent breast and ovarian carcinoma in high-risk populations. These ovarian tissue specimens often contain an increased presence of morphologically abnormal lesions that are believed to be preneoplastic. The authors evaluated archived prophylactic oophorectomy specimens to verify whether the loss of Dab2 expression and the removal of the basement membrane that occur at the ovarian surface and inclusion cyst epithelia are molecular markers of preneoplastic lesions. Of the 36 samples containing identifiable ovarian surface epithelial components on slides, immunostaining was employed to evaluate the intactness of the basement membrane (periodic acid–Schiff [PAS], collagen IV, and laminin) and the expression of Dab2 and Cox-2. Expression of Cox-1 and Cox-2 also were evaluated in cultured ovarian surface epithelial cells prepared from ovarian tissue specimens removed from patients who underwent prophylactic surgery.


The morphologically normal ovarian surface epithelium typically contained a collagen IV- and laminin-positive basement membrane, which also was detected by PAS staining. Many morphologically altered areas, such as papillomatosis, invaginations, inclusion cysts, stratification, adenomas, and microscopic adenocarcinomas, were found in these specimens. Both the morphologically altered and adjacent morphologically normal epithelia consistently exhibited loss of basement membrane and/or Dab2 expression and an increase in Cox-2 staining. Frequently, an increase in Cox-2 staining was correlated with the loss of epithelial basement membrane in morphologically normal areas.


The loss of Dab2 and basement membrane and the overexpression of Cox-2 were observed in presumptive neoplastic precursor areas of oophorectomy specimens obtained from a population at high risk for ovarian carcinoma. Transient loss of collagen IV and laminin in the basement membrane of the preneoplastic epithelium and the loss of Dab2 expression are common early events associated with morphologic alteration and tumorigenicity of the ovarian surface epithelium. The authors concluded that Cox-2 overexpression may play a role in the purging of basement membrane of the ovarian surface epithelium, mimicking the process of ovulation. Further experiments may be able to test the hypothetical model derived from these histologic observations. Cancer 2003;98:2607–23. © 2003 American Cancer Society.

An accumulation of genetic and epigenetic alterations is associated with the stepwise progression of epithelial cell transformation.1, 2 However, there is little understanding of the molecular events associated with the process of ovarian surface epithelial neoplasia.3–7 The lack of understanding of the molecular basis for early neoplastic changes in the ovarian surface epithelium is partially due to the lack of access to preneoplastic tissues and low-grade tumors because of the difficulty in diagnosing ovarian carcinoma at an early stage.8–10 Recently, women from high-risk breast and ovarian carcinoma families, often BRCA1 and BRCA2 carriers, elected prophylactic oophorectomy as a preventive approach.10, 11 Ovarian tissue specimens from women with a high risk for breast and ovarian carcinoma but without clinical diagnoses or symptoms have provided the means to investigate ovarian preneoplastic changes.12–15 Several studies have found microscopic benign-to-malignant morphologic changes in ovarian specimens,14–16 and others have suggested the existence of preneoplastic phenotypes in the cells prepared from ovaries in cancer-prone women.13, 15 However, morphologic changes in the ovarian specimens of BRCA1 carriers may not be correlated with p53 mutations or ErbB-2 overexpression, nor do the specimens from at-risk women differ significantly from control cases in cell proliferation and apoptosis.12 Nevertheless, it should be possible to identify alterations in the microenvironment and gene expression of the ovarian surface and/or cystic epithelial cells that are associated with or account for the preneoplastic morphologic changes. Characterization of the molecular changes in morphologically altered areas of ovarian specimens from cancer-prone women would support the hypothesis that these lesions are indeed precursors of ovarian neoplasia.

The majority of human ovarian neoplasms are derived from the epithelial cells on the surface and/or lining of inclusion cysts.8, 9, 17–19 The ovarian surface epithelial cells attach to and are organized by a well-defined basement membrane.3, 20–22 During ovulation stimulated by pituitary gonadotropins (luteinizing hormone [LH] and follicle-stimulating hormone [FSH]), the basement membranes of both the follicles and surface epithelia are degraded, allowing the release of the ovule.23, 24 The degradative processes in ovulation are mediated by prostaglandins and can be blocked by cyclooxygenase 2 (Cox-2) inhibitors25–29 or by genetic abolition of the enzyme in mice.30–34

The intactness of the basement membrane of the ovarian surface and cyst greatly influences the biology of the epithelial cells attached, because contact with the basement membrane provides regulatory signals for the epithelial cells.35–37 Studies have shown that transgenic expression of the basement membrane–degrading enzyme MMP3/stromelysin-1 in mouse mammary glands promotes tumor development due to the altered microenvironment after basement membrane degradation.38–41 This suggests that the basement membrane is important for the maintenance of the epithelial characters and for tumor suppression.20–22

