Int. J. Cancer (Pred. Oncol.): 69,9-16 (1996) 0 1996 Wiley-Liss, Inc. I Publicationof the International Union Against Cancer Publicationde I’Union Internationale Contre le Cancer ENHANCED PRODUCTION OF MATRIX METALLOPROTEINASES AND ACTIVATION OF MATRIX METALLOPROTEINASE 2 (GELATINASE A) IN HUMAN GASTRIC CARCINOMAS Hidehiro NOMLJRA~,~, Noboru FUJIMOTO~, Motoharu SEIK13, Masayoshi M A Iand ~ Yasunori OKADA’,~ Departments of ’MolecularImmunology and Pathology, 2Surgery and 3MolecularVirology and Oncology, Cancer Research Institute, Kanazawa University, Kanazawa, Ishikawa 920; and 4Biophamaceutical Department, Fuji Chemical Industries Ltd., Takaoka, Toyama 933, Japan. We examined the production and tissue localization of matrix metalloproteinases(MMPs) and tissue inhibitors of metalloproteinases (TIMPs) in gastric carcinoma tissues. MMP- I (tissue collagenase), MMP-9 (gelatinase B) and TIMP-2 were immunolocalized in carcinoma cells and MMP-2 (gelatinase A) on tumor cell membranes,whereas no or little immunostainingfor MMP-3 (stromelysin- I) and TIMP- I was seen in carcinoma cells. Stroma1 cells in carcinoma tissue were also positively stained for these MMPs and TIMPs. MMP-2 immunostainingwas observed exclusively on advanced gastric carcinoma cells and correlated with vascular invasion by tumor cells. Sandwich enzyme immunoassays revealed enhanced production of MMP- I, MMP-2, MMP-3. MMP-9 and TIMP- I by carcinoma tissues. Gelatinolytic activities were significantly higher in carcinoma samples than in normal controls. Using gelatin zymography, active forms of MMP-2 and MMP-9 were more frequently detected in carcinoma tissue, and the activation rate of the zymogen of MMP-2 (proMMP-2). but not that of proMMP-9, correlated well with degree of local invasion and lymphatic permeation. Our data indicate an enhanced production of 4 MMPs in gastric carcinoma tissue and sunnest that activation of pro-MMP-2 may be a key step for spreadEg of gastric carcinoma cells. 0 1996 Wiley-Liss,Inc. Matrix metalloproteinases (MMPs) are a gene family of Zn2+ metalloproteinases composed of at least 11 different gene products and involved in degradation of the extracellular matrix macromolecules associated with tissue destruction in various pathological conditions (Birkedal-Hansen et al., 1993; Stetler-Stevenson et al., 1993; Sato et al., 1994). Previous experimental and clinicopathological studies have shown good correlations between expression of MMPs and the invasive phenotype of tumor cells (Stetler-Stevenson et al., 1993). There are several studies on MMP expression and production by gastric cancer (McDonnell et al., 1991; David et al., 1994; Grigioni et al., 1994; Otani et al., 1994), which is the most common cancer in Japan and some other countries. Gastric carcinoma tissues have been reported to exhibit mRNA expression of MMP-1 (tissue collagenase) and MMP-7 (matrilysin) but not MMP-3 (stromelysin-1) or MMP-10 (stromelysin-2) (McDonnell et al., 1991; Otani et al., 1994). MMP-2 (gelatinase A) has also been immunolocalized within carcinoma cells in stomach cancers (David et al., 1994; Grigioni et al., 1994). However, there are few studies examining the tissue localization of several MMPs in the same specimens of gastric carcinoma. The activity of MMPs is regulated by tissue inhibitors of metalloproteinases (TIMPs), which comprise 3 different molecules, i.e., TIMP-l, TIMP-2 and TIMP-3 (Birkedal-Hansen et al., 1993; Stetler-Stevenson et al., 1993). It is also known that over-expression of TIMPs suppresses invasion and metastasis by tumor cell lines (Stetler-Stevenson et al., 1993). In stomach cancers, the percentage of TIMP-2-immunoreactive carcinoma cells is reported to decrease in advanced carcinomas, showing an inverse correlation of TIMP-2 expression with invasive properties of carcinoma cells (Grigioni et aZ., 1994). This suggests that production of TIMPs at the local tissue level also modulates invasion and metastasis of cancer cells. Although Lokeshwar et al. (1993) have