αv Integrin Expression Is a Novel Marker of Poor Prognosis in Advanced-stage Ovarian Carcinoma

The ability of malignant tumor cells to locally invade host tissue and metastasize to distant organs is a major determinant in the process of tumor progression. Invasion and subsequent dissemination require loss of adhesion between tumor cells, the ability to bind to and degrade basement membranes, and active interaction of tumor cells with the ECM.³ Integrins, a family of heterodimeric glycoproteins involved in these processes, are composed of α and β subunits (1). Integrins participate in both cell-cell adhesion and in binding to a large number of basement membrane and ECM ligands, including laminin, fibronectin, collagen, vitronectin, entactin, tenascin, and fibrinogen (1). Seventeen α and eight β subunits are known to date, forming 22 different combinations (2). Most α subunits associate with a single β subunit, the largest family being the very late activating β1 integrin family. This family includes the α5β1 fibronectin receptor, the αvβ1 fibronectin receptor, and the α6β1 laminin receptor (3). Several α subunits, including α4, α6, and αV, can in turn associate with more than one β subunit. Thus, αv integrin is part of the receptors for fibronectin, vitronectin, fibrinogen, and several other proteins (3).

The roles of integrins extend practically to all main aspects of cellular function. They form a structural link between the ECM and the cellular cytoskeleton through the formation of focal adhesions, where both structural proteins and protein kinases localize. Integrin-mediated signals are involved in cell proliferation, apoptosis, migration, invasion, cellular transformation, angiogenesis, and tumor progression (Refs. 2, 3, 4, 5, 6 and references therein). These effects are closely linked to the activation of metastasis-associated molecules, such as MMPs (7). Altered expression of integrins, in the form of down- or up-regulated expression, have been detected in the majority of malignant tumors but vary considerably according to the origin of the neoplasm (3).

Ovarian cancer is the sixth most common cancer and the sixth most frequent cause of cancer death in women (4.4% of cases and 4.5% of deaths). It is the leading cause of death from gynecological cancer in women in industrialized countries (8). The incidence of ovarian carcinoma appears to be increasing in Western countries, as evidenced by a 30% rise in incidence and an 18% rise in death rate in the United States (8). Despite the inclusion of new chemotherapeutic regimens, the mortality rate from ovarian carcinoma has remained largely unchanged. This results from the late clinical presentation of this tumor, two-thirds of the patients being diagnosed with stage III or IV disease (9).

The expression and the functional role of integrins have been investigated previously in ovarian carcinoma (10, 11, 12, 13, 14), but the prognostic role of these molecules remains largely unknown. In a single study of 19 borderline tumors and 31 carcinomas, no correlation was observed between protein expression of αvβ3 integrin and disease outcome after a mean follow-up period of 26 months (15). We have begun recently a prognostic study of metastasis-related molecules in a cohort of advanced-stage ovarian carcinoma patients, with a follow-up period of ≤20 years. We reported previously the role of mRNA expression of MMPs (MMP-2, MMP-9, and MT1-MMP) and their inhibitor TIMP-2 (16), as well as that of expression of the carbohydrate antigen Sialyl Lewis^x (17), the E-cadherin adhesion complex molecule γ-catenin (plakoglobin; Ref. 18), and the transcription factor Ets-1 (19), as predictors of poor survival. Conversely, protein expression of the membrane protein caveolin-1 (20) and mRNA expression of the angiogenic factors basic fibroblast growth factor, interleukin-8, and vascular endothelial growth factor (21) did not influence the prognosis of patients in this cohort.

The objective of the present study was to evaluate the prognostic role of αv and β1 integrin mRNA expression in advanced-stage ovarian carcinoma. Protein expression of both chains was studied to consolidate mRNA ISH results.

The study population consisted of 34 patients, diagnosed with advanced-stage (Fédération Internationale des Gynaecologistes et Obstetristes III-IV) epithelial carcinoma of the ovary in the Division of Gynecological Oncology at the Sheba Medical Center in the period between 1977 and 1997. The patient cohort was retrospectively selected for good and poor outcomes. The study cohort was thus divided into two groups, consisting of 16 and 18 patients, defined as long-term and short-term survivors, respectively, using a double cutoff of 36 months for disease-free survival and 60 months for overall survival. Inpatient and outpatient charts were available for review for all patients. No patients were lost to follow-up. All patients underwent surgery, followed by standard chemotherapy protocols. Until 1995, patients received adjuvant chemotherapy, including Cisplatinum and Cyclophosphamide. Since 1995, Paclitaxel has replaced Cyclophosphamide.