Disabled-2 (Dab2), a candidate tumor suppressor in ovarian carcinoma, also has a role in the organization of ovarian surface epithelial cells.21, 42, 43 Its expression is lost or greatly diminished in 85% of breast and ovarian tumors and cancer cell lines.42 Dab2 loss occurs in premalignant lesions and is not correlated with tumor progression. Therefore, the inactivation of Dab2 expression is an early event in ovarian tumorigenicity.22, 42 We have speculated that Dab2 functions in epithelial cell organization and that inactivation of Dab2 leads to the loss of the monolayer character of the ovarian surface epithelium, which could contribute to tumorigenicity.21 Gene knockout in mice has confirmed that Dab2 plays a key role in the organization and surface positioning of epithelial cells.44

A recent study of the putative preneoplastic lesions of ovarian tumors suggested that the collagen IV and laminin-containing basement membrane underlying the ovarian surface epithelium is transiently lost before morphologic transformation of the ovarian surface epithelial cells.22 A model proposed that the loss of the basement membrane of the ovarian surface epithelium and the subsequent loss of Dab2 expression are two critical steps in the neoplastic and morphologic transformation of ovarian surface epithelial cells.22 Because the model is based on observations in ovarian tumor tissue specimens, it would be highly informative to evaluate whether preneoplastic changes occur in specimens from cancer-prone women that do not contain overt malignancies. We analyzed preneoplastic ovarian tissue specimens from women with a family history of breast and ovarian carcinoma for changes in epithelial basement membrane and Dab2 expression. The specimens were obtained during prophylactic oophorectomies.


  1. Top of page
  2. Abstract
  6. Acknowledgements

Ovarian Tissue Specimens from Prophylactic Oophorectomies

A representative set of recently archived (years 2000–2002) human ovarian tissue specimens from prophylactic oophorectomies performed at Fox Chase Cancer Center (Philadelphia, PA) and affiliated hospitals was used in the current study. When available, specimens from both the left and right ovaries from the same donor were evaluated as independent samples. For Patients 1, 2, 11, and 13, multiple blocks from the same ovary were evaluated. The majority of surgical oophorectomies were performed as a preventive approach in women with a genetic or perceived increased risk of developing ovarian carcinoma, i.e., BRCA1 or BRCA2 mutation carriers, or those reporting a personal and/or family history of disease.

A portion of each ovarian tissue specimen was fixed in buffered formalin and embedded in paraffin after surgery and visual inspection by pathologists. Signed informed consent for genetic testing was obtained from individuals who were genotyped for mutations in BRCA1 and/or BRCA2 (for persons with a family history of breast and/or ovarian carcinoma) or in MSH2 and/or MLH1 (for individuals with a family history of colon carcinoma). Two tissue specimens were obtained from Patients 7 and 8, unaffected women from clinically diagnosed hereditary nonpolyposis colorectal carcinoma kindreds, who carried a germline mutation in MLH1. Tissue specimens were obtained from a woman with breast carcinoma who did not report a history of breast and ovarian carcinoma (Patient 11) and from a woman who declined to release this information (Patient 15).

Patients 14, 15, and 16 declined genetic testing. A record of family cancer risk history and identified germline mutations was obtained without reference to patients' personal information. The use of human tissue specimens in the current study was evaluated and approved by the institutional review board (IRB). Safety and ethical guidelines were followed according to institutional requirements. To preserve privacy, a series of security procedures were undertaken. Each study subject was given a unique numeric identifier upon study entry. Biologic specimens were labeled with only this unique biosample identifier. Neither the individual's name nor any other identifier appeared on the biosample. In addition, all personnel received HIPAA and human subjects protection training and were certified by the IRB.

Tumors and Tissue Microarray

Archived tumor tissue specimens were used for immunostaining. We used slides of tissue arrays produced in the Fox Chase Cancer Center tumor bank by Dr. Klein-Szanto. In each slide, a selected panel of tissue specimens of 1 mm in diameter was used. The tissue specimens included 18 tumors of the ovary of various histologic subtypes, 9 metastatic tumors of ovarian and other origin found in the omentum, and several human tissue types, including a normal ovary.


The tissue paraffin blocks were cut into 5-μm sections and placed on positively charged glass slides. Adjacent sections from the same tissue block were stained with anti-Dab2, anti-collagen IV, anti-laminin, anti-Cox-2, and anti-Mib1 after steam heat–induced antigen retrieval. The procedure has been described previously.20, 22 The antibodies and dilutions used are listed in Table 1.

Table 1. Antibodies Used in Immunostaining
Collagen IVMouseCIV 221/40Dako (Carpinteria, CA)
LamininMouse4C71/160Dako (Carpinteria, CA)
Disabled-2MouseClone 521/400BD Transduction Lab (Lexington, KY)
Mib-1 (Ki67)RabbitNCL-KI67p1/200Novocastra (Newcastle, United Kingdom)
Cox-2GoatC-201/400Santa Cruz (Santa Cruz, CA)
Cox-2RabbitPolyclonal1/400Cayman (Ann Arbor, MI)

The appropriate antibody dilution was determined using positive and negative control slides before the study. Both positive and negative controls were included during immunostaining of the tested slides. A Cox-2 peptide was tested and found to block staining, indicating that the Cox-2 immunostainings observed are specific. Tissue specimens were stained with periodic acid–Schiff (PAS) after deparaffinization and hydration and then counterstained with hematoxylin.