shown that prostatic cancer tissue produces less TIMP activity than controls, there are no other studies describing the amounts of MMPs and TIMPs produced by cancer tissues. Since MMPs are produced as zymogen forms (pro-MMPs) and activated extracellularly, activation is another key step for each MMP to function in vivo. The occurrence of active MMP-2 species correlates well with an invasive and metastatic phenotype in lung (Brown et al., 1993) and breast (Davies et al., 1993) cancers. We have identified an activator metalloproteinase for pro-MMP-2, i.e., membrane type-MMP (MT-MMP; Sato et al., 1994), and demonstrated MT-MMP-associated activation of pro-MMP-2 in human gastric carcinomas (Nomura et al., 1995). However, no studies are available that correlate activation of proMMP-2 with progression of human gastric carcinomas. In the present work we have examined immunolocalization and production of 4 different MMPs (MMP-1, MMP-2, MMP-3 and MMP-9) and 2 TIMPs (TIMP-1 and TIMP-2) in human gastric carcinomas. The results suggest that overexpression of these MMPs and activation of proMMP-2 are important in the malignant behavior of stomach cancers. MATERIAL AND METHODS Histology Fresh tissue samples were obtained from 46 patients with gastric carcinoma who underwent surgery in the Cancer Research Hospital of Kanazawa University. Blocks of tissue were excised from the margins of carcinoma and normal gastric wall remote from the tumor. Each sample was fixed with periodate-lysine-paraformaldehydefixative for 18-24 hr at 4°C and embedded in paraffin wax. Surgical specimens were stained with hematoxylin and eosin and classified according to “The General Rules for the Gastric Cancer Study” (Japanese Research Society for Gastric Cancer, 1981). Imrnunohistochemistry Monoclonal antibodies (MAbs) against MMP-1 (41-1E5), MMP-2 (75-7F7), MMP-3 (55-2A4), MMP-9 (56-2A4), TIMP-1 (50-3D2) and TIMP-2 (67-4H11) have been characterized and previously used for immunolocalization studies (Zhang et al., 1993; Fujimoto et al., 1993a, b, 1994; Obata et al., 1992). All antibodies for MMPs recognize both zymogen and active forms. For immunohistochemistry, deparaffinized sections were reacted with primary antibodies or non-immune mouse IgG for 15-18 hr at 4°C after blocking endogenous peroxidase by treatment with 0.3% H202 and non-specific binding with 10% normal horse serum. Some sections were treated with 0.04% trypsin in 50 mM Tris-HC1 buffer, pH 7.6, for 10 min at 37°C and then reacted with the antibody against TIMP-1. Concentrations of primary antibodies used for immunostaining were 1 5To whom correspondence and reprint requests should be sent, at Department of Molecular Immunology and Pathology, Cancer Research Institute, Kanazawa University. 13-1 Takara-machi, Kanazawa, Ishikawa 920, Japan. Fax: (81) 0762-34-4508. Received: August 9, 1995 and in revised form October 26, 1995 10 NOMURA ETAL pg/ml for MMP-1,4 pg/ml for MMP-2,8.3 pg/ml for MMP-3, 1 kg/ml for MMP-9, 10 pg/ml for TIMP-1, 1 pg/ml for TIMP-2 and 4 p,g/ml for non-immune mouse IgG. For absorption studies, the antibody against MMP-2 was first subjected to pro-MMP-2-Sepharose or BSA-Sepharose column and the effluent was used for immunostaining. Sections reacted with antibodies were incubated with biotinylated horse IgG to mouse IgG (Vector, Burlingame, CA, diluted 1:200) for 30 min and then with an avidin-biotin-peroxidase complex (Dako, Glostrup, Denmark) for 30 min at room temperature. Color was developed with 0.03% 3,3’-diaminobenzidine tetrahydrochloride in 50 mM Tris-HCI buffer, pH 7.6, containing 0.006% Hz02. Counter-staining was performed with hematoxylin. The ratio (%) of immunoreactive cells to total carcinoma cells was measured by counting cells in 3 different fields at a magnification of ~ 2 0 0 . Tissue cultures Tissue samples of carcinoma and normal gastric mucosa were excised from samples obtained for immunostaining. After rinsing in 0.01 M phosphate buffer, pH 7.5, containing 0.15 M NaCI, they were cut into small blocks (approx. 