Expression of Latent Matrix Metalloproteinase 9 (MMP-9) Predicts Survival in Advanced Ovarian Cancer

Gynecologic Oncology 82, 291–298 (2001) doi:10.1006/gyno.2001.6243, available online at http://www.idealibrary.com on

Expression of Latent Matrix Metalloproteinase 9 (MMP-9) Predicts Survival in Advanced Ovarian Cancer Ernst Lengyel,* ,† ,1 Barbara Schmalfeldt,* Elisabeth Konik,* Kerstin Spa¨the,* Kathrin Ha¨rting,* Anke Fenn,* Ursula Berger,‡ Rafael Fridman,§ Manfred Schmitt,* Dieter Prechtel, ¶ and Walther Kuhn* *Department of Obstetrics and Gynecology, ¶Department of Pathology, and ‡Department of Statistics, Technische Universita¨t Mu¨nchen, Klinikum rechts der Isar, Munich, D-81675 Germany; §Department of Pathology and Karmanos Cancer Institute, Wayne State University, Detroit, Michigan 48201; and †Department of Obstetrics, Gynecology & Reproductive Sciences, University of California, San Francisco, Comprehensive Cancer Center and Cancer Research Institute, San Francisco, California 94143 Received December 4, 2000; published online June 22, 2001

Objective. Matrix metalloproteinases (MMPs) are frequently expressed in malignant tumors and play an important role in tumor invasion and metastasis. MMP-2 and MMP-9 expression has been correlated with poor survival in some tumors, but data for ovarian cancer are lacking, despite clinical trials with MMP inhibitors. The aim of this study was to assess activity of MMP-2 and MMP-9 and correlate it to prognosis in ovarian cancer. Methods. MMP-2 and MMP-9 gelatinolytic activity was analyzed in 84 patients with advanced ovarian cancer FIGO stage III and 19 benign ovarian tumors by gelatin zymography. MMP-9 immunoreactivity was detected by immunohistochemistry and gelatinolytic activity was localized in ovarian cancer tissue by in situ zymography. Results were correlated with patient survival, with a median follow-up period of 55 months. Results. Median pro-MMP-9 activity was at 0.00 U/␮g protein in benign ovarian tissues and 4.82 U/␮g protein in ovarian cancer (P ⴝ 0.001); activated MMP-9 was not detected. Pro-MMP-2 expression in benign ovarian tissue did not differ from that of malignant ovarian tissue, whereas active MMP-2 was present in 52% of ovarian cancers, but absent in benign ovarian tissues. Analyzing all patients high pro-MMP-9 activity was associated with short overall survival (P ⴝ 0.019) while pro-MMP-2 and activated MMP-2 did not predict overall survival. When analyzing the subgroups of patients with and without residual tumor mass at the time of surgery, pro-MMP-9 was of prognostic value only in the subgroup of patients with no residual tumor mass. In univariate analysis pro-MMP-9 activity, residual tumor mass, age, ascites volume, and grading were of prognostic significance for overall survival. However, in multivariate analyses, including all biological and clinicopathologic variables, only pro-MMP-9 and residual disease remained statistically independent prognostic factors. In situ zymography localized gelatinolytic activity predominantly to the tumor cell nests displaying MMP-9 immunoreactivity. Conclusions. Pro-MMP-9 gelatinolytic activity, but not active

MMP-2 or MMP-9, serves as a useful statistically independent prognostic factor in ovarian cancer FIGO stage III, thus helping to identify ovarian cancer patients with an aggressive form of the disease. © 2001 Academic Press Key Words: matrix metalloproteinase 2; matrix metalloproteinase 9; ovarian cancer; zymography; prognosis.

INTRODUCTION

1 To whom correspondence and reprint requests should be addressed at the Department of Obstetrics, Gynecology & Reproductive Sciences, University of California, San Francisco, Comprehensive Cancer Center and Cancer Research Institute, 2340 Sutter Street, San Francisco, CA 94143-0875. Fax: (415) 514-0878. E-mail: [email protected].

Ovarian cancer is the most frequent cause of death by genital cancer in women. Because of their hidden anatomical location such tumors are often undetected unless metastasis has occurred, at which time the disease is difficult to eradicate. Most ovarian cancer cases are of epithelial origin, arising from the single layer of cells that covers the ovary [1] separated by a basement membrane from the underlying ovarian stroma, the tunica albuginea. It has been suggested that the repeated rupture of the surface epithelium during ovulation followed by rapid proliferation and repair may favor the transformation of these cells [2]. Once malignant, ovarian epithelial cells start to invade the ovary and metastasize by surface shedding followed by peritoneal implantation of ovarian cancer cells, which leads to the development of multiple metastatic nodules on the visceral and parietal peritoneum. One prerequisite for invasion and metastasis of ovarian cancer cells is the expression of proteolytic enzymes which degrade components of the basement membrane and the extracellular matrix (ECM) [3, 4]. Malignant cells produce a wide variety of matrix-degrading proteinases [5] that mediate the breakdown of the tissue scaffold. One important group of proteolytic enzymes is the matrix metalloproteinase (MMP) family, which consists of zinc-dependent endopeptidases capable of degrading components of the ECM and basement membrane [6 – 8]. MMPs are secreted as latent precursor proteins and are activated in the extracellular space. They are classified according to their substrate specificity and structure into subgroups of collagenases, stromelysins, membrane-type

291

0090-8258/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved.