Human Ovarian Surface Epithelial Cell and ‘Immortalized’ Lines

Some of the selected ovarian tissue specimens obtained during prophylactic oophorectomies were used to prepare human ovarian surface epithelial (HOSE) cells. Briefly, the surface epithelium was scraped from the surface of the freshly dissected ovarian tissue specimens. The cells released were cultured in medium 199 and MCDB-105 (1:1) supplemented with 4% fetal bovine serum (FBS) and 0.2 units/mL of insulin (Novagen; Madison, WI). The cells were characterized as > 90% epithelial. These cells enter replicative senescence after culturing for 1–2 months. To prolong the culturing period, early passages of the primary HOSE cells were transfected with an SV40 large T antigen expression vector. These cells are referred to as human ‘immortalized’ ovarian surface epithelial (HIO) cells and can undergo an additional 20–30 population doublings before ceasing proliferation. Three preparations of primary HOSE cells and five lines of HIO cells (HIO-80, HIO-120, HIO-114, HIO-118, and HIO-105) from separate preparations were used in the current study.20

Cell Culture and Western Blot Analysis

Ovarian epithelial and tumor cell lines (the OVCAR lines) were previously established.6, 7 The cells were cultured in Dulbecco modified Eagle medium with 10% FBS. The total cell lysate was prepared by collecting monolayered cells grown on a culture dish into sodium dodecyl sulfate (SDS) gel loading buffer and heating in boiling water for 5 minutes. Protein (50 μg) for each lysate was loaded onto SDS-acrylamide gels. Western blotting was performed according to standard procedures as described earlier.42 Monoclonal anti-Cox-1 and anti-Cox-2 antibodies used for Western blotting were purchased from Cayman (Ann Arbor, MI).


  1. Top of page
  2. Abstract
  6. Acknowledgements

Morphologic Characterization of Tissue Specimens from Prophylactic Oophorectomies

We analyzed ovarian tissue specimens from individuals with a high risk for breast and ovarian carcinoma. Most of the women who provided the specimens used in the current study were genotyped as BRCA1 and BRCA2 or MLH1 mutation carriers or had a recognizable family history of breast and ovarian carcinoma risk (Table 2). The presence of epithelial morphologic abnormalities in the ovarian specimens suggests that some areas of the specimens are preneoplastic. We believed that we would find the loss of Dab2 and basement membrane in these lesions. Recently archived ovarian tissue specimens from prophylactic oophorectomies were sectioned for the current study. Few follicles were found in any of these sections, consistent with the fact that most of the ovarian specimens were obtained from women of postreproductive age. Macroscopic evaluations of sections from 36 ovarian tissue samples revealed a variety of morphologic changes similar to those observed in another set of ovarian samples.15 Four (11%) microscopic neoplasms were identified in these 36 ovarian tissue samples. These are microscopic tumors of the endometrioid or serous papillary adenocarcinoma histologic type (Fig. 1A). The lesions are 1–3 mm in diameter, as determined by evaluating 20 consecutive sections. Other morphologic abnormalities, including papillomatosis, inclusion cysts, invaginations, and epithelial pseudostratification/dysplasia, were observed with a wide range of features in most ovarian tissue specimens (Table 2). Thirty-nine percent, 30%, 47%, and 36% of tissue specimens had morphologic characteristics of papillomatosis, inclusion cyst, invagination, and pseudostratification, respectively. Examples of these morphologic changes are shown in Figure 1B. Consistent with several previous investigations,12–15 morphologic changes are present in ovarian tissue samples from high-risk populations with BRCA1/2 or MLH1 mutations or a family history of breast and ovarian carcinoma. However, we have made no attempt to compare these with ovarian specimens from the normal-risk population, because an insufficient number are available for study.

Table 2. Morphologic Characterization of Prophylactic Oophorectomiesa
Sample no.Patient age (yrs)Mutation statusPatient cancer historyFirst-degree relative cancer historySecond-degree relative cancer historyPapillomatosisInclusion cystInvaginationPseudostratificationDysplasia
  • NMD: no mutation detected in BRCA1 or BRCA2; other: malignancy other than breast or ovarian.

  • a

    A set of 36 representative human ovarian tissue specimens from prophylactic oophorectomies performed for 16 individuals were characterized by evaluating the tissue sections for morphologic changes in the epithelial components of the ovaries. In some cases, multiple tissue specimens from two ovaries of an individual were used for analysis. The morphology of the epithelial components was classified as papillomatosis, inclusion cyst, invagination, epithelial pseudostratification, or dysplasia. Some of the women were genotyped for BRCA1, BRCA2, and MLH1 mutations before they underwent prophylactic oophorectomy. Some are members of a family with genetically related family members and siblings with breast and/or ovarian carcinoma. Other samples are from non-cancer-related diseases or from patients whose information was not made available to us. The results of the histologic analysis are scored semiquantitatively on the basis of the degree and extent of the morphologic abnormalities. −: no change; +: slight; ++: moderate; +++: strong.

  • b

    Patient did not consent to genetic susceptibility testing.