2 x 2 x 2 mm), and 18-20 blocks were cultured for 3 hr in 2 ml of serum-free RPMI (GIBCO, Grand Island, NY) containing 0.2% lactalbumin hydrolysate (GIBCO). Culture media were stored at -20°C until used for assays and Zymography, and tissue blocks were weighed after lyophilization. stained with 0.1% Coomassie brilliant blue R250. The intensity of the gelatinolytic bands corresponding to MMP-2 and MMP-9 was measured using computer-assisted image analysis according to the method of Davies et al. (1993). The volume of the media loaded was adjusted acclording to the weight of the tissue used for the culture. Authentic pro-MMP-2 from rheumatoid synovial fibroblasts and pro-MMP-9 from HT-1080 fibrosarcoma cells were also subjected to gelatin zymography for comparison. For inhibition studies, representative culture media were subjected to gelatin qmography and sliced gels were incubated at 37°C in SO mM Tris-HC1, pH 7.4, 0.15 M NaCI, 10 mM CaC12 and 0.02% NaN3 in the presence and absence of 20 mM EDTA, 1 mM o-phenanthroline, 2 mM phenylmethane sulfonyl fluoride, 2 mM diisopropylfluorophosphate, 5 mMN-ethylmaleimide or 0.5 mM pepstatin A. Statistical analysis Statistical analysis was performed using 2-tailed MannWhitney U and x2 tests; p values below 0.05 were considered significant. RESULTS Histology and immunohistochemistly Forty-six gastric carcinomas were examined for histology and immunohistochemistry. They were classified into early and advanced gastric carcinoma groups according to wall penetration: early carcinoma is defined as carcinoma confined to Sandwich enzyme immunoassays mucosa and submucosa and advanced carcinoma as more Concentrations of MMP-1, MMP-2, MMP-3, MMP-9, invasive carcinoma (Japanese Research Society for Gastric TIMP-1 and TIMP-2 in culture media were measured in the Cancer, 1981). Other clinicopathological parameters, includcorresponding sandwich enzyme immunoassay (EIA) systems ing lymphatic permeation, vascular invasion, lymph node for these MMPs and TIMPs as described previously (Zhang et metastasis and peritoneal dissemination, were also examined. al., 1993; Fujimoto et al., 1993a, b, 1994; Obata et al., 1992; Normal control tissues showed various degrees of inflammaKodama et al., 1990). The EIA systems for MMP-1 and tory cell infiltration and foci of intestinal metaplasia of the MMP-3 measure both precursor and active forms of the epithelium. MMPs, but those for MMP-2 and MMP-9 detect only their Immunohistochemically, carcinoma cells showed intracytolatent forms (Zhang et al., 1993; Fujimoto et al., 1993a, 1994; plasmic staining for MMP-1 (Fig. la), MMP-9 (Fig. Id) and Obata et al., 1992). The EIA for TIMP-1 determines the whole (Fig. 2b),and percentages of positive cases were 83%, amount of TIMP-1, including free TIMP-1 and the complexed TIMP-2 35% and 43%, respectively (Table I). However, MMP-2 was forms with MMPs and pro-MMP-9 (Kodama et al., 1990). immunolocalized on the cell membranes of carcinoma cells in However, the EIA for TIMP-2 detects free TIMP-2 and 14 cases (30%) (Fig. lb), though 41 of the 14 cases exhibited TIMP-2 complexed with active MMPs but not the complex intracytoplasmic staining as well. Ratios of immunoreactive with pro-MMP-2 (Fujimoto et al., 19936).Values are expressed cells to total carcinoma cells in the positively stained cases for as nmol/g dry weight tissue. The following m.w. were used for MMP-1, MMP-2, MMP-9 and TIMP-2 were 45 c 26%, 18 ? calculation: 51,929 for MMP-1, 70,952 for MMP-2, 52,220 for 16%, 28 c 24% and 32 ? 