292

LENGYEL ET AL.

MMPs, and type IV collagenases. Of special importance to tumor biology are the type IV gelatinases, M r 72,000 (MMP-2, gelatinase A) and M r 92,000 (MMP-9, gelatinase B) enzymes, because of their preference to degrade type IV collagen which is a major component of basement membranes. This degradation is an essential step in tumor invasion [9]. MMP-2 and MMP-9 have a similar catalytic profile, and besides their specificity for type IV collagen, they have potent gelatinolytic activity toward several other components of the ECM and the basement membrane including collagen types V, VII, IX, and X, fibronectin, and elastin. The important role of MMP-2 and MMP-9 in tumor invasion and metastasis is supported by experiments with cell cultures [10, 11] and animal models [9]. Consistent with their role in experimental metastasis, elevated levels of MMP-2 and/or MMP-9 were found in several types of human tumor to be related to the overall survival of patients afflicted with cancer of the stomach, colon, and prostate [12– 17]. However, the prognostic significance of MMP-2 and MMP-9 in ovarian cancer has not been evaluated so far [8]. Information from other types of cancers can only be partially inferred for ovarian cancer because the tumor biology of ovarian cancer is distinct from that of most epithelial tumors (e.g., breast, colon, and lung cancer). Hematogeneous metastasis is rare in ovarian cancer, and in most cases ovarian tumors remain confined to the peritoneal cavity. Certainly, there is evidence from studies with ovarian cancer cell lines [11, 18, 19] that MMP-2 and MMP-9 activity plays an important role in the invasion and metastasis of ovarian cancer cells. Indeed, Naylor and colleagues [20] showed MMP-2 and MMP-9 activity in almost all the ovarian cancer tissues analyzed, but survival data were not available. We undertook a study to measure MMP-2 and MMP-9 activity and protein expression in ovarian cancer and determine whether they have prognostic significance in patients with FIGO stage III ovarian cancer. MATERIALS AND METHODS Clinical Data Into the study group 84 patients (1990 –1995) were enrolled who were treated for advanced ovarian cancer (FIGO stage III) at the Department of Obstetrics and Gynecology of the Technische Universita¨t Mu¨nchen, Munich, Germany. The surgical staging and debulking consisted of standard procedures and included, where appropriate, the resection of the small and large intestine, peritonectomies, and pelvic and paraaortic lymphadenectomy [21]. Residual disease was defined by the presence or absence of macroscopically visible tumor after completion of surgery. All except 5 patients, who were in poor health, received postoperative chemotherapy with six cycles of a platinum- and/or paclitaxel-containing regimen. Eleven patients received cyclophosphamide. The 84 ovarian cancer patients (Table 1) were compared with a group of 19 patients (median age, 68; range, 44 – 87) with benign ovarian tumors (serous cystadenomas).

TABLE 1 Ovarian Cancer Patient Characteristics (n ⴝ 84) Clinicopathologic parameter

Patients (n)

Median age (range) Median observation time of patients alive, months (range) FIGO stage IIIA IIIB IIIC Histology Papillary serous Mucinous Endometrioid Clear cell Undifferentiated Grading 1/2 3/4 Ascites volume ⬍500 ml ⱖ500 ml Residual disease No Yes Chemotherapy Platinum and/or paclitaxel Other None

62 (20–83)