1-137BRCA1 mutation carrierUnaffected1 ovarian1 ovarian,6 other++
1-3     +++++
1-4     +++
1-5     ++++
1-6     +++++
1-7     +++
2-135BRCA1 mutation carrierBreast carcinoma2 breast,1 other5 ovarian,14 breast, 12 other+
2-2     ++
2-3     +++
349BRCA1 mutation carrierUnknownNot reportedNot reported++
4-144BRCA2 mutation carrierBreast carcinoma2 breast1 other+++++++
4-2     ++++
5-144BRCA2 mutation carrierBreast carcinoma3 breast2 other++++
5-2     +++
6-145BRCA2 mutation carrierUnaffected1 breast4 breast,1 other++++
6-245    ++++
742MLH1 mutation carrierUnaffected1 other5 other+++++
839MLH1 mutation carrierUnaffected1 other3 other++
9-136NMD mutationUnaffected1 breast2 ovarian++++
9-2     ++++
10-171NMDUnaffected1 ovarian1 ovarian,2 breast++++
10-2     +++
11-152NMDBreast carcinoma angiomaNoneNone+++
11-2     ++
11-3     +++
1270NMDBenign ovarian tumor, metastatic cervical carcinoma1 ovarianNone+
13-137NMDBreast carcinoma4 other1 breast,5 other++
13-2     ++++++
13-3     ++
13-4     +
1454Declined testingbBreast carcinoma1 breast 1 ovarian1 breast, 2 other
1571Declined testingbUnknownNot reportedNot reported+++++
1645Declined testingbUnaffected1 ovarian,3 other3 other++++
thumbnail image

Figure 1. Periodic acid–Schiff (PAS) staining of ovarian surface epithelial lesions from high-risk individuals who underwent prophylactic oophorectomy. (A) Four microscopic neoplastic lesions were found in the examination of 36 ovarian tissue specimens from a high-risk population. The lesions are 1–3 mm in diameter, as judged from 20 consecutive sections. Photos of lower (×40) and higher (×400) magnification are shown for each lesion. (B) PAS-stained sections show morphologic abnormalities in surface epithelia from a high-risk population. Examples shown are papillomatosis, inclusion cysts, invaginations, and epithelial pseudostratification. (C) Characteristics of morphologically normal, flat-to-columnar surface epithelia found in ovaries. A typical example of a morphologically normal ovarian surface epithelium is shown for immunostaining and is basement membrane–positive as indicated by PAS and collagen IV (Col IV) staining. The epithelial cells are positive for Disabled-2 (Dab2) and negative for cyclooxygenase-2 (Cox-2).

Download figure to PowerPoint

Loss of Dab2 Expression Is Correlated with Dysplastic Morphologic Alteration of Ovarian Epithelium

All sections containing epithelial components were analyzed for Dab2 expression by immunohistology. In the morphologically normal (flat-to-squamous) ovarian surface epithelial cells, Dab2 status is positive, Cox-2 status is negative, and a well-defined basement membrane can be seen with PAS or collagen IV staining (Fig. 1C). There are various extents of loss of Dab2 expression in epithelial cells identified in 27 of the tissue specimens. Dab2 expression is consistently absent in morphologically transformed multicell-layered or papillary epithelia found in the ovarian tissue specimens from prophylactic oophorectomies of a high-risk population for breast and ovarian carcinoma (Fig. 2A). As shown in ovarian tissue specimens containing both monolayer and transformed epithelia, monolayer epithelial cells often are Dab2 positive, although Dab2-negative cells also are present (Fig. 2A [arrow]). In the same specimen, dysplastic epithelial cells are Dab2-negative. Close evaluation shows that the transition of Dab2-positive to Dab2-negative cells exists on a contiguous epithelium, correlated with the transition of the cell morphology from flat to columnar (Fig. 2B, C). In several examples, the expression of Dab2 is lost in the epithelial cells, although the basement membrane lying underneath is intact, as indicated by collagen IV (Fig. 2B) or PAS staining (Fig. 2C). The loss of Dab2 in morphologically altered noncancerous ovarian specimens indicates that the inactivation of Dab2 is an early step and suggests that the morphologically altered epithelial cells may be the precursors of cancer cells.

thumbnail image

Figure 2. Loss of Disabled-2 (Dab2) expression is correlated with morphologic abnormalities of surface epithelia from high-risk women who received a prophylactic oophorectomy. The ovarian tissue specimens from women at high risk for breast and ovarian carcinoma were analyzed for Dab2 expression by immunostaining. (A) In a representative ovarian tissue section, Dab2 expression is positive in monolayered epithelium (arrowhead), although Dab2 negative cells are present (arrow). Dab2 expression is consistently absent in morphologic transformed epithelia from the same section. (B) Two examples showing the transition from Dab2-positive flat epithelial cells (arrowhead) to Dab2-negative columnar cells (arrow). Adjacent sections were stained for collagen IV to visualize the basement membrane. (C) Another example of the transition of Dab2-positive, flat epithelial cells (arrowhead) linked to Dab2-negative columnar cells (arrow). Adjacent sections were stained with periodic acid–Schiff to visualize the basement membrane.