Formalin-fixed, paraffin-embedded blocks (56) from the archives of the Department of Pathology at the Sheba Medical Center were included in the study. These consisted of 33 primary ovarian tumors and 23 metastatic lesions from 34 patients with advanced ovarian carcinoma. The distribution of the studied material according to biopsy site is shown in Table 1. Sections from all tumors were reviewed by two observers (B. D. and J. K.) in consensus sessions to confirm the diagnosis, histological type, and tumor grade (I-III, corresponding to well, moderately, and poorly differentiated). Established criteria were used for the microscopic diagnosis and tumor classification (22). Tumor staging was established according to Fédération Internationale des Gynaecologistes et Obstetristes criteria (22).

Specific antisense oligonucleotide DNA probes for the mRNA transcript of the αv and β1 integrin chains were obtained from Biognostik (Göttingen, Germany). Probe sequences (5′-3′) were as follows:

αv integrin: TGA CCT TGC CAA TAA AAG CTA CCA GGA CC

β1 integrin: CCA AGT TTC CCA TCT CCA GCA AAG TGA AAC

A poly d(T)20 oligonucleotide (Biognostik) was used to verify the integrity and lack of degradation of mRNA in each sample.

Tissue sections (4-μm thick) of formalin-fixed, paraffin-embedded specimens were mounted on silane-coated slides. Sectioning was performed in RNase-free water. Slides were dewaxed and rehydrated using xylenes (2 × 10 min) and isopropyl alcohol (5 min). Hybridization was carried out as described previously (23). A positive enzymatic reaction in this assay stained dark blue. Known positive controls were used in each hybridization reaction. These consisted of two cases for which positive hybridization was reproducible in pilot studies. Controls for endogenous alkaline phosphatase included treatment of the sample in the absence of the probe and use of chromogen alone.

Strong dark blue staining was interpreted as intense. A minimum of 500 cells, when present, was evaluated. Staining was scored in carcinoma and stromal cells. Staining extent was scored as 0, 1, or 2, using a cutoff of 20%. Staining of ≤20% of tumor/stromal cells was scored as focal (1), whereas staining of >20% of cells was interpreted as diffuse (2). Staining intensity was scored as absent (0), weak/moderate (1), or intense (2). Neither the experimenter performing the tests nor the one evaluating the slides were aware of disease outcome or any other clinical parameter concerning the patients.

Fifty-two specimens (22 from long-term survivors and 30 from short-term survivors) were evaluated for protein expression of αv and β1 integrin chains. An antibody directed against αv integrin was obtained from Chemicon International (Temecula, CA). The antibody directed against β1 integrin was obtained from Serotec (Kidlington, Oxford, UK). Formalin-fixed, paraffin-embedded sections, 4-μm thick, were mounted onto silane-coated slides. After air drying at 37°C for 24 h, slides were deparaffinized and rehydrated. Staining was performed using the labeled Avidin Biotin method. For both antibodies, pretreatment in the form of protein digestion using Trypsin (10 min at 37°C) was used. Negative controls consisted of sections that underwent a similar staining procedure, with the exclusion of primary antibody application. Positive controls consisted of cases in which immunoreactivity was demonstrated in pilot studies.

ISH results in tumor cells and stromal cells were evaluated statistically, applying the SPSS-PC package (version 9.0; SPSS, Chicago, IL). Probability of <0.05 was considered statistically significant. Analyses of the association between ISH results and biopsy site, patient group, and tumor grade were executed using the two-sided χ² test. Univariate survival analyses for all specimens studied were executed using the Kaplan-Meier method and Log-rank test. Both staining extent and intensity were analyzed. Multivariate analyses of survival for all specimens studied were performed using Cox regression model. The parameters included were the expression of Sialyl Lewis^x and γ-catenin protein in tumor cells; mRNA expression of MMP-2, MMP-9, MT1-MMP, TIMP-2, and Ets-1 in tumor and stromal cells; and αv integrin expression in tumor cells (see also “Results”). Clinical parameters (patient age, tumor type, grade of differentiation, and disease stage) were not included, as their prognostic role was nullified by the choice of patients for this study (see also “Results”).

Patient age ranged from 30 to 84 years, with a mean age of 55 and 59 years in the long-term and short-term survivor group, respectively. Patients (28) were diagnosed with stage III tumors and 6 patients with stage IV tumors. These were equally represented in the two patient groups. Thus, the long-term survivor group included 13 patients diagnosed with stage III tumors and 3 with stage IV tumors, whereas the short-term survivor group included 15 and 3 tumors, respectively. Follow-up period ranged from 11 to 224 months (mean = 71 months). Mean disease-free survival and overall survival data, as well as disease status, are presented in Table 2.