%

55 (26–114) 4 5 75

5 6 89

57 6 3 3 15

68 7 4 4 17

25 59

30 70

51 33

61 39

42 42

50 50

68 11 5

81 13 6

Tissue Extraction and Gelatin Zymography Ovarian tissue was obtained at surgery, stored on ice until frozen section diagnosis by the pathologist, then snap-frozen, and stored in liquid nitrogen [4]. Informed written consent was obtained by the patient and tissue procurement was approved by the ethics committee of the Klinikum rechts der Isar, Technische Universita¨t Mu¨nchen. Ovarian tissue was pulverized in the frozen state with a microdismembrator (Braun-Melsungen, Melsungen, Germany) and resuspended in a buffer containing 0.02 M Tris–HCl (pH 8.5), 0.125 M NaCl, and 1% Triton X-100 under gentle rotation at 4°C. The lysate was ultracentrifuged at 100,000g, (45 min, 4°C) and an aliquot of the supernatant assayed for protein content [22]. Gelatinolytic zymography was performed as described [23]. Briefly, tumor extracts (50 ␮g) were denatured in the absence of reducing agent and electrophoresed in 7.5% SDS–PAGE containing 0.1% (w/v) gelatin. Then the gel was incubated for 2 h at room temperature in the presence of 2.5% Triton X-100 followed by incubation at 37°C overnight in a buffer consisting of 10 mM CaCl 2, 0.15 M NaCl, and 50 mM Tris–HCl (pH 7.5). The gel was stained for protein with 0.25% Coomassie blue and dried. Proteolysis was detected as a white zone in a dark field. As an internal reference, in each gel 6.25/12.5/25/50 ng of recombinant MMP-2 and MMP-9 was included. The gelatinolytic activity of MMP-2 and MMP-9 was determined using

293

MMP-9 IN OVARIAN CANCER

densitometric scanning with an Epson GT-9500 scanner and Scan Pack 3.0 software (Biometra, Go¨ttingen, Germany), taking into account the intensity and the width of the band. Each gel was scanned twice, and a standard curve set up for every zymogram in relation to the gelatinolytic activity of recombinant MMP-2 and MMP-9. MMP-2 and MMP-9 gelatinolytic activity present in the ovarian tissues was compared with this curve. MMP-2 and MMP-9 activity was expressed in arbitrary units (U) per microgram of ovarian cancer tissue. Scanning and analysis were done without knowledge of the clinical data. As a control for the activation of pro-MMP-9 to active MMP-9, ovarian tissue extracts were treated with aminophenylmercuric acetate (APMA). Immunohistochemistry For the immunohistochemical detection of the MMP-9 protein, an avidin– biotin peroxidase technique was used on paraffin-embedded, formalin-fixed sections [24] using the DAKO ChemMate Detection Kit DAB (DAKO, Hamburg, Germany). Tissue specimens were dewaxed and digested with 0.1% proteinase K (Sigma, Deisenhofen, Germany), and endogenous peroxidase activity was blocked by incubation in 3% hydrogen peroxide in methanol for 30 min at room temperature. Two antibodies to MMP-9 were used (Triple Points, monoclonal, and CA-209, polyclonal [25]). They were each incubated for 2 h at room temperature (10 ␮g/ml). In Situ Zymography A section was cut from fresh ovarian cancer FIGO stage III tissue specimens and mounted onto gelatin films that were coated with a solution containing 7% gelatin (Fuji Photo Film Co., Tokyo, Japan). The film was incubated for 12 h at 37°C and then stained with 1% amido black 10B. A serial frozen section from the same tissue block was stained with hematoxylin and eosin to distinguish tumor cell nests from the surrounding extracellular matrix. Statistical Analyses Statistical analyses were performed using the SPSS computer program package (SPSS 8.0, Inc., Chicago, IL) or S-Plus (MathSoft 98, Seattle, WA). The level of the biochemical factors as well as other continuous factors is given by the median. The comparison of MMPs in FIGO stage III carcinoma with serous cystadenomas was done applying the Mann– Whitney U test. For all further analyses, MMP-2 and MMP-9 activity were coded as binary variables, using optimized cutoff levels based on log-rank statistic, distinguishing between lowand high-risk patients. A ␹ 2 test of independence and Pearsons R were calculated to test the relationships between traditional clinicopathologic parameters and the MMPs. Kaplan–Meier estimates were used for graphic presentation of survival rates in different groups. The prognostic impact of the analyzed factors on survival time after primary therapy was assessed by

