Clinical Significance of Heparin-Binding Epidermal Growth Factor–Like Growth Factor and A Disintegrin and Metalloprotease 17 Expression in Human Ovarian Cancer

Ovarian cancer is the most common cause of death from a gynecologic malignancy in most countries (1). The high mortality is predominantly due to occult progression of the tumor in the peritoneal cavity, with the initial diagnosis usually being made at an advanced stage. Currently, ∼75% of ovarian cancers are diagnosed at International Federation of Gynecology and Obstetrics stages III and IV (2). Extensive dissemination of a tumor is caused by the peritoneal fluid following the circulatory pathway in the abdominal cavity, and the peritoneal fluid acts as a rich source of growth factor activity for ovarian cancer cells (3). Thus, the dissemination of cancer cells activated by ovarian cancer–activating factors results in an exaggerated increase in peritoneal fluid, which in turn leads to tumor extension in ovarian cancer. Therefore, to develop a targeting therapy, it would be extremely useful to understand the ovarian cancer–activating factor–mediated molecular mechanisms for activating ovarian cancer cells.

Lysophosphatidic acid (LPA) is a simple phospholipid with numerous cellular effects, including growth promotion, cell cycle progression, and cytoskeletal organization (4), and is generated from precursors in membranes. LPA is elevated in the plasma and peritoneal fluid from patients with ovarian cancer in all stages, suggesting that it is a possible candidate for an ovarian cancer–activating factor (5–8). In principle, LPA-induced signaling is mediated by G protein–coupled receptors, including LPA1, LPA2, LPA3, and LPA4 (4). Recent investigations have shown that G protein–coupled receptors are able to use the epidermal growth factor receptor (EGFR) as a downstream signaling partner in the generation of mitogenic signals (9), and EGFR has been recognized to play a pivotal role in the progression of ovarian cancer (10, 11). According to this evidence, it can be considered that EGFR signal transactivation induced by LPA may contribute to the promotion of a tumor in ovarian cancer.

The molecular mechanisms of EGFR signal transactivation involve processing of transmembrane growth factor precursors by metalloproteases, which have been identified as members of the ADAM (a disintegrin and metalloprotease) family of zinc-dependent proteases (9). Seven-transmembrane growth factor precursors have been described as ligands for EGFR: EGF, heparin-binding EGF-like growth factor (HB-EGF), amphiregulin, transforming growth factor-α (TGF-α), betacellulin, epiregulin, and epigen (12, 13). For the metalloproteases, there have been at least 34 adam genes described in a variety of species, and ADAM 9, 10, 12, 17, and 19, which are ubiquitously expressed in somatic tissues, have sheddase activity (14). In particular, ADAM 9, 10, 12, and 17 are involved in the ectodomain shedding of EGFR ligands (15–21). The enhancement of EGFR signal transactivation mediated by EGFR ligands and the ADAM family is linked to the pathogenesis of hyperproliferative disorders, such as cancer. Previously, we shown that HB-EGF is involved in EGFR signal transactivation induced by LPA in ovarian cancer cell lines and that the expression of HB-EGF is attributable to tumor growth on xenografted mice using ovarian cancer cell lines (22). However, no studies have yet comprehensively examined the clinical significance of EGFR ligands and ADAM family expression in human cancers.

To investigate which molecules involved in EGFR signal transactivation are associated with human ovarian cancer, we examined the expression of EGFR ligands and ADAM family members in patients with ovarian cancer, using real-time PCR, and analyzed the clinical significance of these molecules in ovarian cancer.

Patients and surgical specimens. All 108 patients in this study had undergone surgery at the Department of Obstetrics and Gynecology, Kyushu University Hospital (Fukuoka, Japan) between January 1996 and August 2003. All the ovarian cancer specimens were obtained from 68 patients, comprising 16 cases with International Federation of Gynecology and Obstetrics stage I ovarian cancer, 10 cases with stage II, 29 cases with III, and 13 cases with stage IV. None of the patients had received chemotherapy before surgery. After dissection, half of each fresh tumor tissue specimen was immediately snap frozen in liquid nitrogen and stored at −80°C until use, whereas the other half was immediately embedded for the production of frozen or paraffin sections. Diagnosis was based on conventional morphologic examination of paraffin-embedded specimens, and tumors were classified according to the WHO classification (23). Metastases of pelvic lymph nodes in all cases and para-aortic lymph nodes in 60 cases were assessed by pathologic examination using paraffin-embedded specimens after surgery. In 8 cases, metastases of para-aortic lymph nodes were evaluated by the presence or absence of lymph node swelling in an abdominal computed tomographic scan before surgery because surgical specimens were not resected. The median follow-up periods for all patients were 30.0 months (range, 2-83 months) for overall survival and 20.0 months (range, 1-63 months) for progression-free survival. After debulking surgery, all 68 patients had platinum-based chemotherapy (median, 6.0 courses; range, 1-10 courses) as first-line chemotherapy. Normal ovarian tissue specimens were obtained at surgery for benign gynecologic disorders from 40 patients, comprising 12 premenopausal cases and 28 postmenopausal cases. Informed consent was obtained from all patients in this study.