The primary tumor diameter was comparable for both patient groups, having ranged from 2.2 to 15 cm in the long-term survivor group and from 3 to 16 cm in the short-term survivor group. The distribution of tumors according to histological type is shown in Table 1. The distribution of metastatic sites was comparable for all histological types. Tumor differentiation was as follows: 4 grade I tumors, 2 grade II tumors, and 28 grade III tumors. The fraction of poorly differentiated (grade III) tumors was comparable for both groups (12 of 16 and 16 of 18 tumors).

A positive signal using a poly d(T) probe was detected in all cases (Fig. 1-A). Negative controls showed only contrast staining by nuclear fast red (Fig. 1-B). αv and β1 integrin mRNA was detected in tumor and/or stromal cells (Fig. 1-C–F). Expression of αv integrin mRNA was detected in carcinoma cells and stromal cells in 18 of 56 (32%) and 17 of 56 (30%) of the lesions, respectively. β1 integrin mRNA expression was detected in carcinoma cells and stromal cells in 25 of 56 (47%) and 19 of 56 (34%) of the lesions, respectively (Tables 3 and 4). αv integrin expression was more diffuse in both tumor and stromal cells in metastases, whereas intense signal was more often seen in tumor cells of primary tumors (Tables 3,a and 4,a). None of these differences achieved statistical significance (P > 0.05). The expression of β1 integrin mRNA was comparable in primary and metastatic lesions in terms of both intensity and extent (P > 0.05; Tables 3,a and 4,a). αv and β1 integrin mRNA expression in stromal cells was comparable in tumors of long-term and short-term survivors (P > 0.05; Tables 3,b and 4,b). However, marked differences were observed in tumor cells. Diffuse and intense labeling for αv integrin mRNA was detected more often in carcinoma cells in tumors of short-term survivors (16 versus 8% for intensity, 27 versus 0% for extent), significantly so for labeling extent (P = 0.017; Fig. 1-E and 1-F; Tables 3,b and 4,b). Similarly, intense and diffuse expression of β1 integrin in carcinoma cells was more frequently seen in tumors of short-term, as compared with long-term, survivors (20 versus 8% for intensity, 37 versus 19% for extent; Tables 3,b and 4 b). However, these findings failed to reach significance (P > 0.05).

As for mRNA level, αv and β1 integrin protein expression was detected in both tumor and stromal cells (Fig. 1-G and 1-H). Frequent labeling of endothelial cells for β1 integrin was also seen (Fig. 1-I). Protein expression results showed good correlation with those of mRNA expression using ISH. Thus, αv integrin was detected in 23 of 30 (77%) specimens obtained from short-term survivors but only in 10 of 22 (45%) tumors of long-term survivors. It was in general more diffuse and intense in the former group. Expression in fibroblasts was practically universal in tumor-positive cases. In contrast, β1 integrin protein expression was comparable in both patient groups (14 of 30 tumors of short-term survivors, 11 of 21 tumors from long-term survivors; 47 and 52%, respectively). Stromal immunoreactivity for β1 integrin was less frequent than that observed for tumor cells.

In univariate survival analysis of all cases, diffuse (>20% of cells) αv mRNA integrin expression in tumor cells correlated with poor survival (P = 0.012; Fig. 2, -A and -B). αv integrin expression retained its predictive power as a marker of poor survival in a multivariate survival analysis, in which all molecules studied previously in this patient cohort were included (P = 0.031), together with TIMP-2 mRNA expression in stromal cells (P = 0.005) and the expression of MMP-2 (P = 0.004), Sialyl Lewis^x (P = 0.007), γ-catenin (P = 0.003), and Ets-1 (P = 0.043) in tumor cells.

Few human cancers are characterized by distant spread as frequent or extensive as that of ovarian carcinoma. Although this characteristic can be partially explained in light of their late detection, the largely predictable spreading sequence of these tumors, primarily centered on the peritoneal cavity, raises questions regarding the molecules involved in this process. Integrins are natural candidates for this investigation because of their role as ligands for all major constituents of the basement membrane and ECM. Indeed, protein expression of α2 and β1 integrins have been shown to mediate adhesion of ovarian carcinoma cells to collagen type I (10). An additional report documented the role of αvβ3 and α5β1 integrins in binding to fibronectin and that of the β1 chain in binding to laminin and collagen I (13). Protein expression of αv and β3 chains and their ligand vitronectin has been additionally shown in normal ovarian surface cells (11, 13). Analyses of integrin expression focusing on clinical specimens of ovarian carcinoma have been, however, few to date. αvβ3 integrin protein expression was shown to be lower in tumors of low malignant potential, as compared with invasive carcinomas (15), inconsistent with a previous study, in which tumors of the low malignant potential category were grouped together with grade I invasive carcinomas and compared with grade II-III carcinomas (12). Expression of α3β1 integrin has also been shown in 26 of 31 ovarian carcinomas (14). None of the above-mentioned studies evaluated αv and β1 mRNA integrin expression in ovarian carcinoma, and none scored integrin expression in peritumoral fibroblasts. Our findings suggest that synthesis of αv and β1 integrin chains takes place in both carcinoma cells and stromal cells, although it is more prominent in tumor cells. Synthesis in fibroblasts may thus play a role in interaction with ECM ligands in much the same way as it does in cancer cells and could lead to activation of the synthesis of metastasis-associated molecules, such as MMPs. Our recent report, in which MMP mRNA was localized to both tumor and stromal cells (16), supports this hypothesis.