the Cox proportional hazards model, which provides estimates of the relative risk (RR) together with the 95% confidence intervals. To compare the prognostic power of various factors in a multivariate Cox regression, a covariate selection was carried out by a stepwise forward procedure for choosing factors with the highest additional information on prognosis. Interactions between the factors with respect to prognosis were also studied in a stepwise procedure but showed no additional significant influence in multivariate analysis. Probability values ⱕ0.05 were regarded as statistically significant. RESULTS The distribution of traditional clinicopathologic variables, such as patient age, histologic type, histologic grade, volume of ascites, and residual tumor mass after completion of surgery, is shown in Table 1. All 84 patients had advanced ovarian cancer FIGO stage III, the majority (89%) belonging to FIGO stage IIIc. Median survival time for all 84 patients was 47 months. Looking at correlations between the classic clinicopathologic variables, only the presence of ascites (volume ⬎ 500 ml) was highly correlated with residual tumor mass. In univariate survival analysis, grading, ascites volume, age, and residual disease were correlated with survival (Table 2). Macroscopically tumor-free patients (n ⫽ 42) had a significant survival advantage (median survival time 74 months) over patients with postoperative residual tumor mass (n ⫽ 42), their median survival time being 16 months. MMP-2 and MMP-9 Activity Since the proteolytic activity of metalloproteinases is crucial for their biologic effect in tumors we used gelatinolytic zymography to determine the activity of MMP-2 and MMP-9 in benign and malignant ovarian tumors. This technique distinguishes between active and latent forms of MMP-2 and MMP-9 (MMP-2, M r 72,000/68,000, and MMP-9, M r 92,000/ 85,000). In the presence of SDS the enzymes are denatured, exposing their active site, which permits both the latent and active forms of the gelatinases to exhibit gelatinolytic activity after partial renaturation. Equal amounts of protein (50 ␮g) were applied per lane and expression of the MMPs was assessed as shown for 10 different ovarian tumors in Fig. 1. To investigate whether there is a difference in MMP-2 and MMP-9 activity between benign and malignant ovarian tissue, the levels of MMP-2 and MMP-9 gelatinolytic activity were compared. There was no statistically significant difference between pro-MMP-2 expression in benign and malignant tissue (Table 3). Interestingly, the activated form of MMP-2 was detected only in malignant ovarian tissue but was absent in benign serous cystadenomas. In contrast, the median level of pro-MMP-9 was clearly higher in ovarian cancer tissues (P ⬍ 0.001) than in the benign ovarian tissues analyzed (Table 3). Looking at ovarian cancer tissues, pro-MMP-9 was detected in

294

LENGYEL ET AL.

TABLE 2 Univariate and Multivariate Overall Survival Analysis of FIGO III Ovarian Cancer Patients (n ⴝ 84) a Univariate

Multivariate

Prognostic variable

RR

95% CI

P value

RR

95% CI

P value

Pro-MMP-9 activity: ⱕ6 U/␮g vs ⬎6 U/␮g Pro-MMP-2 activity MMP-2 activity Residual tumor mass present vs absent Age: ⱕ60 vs ⬎60 Ascites volume: ⱕ500 ml vs ⬎500 ml Grading: G1 ⫹ G2 vs G3 ⫹ G4

1.99 — — 3.41 1.83 1.80 2.28

1.12–3.52 — — 1.83–6.37 1.03–3.24 1.02–3.17 1.13–4.59

0.019 n.s. n.s. ⬍0.001 0.038 0.042 0.021

1.94 — — 3.40 — — —

1.10–3.46 — — 1.81–6.39 — — —

0.023 — — ⬍0.001 n.s. n.s. n.s.

a

CI, confidence interval; RR, relative risk; n.s., not significant.

57 of 84 (68%), the active form of MMP-2 in 44 of 84 (52%), and pro-MMP-2 in 65 of 84 (77%) ovarian cancer tissue extracts. We did not find any activated MMP-9 (M r 85,000) in benign and malignant ovarian tissues, but could activate proMMP-9 to active MMP-9 by pretreating extracts with APMA, a nonphysiologic pro-MMP-9 activator (data not shown). There was a weak correlation (P ⫽ 0.02) between proMMP-9 activity and age, but not with any of the other traditional clinicopathologic variables investigated. Survival Analysis For the survival analysis, MMPs were coded as binary variables by employing log-rank statistics to determine optimal cutoffs thereby discriminating between low- and high-risk patients (Table 2). Follow-up data were available on all ovarian cancer patients analyzed. The median follow-up time for surviving patients was 55 months. In univariate analysis, elevated pro-MMP-9 expression (Table 2) was significantly associated with overall survival. When calculating Kaplan–Meier survival curves (Fig. 2A) looking at all tumors (n ⫽ 84), patients with high pro-MMP-9 expression have a median survival time of 27 months, compared with patients with low pro-MMP-9 expression who have a median overall survival time of 67 months (P ⫽ 0.019). When analyzing the subgroups of patients with (n ⫽ 42) and without (n ⫽ 42) residual tumor mass at the time of surgery, pro-MMP-9 was of prognostic value only in

the subgroup of patients with no residual tumor mass (Fig. 2B), while for patients with residual tumor mass pro-MMP-9 did not predict survival. Therefore, while pro-MMP-9 activity is of prognostic value in all patients with ovarian cancer, this is due mainly to the strong prognostic power in patients with no residual disease at the end of surgery. MMP-2 and pro-MMP-2 showed no prognostic significance in FIGO stage III ovarian cancer (data not shown). TIMP-1 and TIMP-2 are the physiologic inhibitors of MMP-9 and MMP-2, respectively. We therefore evaluated in a subset of 50 patients their protein expression by Western blotting and did not find any correlation to surival. Complete data for a multivariate Cox regression overall survival analysis were available for all 84 patients. In the Cox analysis age at diagnosis, histologic grade, presence or absence of macroscopic residual tumor, ascites volume, and proMMP-2 and pro-MMP-9 activity were entered into the stepwise selection procedure. Residual tumor mass (P ⬍ 0.001) and pro-MMP-9 activity (P ⫽ 0.023) proved to be statistically independent predictors of overall survival (Table 2). Immunohistochemistry and in Situ Zymography While zymography is a good method for measuring the gelatinolytic activity of MMP-2 and MMP-9, it is not possible

TABLE 3 Comparison of Median MMP-2 and MMP-9 Activity between Benign and Malignant Ovarian Tissues (Mann–Whitney Test)

FIG. 1. Zymographic analysis of gelatinolytic activity in 10 ovarian cancer tissue extracts. Activity is indicated by the white zones where the gelatin has been hydrolyzed.