22% (mean 2 SD), respectively. MMP-3,78,426 for MMP-9, 20,685 for TIMP-1 and 21,755 for These MMPs showed no special1 distribution patterns of TIMP-2 (see references in Birkedal-Hansen et al., 1993). immunoreactive carcinoma cells in tumor cell nests, but staining of TIMP-2 had a tendency to be stronger in the Enzyme assays Gelatinolytic activities in culture media were measured in an marginal carcinoma cells than in the tumor center. Immunoassay using [I4C] acetylated type I gelatin in the presence and staining of MMP-3 (Fig. lc) and TIIMP-1 (Fig. 2a) was weak or absence of 1 mM p-aminophenylmercuric acetate (APMA). negative in carcinoma cells. Stromal and inflammatory cells in Collagenolytic and carboxymethylated transferrin (Cm-Tf)- the tumors, including fibroblasts, endothelial cells, macrodegrading activities in media were also measured using [l4C] phages and neutrophils, were positive for MMP-1 (Fig. l a ) and type I collagen and [3H] Cm-Tf, respectively. Samples were MMP-9 (Fig. Id). In most cases, jibroblasts and endothelial incubated prior to assays with 1mM APMA at 37°C for 18 hr to cells in the tumors were also immuinostained for MMP-2 (Fig. activate pro-MMP-1 and pro-MMP-3. All of these assays were lb), TIMP-1 (Fig. 2a) and TIMP-2 (Fig. 2b),and MMP-3 was performed in the presence of 2 mM phenylmethane sulfonyl occasionally positive in stromal macrophages (Fig. lc). Stainfluoride and 5 mM N-ethylmaleimide to inhibit serine and ing of stromal cells was intensely positive in cancer tissue, but it cysteine proteinases. One unit of activity is defined as the was weak or almost negative in normal tissue remote from the carcinoma. No specimens were positively stained in cancer amount of enzyme that degrades 1 pg substrate/min at 37°C. tissue when reacted with non-immune mouse IgG (Fig. le). In addition, MMP-2 staining on the carcinoma cell membranes Gelatin zymography Zymography in SDS-PAGE containing 0.2% gelatin was was abolished when the primary antibody was absorbed with performed. Culture media were incubated at 37°C for 20 min pro-MMP-2 (Fig. I f ) . When the immunostain in the carcinoma cells was compared in SDS sample buffer without reducing agent and then electrophoresed on 8.5% polyacrylamide gels at 4°C. After between the early and advanced carcinoma groups, MMP-2 electrophoresis, gels were washed in 2.5% Triton-X 100 to was localized exclusively in the advanced carcinomas (42% vs. remove SDS, incubated for 16 hr at 37°C in 50 mM Tris-HCI, 0% of early carcinomas; Table I). MMP-9 was also immunopH 7.4,0.15 M NaCI, 10 mM CaC12 and 0.02% NaN3 and then stained in more cases of advanced (42%) than early (15%) MMPs AND TIMPs IN GASTRIC CANCER 11 FIGURE 1 - Immunolocalization of MMP-1, MMP-2, MMP-3 and MMP-9 in gastric carcinoma tissues. Immunostaining was performed with murine MAbs to MMP-1 (a), MMP-2 (b), MMP-3 (c), MMP-9 (d), non-immune mouse IgG (e) or antibody to MMP-2 absorbed with pro-MMP-2 (f), as described in “Material and Methods”. Note that MMP-1 and MMP-9 are immunostained within the adenocarcinoma cells, whereas MMP-2 is localized on carcinoma cell membranes, and that staining was abolished by absorption. Positively stained stromal fibroblasts (arrows) and inflammatory cells (arrowheads) are indicated. Hematoxylin counter-stain. Bar: 50 pm. N O M U R A ETAL. 12 FIGURE2 - Immunolocalization of TIMP-1 and TIMP-2 in gastric carcinoma tissues. Tissues were inimunostained with antibodies against TIMP-1 (a) and TIMP-2 (b), as described in “Material and Methods”. Note positive staining for TIMP-1 in stromal cells and for TIMP-2 in both stromal and carcinoma cells. Hematoxylin counter-stain. Bar: 50 pm. TABLE 1- IMMUNOHISTOCHEMICAL DATA OF MMPs AND TlMPs IN GASTRIC CARCINOMA CELLS Case Total cases Early carcinoma Advanced carcinoma Percentage of positive cases number MMP-1 MMP-2 MMP-3 MMP-9 TIMP-1 TIMP-2 46 83% 92% 79% 30% 0% 42% 0% 0% 0% 35% 15% 42% 4% 43% 23% 52% 13 33 carcinomas, but there was no statistical significance (Table I). It is notable that MMP-Zpositive cases showed vascular invasion much more frequently (86%) than MMP-2-negative carcinomas (27%; p < 0.01, x2 test). However, there were no correlations between immunolocalization of other MMPs and TIMPs and clinicopathological parameters. Sandwich enzyme immirnoassays To measure the amounts of MMP-1, MMP-2, MMP-3, MMP-9, TIMP-1 and TIMP-2 produced by the carcinoma tissue, EIA systems were applied to culture media. The levels of MMP-I, MMP-2, MMP-3 and MMP-9 in the culture media from tumor tissue were 0.032, 0.197, 0.039 and 0.129 nmol/g weight (median). respectively, and significantly higher than in the control normal tissue ( p < 0.01); molar ratios of MMP-1, MMP-2, MMP-3 and MMP-9 in the tumor samples compared to controls were 5.4 ? 