Download figure to PowerPoint

Regional Losses of Basement Membrane in Ovarian Surface Epithelium from Prophylactic Oophorectomies

From previous observations of preneoplastic lesions (areas of morphologically normal epithelia that are contiguous to neoplastic epithelia), we concluded that the loss of basement membrane is an early event in ovarian tumorigenicity.22 We analyzed the intactness of the basement membrane of the epithelial components of the ovarian specimens by PAS, laminin, and collagen IV immunohistologic staining. A well-defined basement membrane was identified in the morphologic normal epithelia of the ovarian specimens either by PAS, laminin, or collagen IV staining (Fig. 1C). In tissue specimens from prophylactic oophorectomies, basement membranes often are missing in morphologically normal surface or cyst epithelia that lie immediately adjacent to morphologically atypical epithelia. For example, both basement membrane–positive (arrowhead) and negative (arrow) epithelia are observed in the same ovarian section (Fig. 3A). Several PAS-stained ovarian surface and/or cyst epithelia are shown (Fig. 3B), and the areas where the basement membrane is positive (arrowhead) and negative (arrow) are indicated. In these areas, the immunostaining of laminin (Fig. 3C) and collagen IV (Fig. 3D) often is negative in both monolayer and transformed epithelia. A PAS-stained layer of glycoproteins also is undetectable (Fig. 3A–C), indicating that the majority of the components of the basement membrane are lost. Of the 36 ovarian samples analyzed, we observed 51 epithelial regions in which the basement membrane is present, based on staining of PAS, collagen IV, and/or laminin, and 35 cases of epithelial regions were found to be basement membrane–negative. Therefore, 41% of the surface and cyst epithelia observed in the ovarian specimens are devoid of basement membranes.

thumbnail image

Figure 3. Frequent absence of collagen IV and laminin-containing basement membrane of ovarian surface and cyst epithelia. Periodic acid–Schiff (PAS) staining was used to visualize the integrity of epithelial basement membrane in ovarian tissue sections from women at high risk for cancer who received prophylactic oophorectomies. (A) A typical example is shown for a basement membrane-positive (arrowhead) compared with a basement membrane-negative (arrow) epithelium from the same section. (B) The 36 prophylactic oophorectomy specimens were analyzed for the presence of surface and cyst epithelial basement membrane by PAS staining. Examples of basement membrane–positive (arrowhead) and basement membrane–negative (arrow) areas. (C) Two examples of epithelia in ovarian invaginations stained with PAS and laminin. (D) Six examples of collagen IV immunostaining of surface epithelia from prophylactic oophorectomies. Both basement membrane–positive (arrowhead) and basement membrane–negative (arrow) epithelia are present.

Download figure to PowerPoint

Loss of Basement Membrane Is Correlated with Overexpression of Cyclooxygenase-2 in Ovarian Surface Epithelia

Ovulation is suppressed by Cox-2 inhibitors26–29 and is dramatically impaired in Cox-2-deficient mice.30–34 Furthermore, Cox-2 is induced by gonadotropins to initiate ovulation.47, 48 Therefore, Cox-2 expression mediates proteolysis during the inflammatory-like process of ovulation.23, 25, 28, 47 We investigated Cox-2 expression in the tissue specimens from prophylactic oophorectomies to determine whether Cox-2 expression is correlated with the loss of basement membrane. In morphologically normal ovarian surface epithelia, Cox-2 expression is low, as judged by its weak immunostaining (Figs. 1C, 4A). However, we found that in the preneoplastic lesions, in which the basement membrane is absent Cox-2 is consistently overexpressed (Fig. 4B). In these lesions, the basement membrane is negative, as indicated by an absence of PAS and collagen IV staining in adjacent sections (Fig. 4B). In a systematic evaluation of the relation between Cox-2 expression and collagen IV staining of the basement membrane in corresponding epithelium in adjacent sections of each of the 36 ovarian tissue specimens, we documented 19 cases of epithelia that are basement membrane negative and Cox-2 positive, 6 cases that are basement membrane positive and Cox-2 negative, and 8 cases that are positive for both basement membrane and Cox-2. These eight-cases are exclusively papillomas or adenomalike transformed epithelia. In many cases, collagen IV staining localizes to stromal areas and no discrete basement membrane is present. No monolayered epithelia were found Cox-2 positive and basement membrane positive. We conclude that Cox-2 often is overexpressed in preneoplastic epithelia of the ovarian surface and inclusion cysts and is correlated with the loss of basement membrane.

thumbnail image

Figure 4. Correlation between cyclooxygenase-2 (Cox-2) overexpression and loss of basement membrane. (A) In the morphologically normal, basement membrane–positive epithelium, Cox-2 expression is low or absent. An example shows the staining of adjacent sections of an ovary for Cox-2, periodic acid–Schiff (PAS), collagen IV (Col IV), and laminin (Lam). (B) Three examples show that in surface and cyst epithelia from prophylactic oophorectomies, the basement membrane is absent as indicated by PAS and collagen IV immunostaining and Cox-2 staining is strongly positive in the epithelia.