Parameter

Benign ovarian tissue, median (n ⫽ 19)

Ovarian cancer tissue, median (n ⫽ 84)

P value

Pro-MMP-2 activity (U/␮g) MMP-2 activity (U/␮g) Pro-MMP-9 activity (U/␮g) MMP-9 activity (U/␮g)

3.70 0.00 0.00 0.00

5.92 1.6 4.82 0.00

n.s. a ⬍0.001 ⬍0.001 n.s.

a

n.s., not significant.

MMP-9 IN OVARIAN CANCER

295

FIG. 2. Kaplan–Meier curves of overall survival rates for low and high levels of pro-MMP-9 activity. (A) Pro-MMP-9 activity in all patients (n ⫽ 84). (B) Pro-MMP-9 activity in the subgroup of patients with no residual disease (n ⫽ 42). Numbers in parentheses indicate number of deaths per total number of patients in each group.

to relate the data to cellular localization of MMP expression in the tumor tissue. Since pro-MMP-9 showed prognostic significance in uni- and multivariate analysis, an immunohistochemical study on tumor tissue of 10 ovarian cancer patients was performed to identify the cellular localization of MMP-9 with an antibody that recognizes pro-MMP-9 and active MMP-9. Distribution of MMP-9 was mostly homogenous and mainly in

the epithelial tumor compartment (Fig. 3A). In contrast to previous studies on other cancer types, there was no stronger protein staining at the periphery of the tumor [26]. At higher resolution MMP-9 is located in the cytoplasm of the tumor cells and in adjacent stromal cells (Fig. 3B). Immunohistochemical staining was repeated with a polyclonal antibody [25] against MMP-9, yielding identical staining results (data not

296

LENGYEL ET AL.

FIG. 3. Immunohistochemical localization of MMP-9 in formalin-fixed, paraffin-embedded ovarian cancer tissue. The tissue section was incubated with a monoclonal antibody to MMP-9 and immune complexes were visualized by the avidin– biotin technique. A positive reaction is indicated by brown staining. (A) Ovarian cancer, FIGO stage IIIc, 1:40 magnification. (B) Metastasis of the same ovarian cancer tissue at 1:400 magnification.

shown). To locate areas of gelatinolytic activity in the ovarian cancer tissue specimen in situ zymography was performed using a frozen section mounted on gelatin films and an adjacent section stained with hematoxylin and eosin (Figs. 4A, 4B). Gelatinolytic activity in ovarian cancer is located predominantly in the epithelial tumor compartment. However, weak activity was reproducibly detectable in single cells between the epithelial tumor cell nests. In conclusion, gelatinolytic activity and MMP-9 protein expression in ovarian cancer was distributed mainly in the epithelial tumor compartment. DISCUSSION A major aim of this study was to assess the prognostic impact of MMP-2 and MMP-9 expression on survival while accounting for generally accepted clinical and histopathologic factors. Univariate analysis confirmed the significant prognostic value of the traditional clinicopathologic variables (i.e., residual tumor mass, histologic grade, ascites volume, age) as predictors of overall survival [21, 27]. In multivariate analysis residual disease was the only statistically independent prognostic factor, thus having a stronger clinical value than the biological parameters examined. A prognostic significance of MMP-2 in human ovarian can-

cer has been sparsely reported [28, 29] and to the best of our knowledge there have been no reports on the prognostic significance of MMP-9 in ovarian cancer (reviewed in [8]). In vitro, ovarian cancer cell lines that secrete pro-MMP-9 comparable to tumors from patients with high risk for recurrence are highly invasive. Our results demonstrate that in univariate and multivariate analysis low pro-MMP-9 activity correlates with a favorable prognosis in ovarian cancer, while high proMMP-9 indicates early death. Interestingly, pro-MMP-9 is of prognostic value in the subgroup of patients without residual tumor mass but does not predict prognosis in patients with residual disease. Further studies in another set of patients are needed, however, to validate results, as patients with no residual tumor mass and high pro-MMP-9 expression should be followed up more closely and eventually receive more aggressive individualized adjuvant therapy. The absence of the activated form of MMP-9 in benign and malignant ovarian tumors is not unique; this phenomenon has been observed in cancer of the stomach, breast, and colon as well [13, 16, 30, 31]. This feature strongly suggests that MMP-9 in ovarian cancer is not causing an invasive phenotype but is associated with it and that, at least for ovarian cancer, the prognostic value of MMP-9 is independent of its activation state. Possibly, the high pro-

FIG. 4. In situ zymography of an ovarian cancer case, FIGO stage IIIc. Serial frozen sections were subjected to staining with (A) hematoxylin/eosin or (B) in situ zymography. Note the strong gelatinolytic activity in the carcinoma cell nests.