7.7, 3.3 f 3.0, 5.7 f 7.5 and 6.2 2 9.7 (mean 2 SD), respectively (Fig. 3). The TIMP-1 level in the tumor samples (median = 0.179 nmol/g weight) was also significantly higher than that in the normal (median = 0.064), but there was no difference in the TIMP-2 level between the tumor (median = 0.035 nmol/g weight) and control samples (median = 0.033; Fig. 3). Since our EIA for TIMP-2 does not measure TIMP-2 complexed with pro-MMP-2, TIMP-2 amounts were corrected by adding the values of MMP-2 to those of TIMP-2, assuming that all of the MMP-2 detected by the assay was the pro-MMP-2-TIMP-2 complex. When the total amounts of MMP-1, MMP-2, MMP-3 and MMP-9 were divided by the sum of TIMP-1 and corrected TIMP-2 in each case, the following molar ratios were obtained: 1.25 f 1.00 (mean 2 SD) for carcinoma samples and 1.15 f 0.62 for controls, which were not significantly different. 0% 6% Enzyme activities Proteinase activities in culture media were measured using ‘jC-gelatin, 14C-collagenand -?H-Cm-Tfsubstrates in the presence of serine and cysteine prol~einasesinhibitors. Figure 4 shows gelatinolytic activities in the media. The level of the total activities detected in the presence of APMA was significantly higher in the carcinoma samples (median = 33.0 mU/mg weight) than in the control specimens (median = 8.5) ( p < 0.01). The level assayed in the absence of APMA showed that tumor samples (median = 10.5 mU/mg weight) contained more active enzymes than the normal tissue group (median = 4.0) ( p < 0.01) (Fig. 4). However, no definite correlations between activities and clinicopathological parameters were found. Inhibition studies using 4 representative samples from the carcinoma and control normal tissues revealed that EDTA (20 mM) remarkably inhibits gelatinolytic activities (82-100% of activity). Cm-Tf-degrading activity was minimal in both carcinoma (median = 1.4 mU/mg weight) and normal control samples (median = 1.1, n = 33), and no significant difference was seen. Collagenolytic activity was detectable in 3 cases only (n = 30). Gelatin zymography Gelatinolytic activities in culture media from 39 gastric cancer cases were also analyzed by gelatin Zymography. ProMMP-9 of 92 kDa and pro-MMP-2 of 68 kDa were detected in all carcinoma and normal tissue samples, but carcinoma tissue produced remarkably larger amounts of pro-MMP-9 than normal tissue (Fig. 5). Production of pro-MMP-2 appeared to be also higher in most carcinoma specimens than in controls (Fig. 5). Computer-assisted image analyses of the proteolytic bands of pro-MMP-2 and pro-MIUP-9 showed that proteolytic activities of pro-MMP-2 and pro-MMP-9 in carcinoma samples were about 2-fold higher than in normal counterparts (data not 13 MMPs AND TIMPs IN GASTRIC CANCER nmol I g weight ** n . 1.0 F-l 0.8 .. 0.6 n 0.4 a t ** . n 8 ** . I, n .I . 0 t 0.2 . . . . . .. .. NS n 8 . 4 - .. 0 i at 0 T N MMP-I T N MMP-2 T N MMP-3 T N MMP-9 (n=26) (n=31) (n=23) (n=32) T N TIMP-I (n=27) T N TIMP-2 (n=34) FIGURE 3 - Amounts of MMPs and TIMPs in the culture media secreted by carcinoma and normal control tissues. Culture media were prepared by culturing fragments from gastric carcinoma (T) and normal controls (N). MMP-1, MMP-2, MMP-3, MMP-9, TIMP-1 and TIMP-2 were measured by the corresponding EIA systems. Values (nmolig weight) were calculated as described in “Material and Methods”. Bars indicate mean values. **p < 0.01; NS, not significant (Mann-Whitney U test). shown). Since there was a direct correlation between the data obtained by zymography and EIA systems, the assays are considered mutually supportive. The active MMP-2 species of 62 kDa was found on gelatin zymography in 