Download figure to PowerPoint

Aberrant Expression of Cox-1 and Cox-2 in Ovarian Surface Epithelial and Tumor Cell Lines

We analyzed the expression of Cox-1 and Cox-2 in several preparations of HOSE cells and several lines of HIO and tumor cells. Among the three preparations of primary HOSE cells and the five lines of HIO cells analyzed, Cox-2 is absent in four cell lines (HOSE 2, HOSE 3, HIO-117, and HIO-80), weakly detectable in HOSE 1, and strongly expressed in HIO-118, HIO-121, and HIO-135 (Fig. 5A). Cox-1, however, is expressed in two primary HOSE preparations but not in any HIO cells. In tumor cells, Cox-1 is more frequently expressed. Cox-2 is detected strongly in OVCAR5 and weakly in OVCAR8. Thus, the overexpression of Cox-2 is observed in a fraction of isolated and cultured ovarian surface epithelial cells. All tumor lines express either Cox-1 or Cox-2. Perhaps in the ovary, Cox-2 expression is involved in physiologic function such as ovulation and the early stages of cell transformation. In later stages of tumor development, overexpression of either Cox-1 or Cox-2 may contribute to malignancy. Nevertheless, the above observations were derived from cultured cells, which exist in a very different environment and may not be able to accurately reflect gene expression status in tissue specimens.

thumbnail image

Figure 5. Expression of cyclooxygenase-2 (Cox-2) in ovarian surface epithelial cells and tumors. (A) Expression of Cox-2 in primary human (HOSE) and immortalized (HIO) ovarian surface epithelial and tumor cell lines. The expression of Cox-1 and Cox-2 in the three HOSE, five HIO, and five tumor cell lines of ovarian surface epithelial origin (OVCAR) were analyzed by Western blotting. Total cell lysates (50 μg) were used, with beta-actin serving as a loading control. (B–J) Tissue microarray was stained to detect Cox-2 expression (see the results in Table 3). (B) 00-099, clear cell adenocarcinoma. (C) 96-311, serous endometrioid carcinoma. (D) 95-187, endometrioid adenocarcinoma. (E) 98-209, papilloendometrioid carcinoma. (F) 99-017, papilloserous adenocarcinoma. (G) 99-022, endometrioid adenocarcinoma. (H) 96-274, metastatic pancreatic adenocarcinoma. (I) liver. (J) 95-211, metastatic colon adenocarcinoma.

Download figure to PowerPoint

Table 3. Cyclooxygenase-2 Staining in Ovarian Carcinomasa
TB no.LocationTissue specimenDiagnosisWHO GradeCox-2
  • WHO: World Health Organization; TB: Tumor Bank; OSE: ovarian surface epithelium.

  • a

    Summary of cyclooxygenase-2 immunostaining of a tissue microarray slide.

SP-98-4475 BC2–C3OvarySerous cystadenocarcinoma2
S-95-9901 A4C4–C5OvaryModerately-to-poorly differentiated mucinous cystadenocarcinoma2
94-3575-1EC6–C7OvaryModerately differentiated infiltrating mucinous papillary cystadenocarcinoma2
95-187C9–C10OvaryEndometrioid adenocarcinoma2
99-022C11, D2OvaryEndometrioid adenocarcinoma1+
00-225C12, D3OvaryEndometrioid adenocarcinoma1
96-311D4–D5OvarySerous endometrioid adenocarcinoma3
97-247D7–D8OvaryEndometrioid adenocarcinoma3
98-047D9–D10OvarySerous endometrioid adenocarcinoma4
98-209D11–D12OvaryPapillary endometrioid carcinoma3+
99-050E2–E3OvaryGrade III endometrioid adenocarcinoma3
00-034E5–E6OvaryClear cell carcinoma/
00-070E7–E8OvaryClear cell adenocarcinoma3
00-099E9–E10OvaryClear cell adenocarcinoma3
98-095E11–E12OvaryMixed serous mucinous endometrioid carcinoma2
00-211F2–F3OvaryMixed mesodermal tumor/
97-177F4–F5OvaryPoorly differentiated carcinoma/
98-030F6–F7OmentumMetastatic adenocarcinoma/
95-211F8–F9OmentumMetastatic colon adenocarcinoma/++
96-201F11–F12OmentumMetastatic colon adenocarcinoma/
96-237G2–G3OmentumMetastatic colon adenocarcinoma/
96-274G4–G5OmentumMetastatic pancreatic adenocarcinoma/++
97-229G6–G7OmentumMetastatic adenocarcinoma/
97-273G9–G10OmentumMetastatic appendix adenocarcinoma/
99-017H2–H3OmentumPapillary serous adenocarcinoma/+
SO-95-2086H5–H6OvarySerous cystoadenoma/
01-280 ND1OvaryOSE normal0

Lack of Expression of Cox-2 in Ovarian Tumors

Whether Cox-2 is expressed in ovarian tumors and whether Cox-2 expression is involved in ovarian tumorigenicity is not certain. Both positive45, 46, 49–53 and negative54 Cox-2 expression in ovarian tumors have been reported. We analyzed the expression of Cox-2 in a panel of ovarian tumors on slides of tissue microarrays (Table 3). Of the 18 tumors of various histologic subtypes confined to ovaries, most tumors do not express levels of Cox-2 that are detectable by staining, and Cox-2 is weakly positive in two tumors (Fig. 5B–J). Nevertheless, several tumors that had metastasized to the omentum were found to be Cox-2 positive (Table 3). Among tumors of ovarian surface epithelial origin, only one (99-017), a papilloadenocarcinoma that metastasized to the omentum, is Cox-2 positive. We conclude that unlike in other tumors, Cox-2 overexpression is not common in established ovarian tumors, consistent with a recent report.54 We speculate that Cox-2 is involved in the initiation but not the malignant progression of ovarian carcinoma.