Studies of integrin expression in ovarian carcinoma have focused mainly on primary tumors to date (12, 14, 15). However, Moser et al. (10) and Cannistra et al. (13) analyzed cultures of cells obtained from ascites. No comparative studies of primary and metastatic tumors are available. Our results disclose no significant differences in the expression of αv and β1 integrins at primary and metastatic sites. On the basis of these findings, integrin synthesis appears to be an early event in tumor progression in ovarian carcinoma, in agreement with the role of these molecules in local invasion. The detection of αv integrin reported previously in ovarian surface cells supports the possibility of up-regulation in the synthesis of a molecule constitutively expressed in the benign ovarian epithelium.

The prognosis of ovarian carcinoma remains poor, partially because of its late detection, often associated with widespread i.p. disease. Although a considerable amount of data has been acquired by in vitro studies of ovarian carcinoma cells, the prognostic role of the most important metastasis-associated molecules in this malignancy remains unknown. Our study evaluated two groups of patients diagnosed with advanced-stage ovarian carcinoma with a markedly different disease outcome, with a follow-up period of ≤20 years. Established prognostic factors, such as age, stage, grade, and tumor type, were all controlled by patient selection criteria in the design of the study. This selection was meant to facilitate the study of potential prognostic markers. Studying this patient cohort, we have identified recently the prognostic role of mRNA expression of MMPs (MMP-2, MMP-9, and MT1-MMP) and their inhibitor TIMP-2 (16). The carbohydrate antigen Sialyl Lewis^x (17), the adhesion complex molecule γ-catenin (plakoglobin; Ref. 18), and the transcription factor Ets-1 (19) were similarly found to predict poor survival. Conversely, protein expression of the membrane protein caveolin-1 (20) and mRNA expression of the angiogenic factors basic fibroblast growth factor, interleukin-8, and vascular endothelial growth factor did not predict the prognosis of patients in this cohort (21).

In the present study, expression of αv integrin mRNA was a predictor of poor survival, a finding that retained its power in a multivariate survival analysis. The larger number of IHC-positive cases in both groups, as compared with mRNA results, may be attributable to the pretreatment of sections or the inherent sensitivity of the antibodies. Nevertheless, the differences between the two prognostic patient groups were similarly observed on protein level, supporting our findings using ISH. The prognostic significance of integrin expression has to date been investigated in a single study (15). Liapis et al. (15) evaluated the prognostic role of αvβ3 integrin expression in 31 ovarian carcinomas and found no correlation between protein expression and disease outcome after a mean follow-up period of 26 months. However, 13 of 23 immunoreactive tumors proved fatal, as compared with 2 of 8 negative tumors (15). The absence of correlation with survival in the latter study may be attributed to the short follow-up period, possibly combined with a small number of cases. The cohort studied by us was followed up for an average period of 71 months, in effect considerably longer for nonlethal cases. It is noteworthy that two studies of αv integrin expression in lung carcinoma have led to inconclusive results. Clarke et al. (24) reported an association (though nonsignificant) between protein expression of αv integrin and nodal metastasis in a study of 31 carcinomas, whereas Smyth et al. (25) reported a significant association between loss of expression of this integrin and nodal metastasis. In an additional study, expression of αvβ3 integrin in vessels predicted disease relapse in breast carcinoma (26). To our best knowledge, our findings present the first evidence linking αv integrin with poor survival in human epithelial malignancy. αv integrin is thus a novel prognostic marker in ovarian carcinoma.

Although failing to reach significance, the more frequent expression of β1 integrin mRNA in carcinoma cells in tumors of short-term survivors merits additional attention, as larger studies may reveal a similar role for this integrin. However, the similar expression on protein level in tumors from the two groups argues against a central role in determining disease outcome in ovarian carcinoma. The more pronounced expression of both αv and β1 integrin chain mRNA, as well as αv protein in carcinoma cells rather than in stromal cells, suggests that tumor cell, rather than stromal production of these molecules, is central in the biology of ovarian carcinoma.

We thank Ellen Hellesylt, Bruno Guggiana, and Asle Bjåmer at the Department of Pathology, The Norwegian Radium Hospital, for their technical help. We also thank Ula Bartenstein for her help with the ISH analysis.