297

MMP-9 IN OVARIAN CANCER

MMP-9 expression within aggressive tumors reflects the presence of inflammatory cells, which have been shown to promote tumor progression [32]. In contrast to MMP-9, the activated form of MMP-2 was present in about half of the ovarian cancer tissue specimens, but this finding is of no prognostic significance. Moreover, none of the benign tumors showed any active MMP-2. These findings are in agreement with studies in breast cancer [16, 33] showing that activation of MMP-2 in cancer is a more common event than activation of MMP-9. Expression of MMP-2 in ovarian cancer was studied by Garzetti and colleagues [28], who analyzed 21 ovarian cancers for MMP-2 expression by immunohistochemistry and found a significant difference in disease-free survival between the two groups with high and low MMP-2 expression. However, when analyzing more patients (n ⫽ 43), the same group reported later [29] that MMP-2 is not of prognostic significance. We also found no prognostic value for MMP-2 or pro-MMP-2 activity or protein expression (data not shown) in ovarian cancer. Since we did not find a prognostic role for active MMP-2 or MMP-9 it might well be that other proteases like serine proteases [4] are more important for tumor invasion and metastasis in ovarian cancer, which has a unique tumor biology within the group of epithelial cancers [1]. By immunohistochemistry, we show that in ovarian cancer FIGO stage III, MMP-9 protein expression is confined mainly to the epithelial tumor cell nests. Some MMP-9 immunostaining is also observed in cells of the peritumorous area, but by this approach it was not possible to define the cell types involved. MMP-9 was detected in ovarian [20] and lung cancer [34] by in situ hybridization equally both in epithelial tumor cells and in host stromal cells surrounding the tumor. One possible explanation for the difference between detection of MMP-9 mRNA inside and outside the tumor and the detection of MMP-9 protein predominantly in the tumor is the fact that tumor cells recruit stromal cells to produce MMP-9, which is released and then binds to the surface of the tumor cell, suggesting intensive cross-talk between the stroma and the tumor [32]. Indeed, previous studies have shown a specific binding of MMP-9 to the surface of tumor cells [35], which has been shown to be mediated in part by the ␣2(IV) chain of collagen IV [36] and CD44 [37]. Immunohistochemical expression of MMP-9 mirrors the gelatinolytic activity as shown by in situ zymography. However, in the zymograms only the latent form of MMP-9 was observed. Therefore it cannot be ruled out that the gelatinolytic activity exerted by ovarian cancers in the in situ zymogram is caused by MMPs other than MMP-9 as gelatin is also a substrate for MMP-2, MMP-13, MT1-MMP, and MT2-MMP. Still, the fact remains that in ovarian cancer FIGO stage III, the major gelatinolytic activity stems from the tumor cell area. The evaluation of the prognostic relevance of MMP protein expression and/or activity in human ovarian cancer is extremely relevant given the fact that MMP inhibitors are already

in phase II/III clinical trials. However, the preliminary clinical results are disappointing [8], which was rather unexpected in view of the good antitumor effects of MMP inhibitors in animal studies [38]. Taking into account that we saw a prognostic significance only for the latent form of MMP-9 it is questionable if synthetic inhibitors against MMP-2 and/or MMP-9 will inhibit tumor growth or dissemination in patients with FIGO stage III ovarian cancer. ACKNOWLEDGMENTS We are grateful to Dr. K. Iwata and Dr. R. Nemori (Fuji Chemicals Industries, Japan) for providing us the gelatin films. We thank Dr. H. Allgayer and R. Manson, M.A., for their critical appraisal of the manuscript and Prof. Dr. K. Ulm for helpful discussions. The work would not have been possible without the contribution of Juliane Scha¨fer in the statistical analysis. This work was supported by grants from the Deutsche Krebshilfe/Dr. Mildred Scheel Stiftung (10-1197 and 10-1637 to E.L., B.S., and W.K.), the Medical Faculty of the Technische Universita¨t Mu¨nchen (KKF 8756156 to E.L. and M.S.), and the National Cancer Institute (CA-61986 to R.F.).