Criteria for chemotherapy response and definition of progression-free survival interval. The response to chemotherapy induction was assessed by second-look surgery, clinical and/or radiographic evaluation according to the WHO criteria, or CA125 response using Rustin et al.'s criteria (23, 24). The progression-free interval was defined as the duration from the date at surgery to the final date observed in this study (March 31, 2004) or the duration from the date at surgery to the date when progression was diagnosed, according to the proposed definitions of progression by the Gynecologic Cancer Intergroup (25).

Preparation of RNA. To ascertain the presence of cancer cells, half of each fresh tumor tissue specimen was immediately embedded in Tissue-Tek OCT compound (Sakura, Tokyo, Japan). Frozen sections were cut on the cryostat to a thickness of 6 μm and immediately stained with H&E. More than 80% of any given tumor specimen, which contained cancer cells, were used for cDNA synthesis. RNA was extracted using TRIzol (Invitrogen Corp., Carlsbad, CA) according to the manufacturer's protocol. First-strand cDNA synthesis was done with 0.8 μg total RNA using SuperScript II reverse transcriptase (Invitrogen) following the manufacturer's protocol.

Performance of reverse transcription-PCR and real-time quantitative PCR for epidermal growth factor receptor ligands or a disintegrin and metalloprotease family members. Sense and antisense primers based on the nucleotide sequences of HB-EGF cDNA, TGF-α cDNA, amphiregulin cDNA, epiregulin cDNA, betacellulin cDNA, and EGF cDNA were used, and the PCR protocol for each EGFR ligand followed those described by Adam et al. (26) or Sorensen et al. (27). The PCR products were electrophoresed in 2% agarose gels, and the bands were visualized with ethidium bromide and photographed with a camera (Funakoshi, Tokyo, Japan). When no bands were detected, the number of amplifications was increased by 50 cycles. Real-time PCR (TaqMan PCR) was done using an ABI PRISM 7000 Sequence Detection System (Applied Biosystems, Foster City, CA) as described previously (28). The sequences of the oligonucleotide primer pairs and TaqMan probes for each EGFR ligand and ADAM family member are summarized in Table 1. Serial 1:10 dilutions of plasmid DNA containing each target cDNA (10⁷-10¹ copies/μL) were analyzed and served as standard curves, from which we determined the rate of change of the threshold cycle values. The correlation coefficients of the standard curves were >0.995, thus ensuring the accuracy of our data. Plasmid DNA played the role of a positive control for each reaction. Copy numbers of the target cDNAs (HB-EGF, amphiregulin, TGF-α, epiregulin, betacellulin, ADAM 9, ADAM 10, ADAM 12, and ADAM 17) were estimated from the standard curves. All reactions for the standard and patient samples were done in triplicate, and the data were averaged from the values obtained in each reaction. To determine the mRNA levels of four EGFR ligands and four ADAM family members, we used the mRNA expression index, which is a relative mRNA expression level standardized by glyceraldehyde-3-phosphate dehydrogenase. The mRNA expression index was calculated as follows (in arbitrary units): mRNA expression index = (copy number of each EGFR ligand or each ADAM family member mRNA / copy number of glyceraldehyde-3-phosphate dehydrogenase mRNA) × 10,000 arbitrary units. When the expression index was over the maximal value in patients with normal ovaries, it was regarded as a high expression status of the molecule under analysis.