  1. Top of page
  2. Abstract
  6. Acknowledgements

During the transformation of a simple epithelium such as that of the ovarian surface to multilayered cells characteristic of dysplasia, the epithelial cells need to overcome two organizational forces: 1) apical-basal polarity of the epithelial cells that restricts the interacting surface domain of the cells and 2) attachment to a basement membrane. From the observations of transitional epithelia linking morphologically normal lesions to neoplastic lesions in ovarian tumors, two alterations were found and proposed to be the driving force in the transformation process, which may overcome the two forces controlling epithelial positioning.22 Preceding transformation, the basement membrane disappears and subsequently the expression of Dab2 is lost. Dab2 is a cargo-specific endocytic adaptor55 believed to function in establishing and maintaining epithelial polarity44 (unpublished data). Presumably, the loss of basement membrane and polarity (Dab2 expression) enables the cells to escape the restriction imposed by the architectural constraint of epithelial tissue organization and allows them to undergo morphologic transformation.

We report examples of preneoplastic molecular changes in oophorectomy specimens obtained from women at high risk for breast and ovarian carcinoma. These specimens often contain morphologic changes in the surface epithelia that are considered preneoplastic and serve as an excellent resource in which to evaluate preneoplastic molecular changes. The loss of basement membrane and the loss of Dab2 expression can occur independently of each other. Morphologic transformation presumably occurs when both basement membrane and Dab2 expression are lost.

Expression and Functions of Cox-2 in Ovaries and Tumors

Two prostaglandin-producing enzymes, Cox-1 and Cox-2, are present in mammals.56 It is believed that Cox-1 is expressed constitutively and that Cox-2 can be induced by hormones and cytokines in processes such as ovulation and inflammation.56 The functions of both Cox-1 and Cox-2 were investigated using knockout mouse models.32–34 Although prostaglandin levels are reduced by 99% in Cox-1 knockout mice, no significant defects were observed.57 Cox-2 knockout mice are female infertile with multiple reproductive defects,31 including anovulation,30 and it is concluded that Cox-2 plays a regulatory role in the activation of proteolysis during ovulation. Cox-2 was found to be overexpressed in several malignancies, including those of the colon, breast, skin, and ovary.46, 51–53, 58 Whether Cox-2 is expressed in ovarian tumors is somewhat controversial. Some reports found overexpression of Cox-2,45, 46, 49–53 but another stated that Cox-2 is not overexpressed in ovarian tumors and cell lines.54

Our investigation indicates that Cox-2 often is expressed in ovaries undergoing early preneoplastic changes, in both tissues and cell lines. However, there is no evidence of frequent Cox-2 overexpression in ovarian tumors. In fact, it appears that Cox-2 is expressed more frequently in preneoplastic cells. It is likely that the increase in Cox-2 expression in the preneoplastic ovarian surface epithelium resembles that of ovulation, whereas the tumor cells, through much additional change, may have lost the characteristic of responding to gonadotropins for ovulation and Cox-2 induction. One unexpected finding is that in tumor cells in which Cox-2 is not expressed, Cox-1 is expressed. This is somewhat surprising, because the expression of Cox-1 is not known to be regulated.56 Nevertheless, Cox-1 expression may contribute to the malignancy of ovarian tumor cells, as suggested by a recent report.54

A Possible Mechanism of Cox-2 in Ovarian Tumorigenicity

The cancer-promoting activity of Cox-2 expression and the antineoplastic activity of Cox-2 inhibitors have been recognized in general,58–61 as well as in ovarian carcinoma.45, 49–53 Furthermore, transgenic overexpression of Cox-2 is sufficient to induce mammary tumors in mice.62 The mechanisms for the action of Cox-2 in tumor promotion are not well understood. It has been proposed that Cox-2 stimulates tumor or preneoplastic cell proliferation,63, 64 suppresses apoptosis,63, 65 and/or induces angiogenes.66, 67 These mechanisms, however, are not well correlated with in vivo results,68 and additional mechanisms should be considered.