REFERENCES 1. Auersperg N, Edelson MI, Mok SC, Johnson SW, Hamilton TC. The biology of ovarian cancer. Semin Oncol 1998;25:281–304. 2. Fathalla M. Factors in the causation and incidence of ovarian cancer. Obstet Gynecol Surv 1972;27:751– 68. 3. Stack MS, Ellerbroek SM, Fishman DA. The role of proteolytic enzymes in the pathology of epithelial ovarian carcinoma [review]. Int J Oncol 1999;12:569 –76. 4. Kuhn W, Schmalfeldt B, Reuning U, Pache L, Berger U, Ulm K, Harbeck N, Spa¨the K, Dettmar P, Ho¨fler H, Ja¨nicke F, Schmitt M, Graeff H. Prognostic significance of urokinase (uPA) and its inhibitor PAI-1 for survival in advanced ovarian carcinoma stage FIGO IIIc. Br J Cancer 1999;79:1746 –51. 5. Woodhouse EC, Chuaqui RF, Liotta L. General mechanisms of metastasis. Cancer 1997;80:1529 –37. 6. Nagase H, Woesner JF. Matrix metalloproteinases. J Biol Chem 1999; 274:21491– 4. 7. Shapiro S. Matrix metalloproteinase degradation of extracellular matrix: biological consequences. Curr Opin Cell Biol 1998;10:602– 8. 8. Nelson AR, Fingleton B, Rothenberg ML, Matrisian LM. Matrix metalloproteinases: biologic activity and clinical implications. J Clin Oncol 2000;18:1135– 49. 9. Bernhard EJ, Gruber SB, Muschel RJ. Direct evidence linking expression of matrix metalloproteinase 9 (92-kDa gelatinase/collagenase) to the metastatic phenotype in transformed rat embryo cells. Proc Natl Acad Sci USA 1994;91:4293–7. 10. Sato H, Kida Y, Mai M, Endo Y, Sasaki T, Tanaka J, Seiki M. Expression of genes encoding type IV collagen-degrading metalloproteinases and tissue inhibitors of metalloproteinases in various human tumor cells. Oncogene 1999;7:77– 85. 11. Moser TL, Young TN, Rodriguez GC, Pizzo SV, Bast RC, Stack MS. Secretion of extracellular matrix-degrading proteinases is increased in epithelial ovarian carcinoma. Int J Cancer 1994;56:552–9. 12. Liabakk NB, Talbot I, Smith RA, Wilkinson K, Balkwill F. Matrix metalloprotease 2 (MMP-2) and matrix metalloprotease 9 (MMP-9) type IV collagenases in colorectal cancer. Cancer Res 1996;56:190 – 6. 13. Duffy MJ, McCarthy K. Matrix metalloproteinases in cancer: prognostic markers and targets for therapy [review]. Int J Oncol 1998;12:1343– 8.

298

LENGYEL ET AL.

14. Duffy MJ. Proteases as prognostic markers in cancer. Clin Cancer Res 1996;2:613– 8. 15. Zeng ZS, Huang Y, Cohen AM, Guillem JG. Prediction of colorectal cancer relapse and survival via tissue RNA levels of matrix metalloproteinase-9. J Clin Oncol 1996;14:3133– 40. 16. Remacle A, Noe¨l A, Duggan C, McDermott E, O⬘Higgins N, Foidart J, Duffy MJ. Assay of matrix metalloproteinases types 1, 2, 3 and 9 in breast cancer. Br J Cancer 1998;77:926 –31. 17. Allgayer H, Babic R, Beyer BC, Gru¨tzner KU, Tarabachi A, Schildberg F, Heiss MM. Prognostic relevance of MMP-2 (72-kD collagenase IV) in gastric cancer. Oncology 1998;55:152– 60. 18. Young TN, Rodriguez GC, Rinehart AR, Bast RC, Pizzo SV, Stack MS. Characterization of gelatinases linked to extracellular matrix invasion in ovarian adenocarcinoma: purification of matrix metalloproteinase 2. Gynecol Oncol 1996;62:89 –99. 19. Shibata K, Kikkawa F, Nawa A, Suganuma N, Hamaguchi M. Fibronectin secretion from human peritoneal tissue induces Mr 92,000 type IV collagenase expression and invasion in ovarian cancer cell lines. Cancer Res 1997;57:5416 –20. 20. Naylor MS, Stamp GW, Davies BD, Balkwill FR. Expression and activity of MMPs and their regulators in ovarian cancer. Int J Cancer 1994;58: 50 – 6. 21. Kuhn W, Florack G, Roder J, Schmalfeldt B, Pache L, Rust M, Ulm K, Spa¨the K, Ja¨nicke F, Siewert JR, Graeff H. The influence of upper abdominal surgery on perioperative morbidity and mortality in patients with advanced ovarian cancer FIGO III and FIGO IV. Int J Gynecol Cancer 1998;8:56 – 63. 22. Lengyel E, Stepp E, Gum R, Boyd D. Involvement of a mitogen-activated protein kinase signaling pathway in the regulation of urokinase promoter activity by c-Ha-ras. J Biol Chem 1995;270:23007–12.