In situ hybridization.In situ hybridization was done as described previously (29). Briefly, digoxigenin-labeled antisense and sense riboprobes were generated by in vitro transcription using a DIG RNA Labeling kit (SP6/T7; Roche Diagnostics GmbH, Penzberg, Germany) according to the manufacturer's instructions. The frozen samples were sectioned to a thickness of 6 μm. After fixation in 4% paraformaldehyde for 10 minutes, slides were immersed in 0.2 mol/L HCl for 10 minutes and then rinsed in phosphate buffer. Acetylation was done in 0.1 mol/L triethanolamine in 0.25% acetic anhydride for 15 minutes. After rinsing in phosphate buffer, sections were dehydrated in an ethanol gradient and dried. Sections were hybridized with the HB-EGF probe (diluted 1:10) at 55°C overnight. After high-stringency washing, the signal was visualized using an alkaline phosphatase–conjugated anti-DIG antibody (Roche Diagnostics). Three examiners (Y.T., S.M., and K.S.) separately evaluated the HB-EGF mRNA staining by cell counting. At least 20 high-magnification fields were chosen randomly, and 1,000 cells in total were counted.

Immunohistochemistry. Immunohistochemistry was done on frozen sections using a goat polyclonal antibody against HB-EGF (R&D Systems, Inc., Minneapolis, MN) and on formalin-fixed, paraffin-embedded sections using a goat polyclonal antibody against ADAM 17 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). Frozen sections were cut on a cryostat to a thickness of 6 μm, mounted on poly-l-lysine-coated slides, and either used immediately or stored at −80°C until needed. Paraffin-embedded sections were cut on a microtome to a thickness of 4 μm, mounted on poly-l-lysine-coated slides, and then dewaxed and rehydrated through xylene, graded ethanol solutions (100%, 90%, and 70%), and water. Briefly, the frozen and paraffin-embedded sections were subsequently immersed for 30 minutes in 0.3% H₂O₂ in absolute methanol, treated with 5% normal rabbit serum for 30 minutes, and incubated with the primary antibody against HB-EGF or ADAM 17 overnight at 4°C. The sections were then incubated with biotinylated rabbit anti-goat IgG (Nichirei Corp., Tokyo, Japan) for 30 minutes followed by an avidin-biotin-peroxidase complex solution. The peroxidase reaction was developed using 3,3′-diaminobenzidine tetrahydrochloride in the presence of 0.05% H₂O₂, and the sections were then counterstained in Mayer's hematoxylin, washed in tap water, dehydrated in graded ethanol, cleared in xylene, and coverslipped. Control staining was done using nonimmune goat IgG as the primary antibody. Three examiners (Y.T., S.M., and K.S.) separately evaluated the HB-EGF and ADAM 17 staining by counting the immunoreactive cells. At least 20 high-magnification fields were chosen randomly, and 1,000 cells in total were counted.

Statistical analysis. Statistical analysis was done with StatView software version 5.0 (Abacus Concepts, Berkeley, CA). The Mann-Whitney test was done to test the equality of the distribution of age and the mRNA expression index of five EGFR ligands and four ADAM family members among patients with normal ovaries, early ovarian cancer, and advanced ovarian cancer. Progression-free survival curves were estimated using the Kaplan-Meier method and analyzed by the log-rank test. Correlation between the mRNA expression indices of molecules was analyzed using Spearman's correlation analysis. Statistical significance was based on two-tailed statistical analyses, and Ps < 0.05 were considered statistically significant.

Expression of epidermal growth factor ligands in normal ovaries and ovarian cancer. Using real-time PCR, large differences in the mRNA expression index were found for HB-EGF and four other EGFR ligands between normal ovaries and ovarian cancer (Fig. 1; Table 1). For HB-EGF, the mRNA expression index was significantly elevated in advanced ovarian cancer compared with that in normal ovaries, although there was no significant difference between early ovarian cancer and normal ovaries (Fig. 1; Table 1). For TGF-α, amphiregulin, epiregulin, and betacellulin, no significant differences in the mRNA expression index were found among the three groups (Fig. 1; Table 1). No clear expression of EGF was detected in 10 patients with normal ovaries or 30 patients with ovarian cancer by reverse transcription-PCR, although EGF expression was confirmed in human placenta tissue using the same primer sets (26). Therefore, real-time PCR for EGF was not done in this study. To further investigate the expression of HB-EGF in surface normal ovarian epithelial cells, we examined the expression index of HB-EGF using 10 samples extracted by brushing normal ovarian epithelial cells. The expression index of HB-EGF mRNA was 7.6 ± 7.9 (mean ± SE), which was not significantly changed from that in samples extracted from whole normal ovaries. In a cancerous state, the expression of HB-EGF significantly increased compared with that in a normal state. These results suggest that HB-EGF contributes to the progression of ovarian cancer among the EGFR ligands.