59% of carcinomas (23/39 cases) but in only 23% of controls (9/39 cases) ( p < 0.01, x2 test). When the intensity of the proteolytic bands of pro-MMP-2 and its active species was compared by zymography, the ratio of the 62 kDa-active form to pro-MMP-2 plus active MMP-2 (activation ratio of pro-MMP-2) was significantly higher in carcinoma specimens than in normal controls ( p < 0.01) and in the MMP-2 staining-positive carcinoma group (n = 13) than in the negative group (n = 20) ( p < 0.05). In addition, analysis of the relationships between zymography data and wall penetration of carcinoma cells demonstrated that the activation ratio of pro-MMP-2 was significantly higher in advanced carcinomas (n = 32) than in early carcinomas (n = 7 ; p < 0.05; Fig. 6a). Appearance of active MMP-2 also correlated well with lymphatic permeation (ly): the activation ratio of pro-MMP-2 was remarkably higher in ly-positive cases (n = 27) than in lynegative ones (n = 12;p < 0.05; Fig. 6b).However, no correlation was seen between activation and vascular invasion. The 83 kDa-form of MMP-9 was detected in 31% of carcinoma samples (12/39 cases) but in only 8% of controls (3/39 cases) ( p < 0.01, x2 test), though the ratio of the 83 kDa-MMP-9 to pro-MMP-9 plus MMP-9 exhibited no significant difference between carcinoma and control samples and showed no correlations with any clinicopathological factor. Inhibition studies showed that all gelatinolytic activities of 92, 83,68 and 62 kDa, in addition to higher m.w. species, were almost completely inhibited with 20 mM EDTA and 1 mM o-phenanthroline but not with 2 mM phenylmethane sulfonyl fluoride, 2 mM diisopropylfluorophosphate, 5 mM N-ethylmaleimide or 0.5 mM pepstatin A, indicating that all of the activities belong to a class of metalloproteinases (data not shown). DISCUSSION Our immunolocalization studies have shown that stomach carcinoma cells stain positively for several MMPs including MMP-I, MMP-2 and MMP-9 as well as TIMP-2. The staining pattern of MMP-2 on the plasma membranes of carcinoma cells was unique. Previous studies using in situ hybridization and immunohistochemistry for MMP-2 in human carcinomas (Autio-Harmainen et al., 1993; David et al., 1994; Davies et a[., 1993; Grigioni et al., 1994; Poulsom et al., 1992; Pyke el al., 1993; Wolf et al., 1993) have yielded contradictory datanamely, that stromal fibroblasts, but not carcinoma cells, give strong signals for mRNA, whereas MMP-2 staining is present in the cytoplasm of both carcinoma cells and fibroblasts. This discrepancy has been explained by several possibilities: (i) uptake of fibroblast-derived MMP-2 by carcinoma cells, (ii) differences in the rates of mRNA translation and capacity for intracellular storage of the protein or (iii) differences in the threshold of detection by in situ hybridization and immunohistochemistry (Poulsom et al., 1992). However, the antibody used in our study revealed a novel staining pattern: localization on plasma membranes of the carcinoma cells and within stromal fibroblasts. lmmunostaining was considered to be specific since the antibody absorbed with pro-MMP-2 resulted in no staining. Similar membrane staining of carcinoma cells has been reported in breast cancers (Visscher et al., 1994). A growing body of evidence indicates that carcinoma cells possess binding 14 rnU I mg weight - NOMURA ETAL ** m 400 ** 300 I 1 m 0 m 200 m m 100 m 0 T N APMA( +) (n= 32) T N APMA(-) (n= 32) FIGURE 4 - Gelatinolyticactivities in the culture media of tumor (T) and normal control (N) samples. The assay was performed using I4C-gelatinat 37°C in the presence and absence of APMA [APMA (+) and APMA (-), respectively]. All assays were performed in the presence of 2 mM phenylmethane sulfonyl fluoride and 5 mM N-ethylmaleimide; activities were expressed on the basis of dry weight of each sample. **p < 0.01 (Mann-Whitney U test). sites for pro-MMP-2 on the membranes (Emonard et a/., 1992). In the present study, membrane localization of MMP-2 was observed