In the ovaries, Cox-2 mediates the gonadotropin-induced ovulation and the proteolysis of the basement membrane.47 The loss of basement membrane, as observed in the current study of ovarian specimens from prophylactic oophorectomies, may dramatically alter the biology of the epithelial cells in cell contact signaling and positional organization.21, 22, 35, 36 It has been shown that transgenic expression of either epithelial or stromal matrix metalloproteinases leads to the loss of mammary gland epithelial basement membrane and the induction of mammary tumorigenicity.38–41, 69 We propose that in ovarian surface epithelial cells, the carcinogenic mechanism of Cox-2 involves the induction of ovulation-like basement membrane loss (Fig. 6A). Cox-2 is induced in both surface epithelial and follicle granulosa cells before ovulation.47, 48, 70 Ovulation is suppressed by inhibitors (indomethacin) of Cox-1 and Cox-2 and can be restored by the addition of prostaglandin E2.26, 27, 29, 47, 71 Prostaglandins produced by Cox-2 also mediate the inhibition of collagen synthesis.25, 28 The loss of basement membrane and tissue remodeling lead to the ultimate rupture in ovulation (Fig. 6A). We hypothesize that high Cox-2 expression may stimulate ovulation-like proteolytic activity, leading to loss of basement membrane without rupture, especially in the ovaries of postmenopausal women (Fig. 6B). Thus, we speculate that one possible mechanism for the carcinogenic activity of Cox-2 in ovarian epithelia involves the stimulation of ovulation-like loss of basement membrane, which may provide an opportunity for the genetic and/or epigenetic altered epithelial cells to undergo transformation and increase the risk of ovarian tumorigenicity.

thumbnail image

Figure 6. Model for the regulatory roles of gonadotropins in ovulation and epithelial transformation. (A) Gonadotropins initiate ovulation by stimulating the expression of cyclooxygenase-2 (Cox-2) in ovarian surface epithelial and follicle cells. Cox-2 is required and mediates the proteolytic degradation of basement membranes of both the surface and follicle wall. Prostaglandins produced by Cox-2 may suppress the production of basement membrane components (collagen IV and laminin). The loss of basement membranes and tissue remodeling result in the ultimate rupture in ovulation. (B) We hypothesize that a high level of gonadotropins may stimulate the Cox-2–dependent proteolytic activity, leading to loss of basement membrane without ovulation, especially in postmenopausal women. The frequent loss of the basement membrane of the ovarian surface epithelium may increase the likelihood of genetic and epigenetic alteration of the epithelial cells and, subsequently, the risk of ovarian tumorigenicity. We hypothesize that this mechanism may underlie the carcinogenic activity of gonadotropin stimulation and the protective function of Cox-2 inhibitors for ovarian carcinoma. CNS: central nervous system; GnRH: gonadotropin releasing hormone; FSH: follicle-stimulating hormones; LH: luteinizing hormone; MMP: matrix metalloproteinase.

Download figure to PowerPoint

Hypothesis: Cox-2 Expression and Loss of Basement Membrane Underline the Etiology of Cancer Risk by Gonadotropin Stimulation

Epidemiologic evidence suggests that the risk of ovarian carcinoma is associated with the number of ovulatory events.72 Two major theories, namely, the incessant ovulation hypothesis73 and the gonadotropin stimulation hypothesis,74 have been postulated to explain the cancer risk association. Laboratory experiments demonstrated the neoplastic transformation of rat ovarian surface epithelial cells by simply passaging the cells in culture in vitro, providing experimental support for the hypothesis.75, 76 However, ovarian neoplasms often arise from cells lining inclusion cysts or deep invaginations.8, 9, 17–19 Once trapped inside, these precursor lesions may no longer be affected by surface events during subsequent ovulations.

The competing theory supported by the same epidemiologic evidence is the gonadotropin stimulation hypothesis,74 which postulates that the surges of pituitary gonadotropins (FSH and LH) that initiate each ovulation also stimulate the ovarian surface epithelium and induce cell transformation. The speculated role of gonadotropins also is consistent with the finding that ovarian carcinoma occurs most frequently in postmenopausal women, when ovulation ceases but plasma gonadotropin levels are elevated.72 The gonadotropin stimulation hypothesis has been received less enthusiastically, because the hormones have unremarkable effects on the growth of ovarian surface epithelial cells in culture77, 78 and because a conceivable mechanism is lacking. Nevertheless, it is known that receptors for FSH and LH are present in the ovarian surface epithelial cells and that the cells are responsive to gonadotropins.78, 79

In preovulatory stimulation of the ovarian surface epithelium by gonadotropins (LH and FSH), the basement membrane of the ovarian surface is lost. This is similar to the basement membrane of the follicles,28, 47 due to both degradation and suppression of collagen IV synthesis.25–29, 71 Even after cessation of ovulation, the high levels of gonadotropins immediately after menopause may still stimulate the loss of basement membrane of the surface and inclusion cysts of ovarian epithelium. We propose that high Cox-2 expression induced by gonadotropins may underlie the carcinogenic activity of gonadotropin stimulation and that subsequently, Cox-2 may stimulate the loss of basement membrane of ovarian surface epithelia, providing a predicament for neoplastic transformation (Fig. 6B). This hypothesis may be worthy of further testing in animal models and in human clinical studies to evaluate the protective function of Cox-2 inhibitors with respect to ovarian carcinoma risk and the proposed mechanism.


  1. Top of page
  2. Abstract
  6. Acknowledgements

The authors thank Dr. Elizabeth Smith for reading and commenting during the process of preparing the article. They have been greatly assisted by Dr. Andres Klein-Szanto and the Histopathology Core (Core B of FCCC Ovarian SPORE) in tissue analysis. They also thank Malgorzata Rula, Jennifer Smedberg, Carolyn Slater, and Lisa Vanderveer for their technical assistance and Ms. Patricia Bateman for her excellent secretarial support.


  1. Top of page
  2. Abstract
  6. Acknowledgements