Expression of a disintegrin and metalloprotease family members in normal ovaries and ovarian cancer. A large difference in the mRNA expression index was found for ADAM 17 and three other ADAM family members in normal ovaries and ovarian cancer (Fig. 2; Table 1). For ADAM 17, the mRNA expression index in early or advanced ovarian cancer was significantly elevated compared with that in normal ovaries, and there was no significant difference between early and advanced ovarian cancers (Fig. 2; Table 1). For ADAM 9, 10, and 12, no significant differences in the mRNA expression index were found among the three groups (Fig. 2; Table 1). To further investigate the expression of ADAM 17 in normal surface ovarian epithelial cells, we examined the expression index of ADAM 17 using 10 samples extracted by brushing normal ovarian epithelial cells. The expression index of ADAM 17 mRNA was 440 ± 380 (mean ± SE), which was not significantly changed from that in samples extracted from whole normal ovaries. In a cancerous state, the expression of ADAM 17 significantly increased compared with that in a normal state. Taken together, these results suggest that ADAM 17 is involved in the occurrence of ovarian cancer.

Clinical significance of heparin-binding epidermal growth factor–like growth factor or a disintegrin and metalloprotease 17 expression in ovarian cancer. HB-EGF expression ranged from a mRNA expression index of 10 to 39 in patients with normal ovaries. Ovarian cancer patients with a HB-EGF mRNA expression index of >40 were therefore regarded as cases with a high expression status of HB-EGF. In the progression-free survival curves, ovarian cancer patients with a high expression status of HB-EGF showed a significantly less favorable prognosis than those with a low expression status (Fig. 3A). Taken together, these results suggest that HB-EGF plays a significant role in the progression of ovarian cancer. ADAM 17 expression was within a mRNA expression index of 999 in patients with normal ovaries. Ovarian cancer patients with an ADAM 17 mRNA expression index of >1,000 were therefore regarded as cases with a high expression status of ADAM 17. No significant difference in the progression-free survival curves was found between ovarian cancer patients with a low and high expression status of ADAM 17 (Fig. 3B), suggesting that ADAM 17 is not significantly associated with the clinical outcome.

Relationships between the expression indices of heparin-binding epidermal growth factor–like growth factor and a disintegrin and metalloprotease family members in ovarian cancer. Significant correlations were found between the mRNA expression indices for HB-EGF and ADAM 9, 12, and 17 (Fig. 4A,, B, and D). The mRNA expression index of ADAM 10 showed no significant correlation with that of HB-EGF in ovarian cancer (Fig. 4C). The correlation coefficient between the mRNA expression index of HB-EGF and that of ADAM 17 was increased compared with those between HB-EGF and ADAM 9 or 12, suggesting that ADAM 17 has a more significant contribution to the ectodomain shedding of HB-EGF than ADAM 9 or 12.

Localization of heparin-binding epidermal growth factor–like growth factor and a disintegrin and metalloprotease 17 proteins in normal ovarian epithelial cells and ovarian cancer cells. Abundant HB-EGF protein appeared as positive in interstitial tissues surrounding the ovarian cancer cells, whereas no definite expression of HB-EGF was observed in normal ovarian epithelial cells or interstitial tissues (Fig. 5A and B). In in situ hybridization, diffuse staining for HB-EGF mRNA was only found in ovarian cancer cells and not in either interstitial cells surrounding the cancer cells or normal ovarian epithelial cells (Fig. 5C and D), suggesting that HB-EGF protein is only produced by cancer cells, and not by interstitial cells, and that the proteolytic form of HB-EGF accumulates in the extracellular matrix with heparin sulfate in the interstitial tissues surrounding the cancer cells. The correlation coefficient between the mRNA expression index of HB-EGF and the in situ hybridization score of HB-EGF was 0.876 in ovarian cancer (P < 0.001). In addition, positive staining for ADAM 17 was observed in cancer cells, whereas no cells showed positive expression of ADAM 17 in normal ovarian epithelium (Fig. 5E and F). In ovarian cancer, the correlation coefficient between the ADAM 17 immunostaining and the mRNA expression index of ADAM 17 was 0.839 (P < 0.01).