27.

28.

29.

30.

31.

32. 33.

34.

35.

23. Lengyel E, Gum R, Juarez J, Clayman G, Seiki M, Sato H, Boyd D. Induction of Mr 92,000 type IV collagenase expression in a squamous cell carcinoma cell line by fibroblasts. Cancer Res 1995;55:963–7. 24. Schnelzer A, Prechtel D, Knaus U, Dehne K, Gerhard M, Graeff H, Harbeck N, Schmitt M, Lengyel E. Rac1 in human breast cancer: overexpression, mutation analysis, and characterization of a new isoform, Rac1b. Oncogene 2000;19:3013–20. 25. Grignon DJ, Sakr WA, Toth M, Ravery M, Angulo J, Shamsa F, Pontes JE, Crissman JC, Fridman R. High levels of tissue inhibitor of metalloproteinase-2 (TIMP-2) expression are associated with poor outcome in invasive bladder cancer. Cancer Res 1996;56:1654 –9. 26. Ashida K, Nakatsukasa H, Higashi T, Ohguchi S, Hino N, Nouso K, Urabe Y, Yoshida K, Kinugasa N, Tsuji T. Cellular distribution of 92-kd type IV

36.

37.

38.

collagenase/gelatinase B in human hepatocellular carcinoma. Am J Pathol 1996;149:1803–11. Anttila M, Kosma VM, Ji H, Wei-Ling X, Puolakka J, Juhola M, Saarikoski S, Syrja¨nen K. Clinical significance of ␣-catenin, collagen IV, and Ki-67 expression in epithelial ovarian cancer. J Clin Oncol 1998;16:2591– 600. Garzetti G, Ciavattini A, Lucarini G, Goteri G, Nictolis MD, Garbisa S, Masiero L, Romanini C, Graziella B. Tissue and serum metalloproteinase (MMP-2) expression in advanced ovarian serous cystadenocarcinomas: clinical and prognostic implications. Anticancer Res 1995;15:2799 – 804. Nictolis MD, Garbisa S, Lucarini G, Goteri G, Masiero L, Ciavattini A, Garzetti G, Stetler-Stevenson WG, Fabris G, Biagini G, Prat J. 72Kilodalton type IV collagenase, type IV collagen, and Ki67 antigen in serous tumors of the ovary: a clinicopathologic, immunohistochemical, and serological study. Int J Gynecol Pathol 1999;15:102–9. Sier CFM, Kubben FJGM, Ganesh S, Heerding MM, Griffioen G, Hanemaaijer R, van Krieken JHJM, Lamers C, Verspaget HW. Tissue levels of matrix metalloproteinases MMP-2 and MMP-9 are related to the overall survival of patients with gastric carcinoma. Br J Cancer 1996;74:413–7. Kumar A, Collins H, Scholefield J, Watson SA. Increased type-IV collagenase (MMP-2 and MMP-9) activity following preoperative radiotherapy in rectal cancer. Br J Cancer 2000;82:960 –5. Bergers G, Coussens LM. Extrinsic regulators of epithelial tumor progression: metalloproteinases. Curr Opin Genet Dev 2000;10:120 –7. Davies B, Miles DW, Happerfield LC, Naylor MS, Bobrow LG, Rubens RD, Balkwill FR. Activity of type IV collagenases in benign and malignant breast disease. Br J Cancer 1993;67:1126 –31. Canete-Soler R, Litzky L, Lubensky I, Muschel RJ. Localization of the 92 kd gelatinase mRNA in squamous cell and adenocarcinoma of the lung using in situ hybridization. Am J Pathol 1994;144:518 –27. Mazzieri R, Masiero L, Zanetta L, Monea S, Onisto M, Garbisa S, Mignatti P. Control of type IV collagenase activity by components of the urokinase–plasmin system: a regulatory mechanism with cell-bound reactants. EMBO J 1997;16:2319 –32. Olson M, Toth M, Gervasi DC, Sado Y, Ninomiya Y, Fridman R. High affinity binding of latent matrix metalloproteinase-9 to the ␣2(IV) chain of collagen IV. J Biol Chem 1999;273:10681. Yu Q, Stamenkovic I. Localization of matrix metalloproteinase 9 to the cell surface provides a mechanism for CD44-mediated tumor invasion. Genes Dev 1999;13:35– 48. Davies B, Brown PD, East N, Crimmin MJ, Balkwill F. A synthetic matrix metalloproteinase inhibitor decreases tumor burden and prolongs survival of mice bearing human ovarian carcinoma xenografts. Cancer Res 1993; 53:2087–91.