Impairment of the EGF system is involved in the pathogenesis of different types of carcinomas (10, 11). Univariate and multivariate statistical analyses have confirmed that EGFR overexpression is significantly associated with a high risk of progression in ovarian cancer patients (30). Relatively high frequencies of TGF-α and amphiregulin have been described in ovarian carcinomas, although the staining varied from weak to strong in tumors (31, 32). Ovarian cancer cells are sensitive to the diphtheria toxin, indicating the expression of pro-HB-EGF (33). No significant expression of EGF is present in normal ovaries or ovarian cancer. Thus, it remained unclear which EGFR ligands were predominantly expressed in ovarian cancer. In this study, however, abundant expression of HB-EGF was found in ovarian cancer compared with other EGFR ligands. Recently, several studies have revealed that HB-EGF is involved in a variety of cancers. In bladder cancer, HB-EGF is abundantly expressed and a significant prognostic marker for survival (34). HB-EGF expression is also associated with the clinical outcome in gastric, pancreatic, and breast cancers, in which HB-EGF expression is markedly abundant (35–37). In addition, Helicobacter pylori infection in human gastric carcinogenesis and the inflammatory processes associated with this type of infection have been linked to HB-EGF-dependent EGFR signal transactivation in human gastric epithelial tumor cells (38, 39). According to these studies, HB-EGF has been implicated in the occurrence and progression of human cancers. In this study, HB-EGF expression was significantly associated with the clinical outcome in ovarian cancer, suggesting that HB-EGF plays a crucial role in the aggressive behavior of a tumor in ovarian cancer.

The ADAM family has been implicated in diverse processes, including membrane fusion, cytokine and growth factor shedding, and cell migration (14). In particular, recent findings have revealed that the ADAM family is involved in cancer. ADAM 9 expression is associated with the clinical significance of human breast and pancreatic cancers (40, 41), whereas abundant ADAM 17 protein is expressed in human breast cancer (42). In human gastric carcinoma, high levels of transcripts for ADAM 10, 17, and 20 are present (43), whereas, in human liver cancer, expression of ADAM 9 and 12 is associated with tumor aggressiveness and progression (44). Thus, a few members of the ADAM family may be simultaneously associated with the acceleration and progression of human cancers. Therefore, any ADAM family members with similar functions should be examined to identify those involved in the pathogenesis of cancer. In this study, the expression of each ADAM family member involved in the ectodomain shedding of HB-EGF (15–21) was quantitatively estimated in human ovarian cancer. ADAM 17 was abundantly expressed compared with the other three ADAM family members, and its expression was enhanced in ovarian cancer. Therefore, this elevation of ADAM 17 expression in cancer might facilitate the proteolytic cleavage of EGFR ligands that are involved in the progression of cancer.

LPA can mediate EGFR signal transactivation through different combinations of EGFR ligands and ADAM family members. In NCI-H292 lung cancer cells, LPA transactivates EGFR through the ectodomain shedding of HB-EGF or amphiregulin, which is cleaved by ADAM 17 (45). In kidney cancer cells, EGFR transactivation is mediated by LPA, in association with HB-EGF and ADAM 10 or 17 (46). In bladder cancer cells, ADAM 15 has a role in EGFR transactivation mediated by LPA via soluble forms of amphiregulin or TGF-α (46). Thus, in the same cell system, there is a functional redundancy between EGFR ligands and ADAM family members that depends on a variety of stimuli. In ovarian cancer cells, HB-EGF and ADAM 17 were abundantly expressed compared with other EGFR ligands and other members of the ADAM family, respectively, and LPA activated EGFR through the ectodomain shedding of HB-EGF (22). In this study, the expressions of both HB-EGF and ADAM 17 were also abundant compared with those of other EGFR ligands and other members of the ADAM family in human ovarian cancer. In addition, HB-EGF protein appeared to accumulate in the interstitial tissues surrounding cancer cells and abundant ADAM 17 was also expressed in cancer cells, leading to the speculation that most HB-EGF expressed in cancer cells is quickly cleaved by ADAM 17. In fact, a large amount of HB-EGF was observed in the peritoneal fluid of ovarian cancer patients compared with the levels of amphiregulin and TGF-α (22). Taken together, these results suggest that proteolytic cleavage of HB-EGF was extensively provoked by ADAM 17 in human ovarian cancer.

This is the first study to show that both HB-EGF and ADAM 17 are significantly expressed among EGFR ligands and ADAM family members in human ovarian cancer. We have shown that tumor formation of ovarian cancer was completely blocked by pro-HB-EGF gene RNA interference and that the release of soluble HB-EGF is essential for tumor formation (22). Therefore, the development of therapeutic tools against HB-EGF and ADAM 17 would allow us to explore novel targeting therapy to human ovarian cancer.

We thank E. Hori and Y. Kubota for technical assistance and L. Saza and Drs. H. Baba, R. Tsunematsu, M. Tsuneyoshi, K. Sueishi, and T. Iwaki for critical comments.