only in advanced gastric carcinomas, which give tissue reactions with MMP-2-positive fibroblasts. Thus, our data suggest that in gastric carcinoma tissue MMP-2 is produced mainly by stromal fibroblasts and captured on the membranes of carcinoma cells. Inconsistency between immunohistochemistry and in situ hybridization results was also apparent for MMP-1 localization. Although our study showed that MMP-1 is immunolocalized in carcinoma cells and stromal fibroblasts in most cases of stomach cancer, Otani et a/. (1994) have reported that mRNA expression is present only in gastric cancer stromal fibroblasts. However, some colorectal carcinoma cells (Hewitt et a/., 1991) also express MMP-1 at the invasive edges. Our findings on MMP-1 immunolocalization, in addition to previous studies, suggest that MMP-1 is produced by both carcinoma cells and stromal fibroblasts in gastric cancer. However, it is also possible that our antibody recognizes collagenase-3 (MMP-13; Freije et a/., 1994) as well as MMP-1. Our antibody was prepared against the synthetic oligopeptide corresponding to the COOH-terminal domain of pro-MMP-1 (residues 332350) (Zhang et a/., 1993). Although homology of the peptide to the corresponding sequence of MMP-13 is 37%, we cannot RGURE 5 - Gelatin zymography of culture media from stomach tissues. Culture media were electrophoresed under non-reducing conditions on 8.5% SDS-polyacrylamidegels containing gelatin at 4°C; gels were incubated in 50 mM Tris-HCI,pH 7.5,0.15 M NaCI, 10 mM CaCI, and 0.02% NaN, for 16 hr at 37°C after removal of SDS, as described in “Material and “ethods”. Three representative samples of tumors (T) and their normal counterparts (N) are shown. Protein standards include phosphorylase b (94 kDa , human transferrin (77 kDa), BSA (68 kDa), IgG-heavy chain ( 5 kDa), ovalburnin (43 kDa) and carbonic anhydrase (29 kDa). Major gelatinolytic activities with 92 and 83 kDa corresponding to MMP-9 (arrows) and with 68 and 62 kDa corresponding to MMP-2 (arrowheads) are indicated. Authentic pro-MMP-9 (B) and pro-MMP-2 samples (A), the latter of which contains active MMP-2 species of 62 kDa, are also electrophoresed for comparison. 2 exclude the possibility that the epifope of the antibody has a higher homology to some portions of MMP-13. Carcinoma tissues produce MMP-9 (Brown et a/., 1993; Davies et a/., 1993). Previous studies using in situ hybridization have shown that carcinoma cells and/or stromal cells, such as macrophages, express the mRNA for MMP-9 in carcinoma tissues (Wolf et a/., 1993; Pyke et a/., 1993). However, very little is known about the immunolocalization of MMP-9, except one report showing localization in carcinoma cells of breast cancer (Visscher et a/., 1994). Our study has shown that MMP-9 is immunolocalized in gastric carcinoma cells as well as in fibroblasts and macrophages infiltrading tumor tissue, suggesting production by these cells. In co’ntrast, MMP-3 was barely immunostained in gastric carcinomas. Although MMP-3 is expressed in esophageal carcinomas (Shima et a/., 1992), MMP-3 mRNA expression is undetectable in stomach, colon (McDonnell et al., 1991) and breast (Wolf et a/., 1993) adenocarcinomas. These and our results suggest that MMP-3 expression in tumors is more specifiic of squamous cell carcinomas. TIMP-1 and TIMP-2 share biochemical properties as inhibitors of MMPs (Birkedal-Hansen et a/., 1993; StetlerStevenson et a/., 1993), and both TIMPs tended to be expressed mainly in stromal cells of carcinoma tissues (Poulsom et al., 1992; Visscher et a/., 1994; Hewitt et a/., 1991). In gastric cancers, TIMP-2 is immunolocalized in carcinoma cells, inversely correlating with invasiveness and metastasis (Grigioni et a/., 1994). In the present study, TIMP-1 was immunolocal- 15 MMPs AND TIMPs IN GASTRIC CANCER (9) Activation ratio of proMMP-2 * m 0 0.5 Activation ratio of proMMP-2 * I 0.5 A A 0.4 y. 0 m 0.4 A 0 & 0.3 0.2 0 i, 0.3 0.2 A A Y A A A A A t? 6 0.1 0 0 0 t? 0. I 0 lY (-1 (n=12) lY (+I (n=27) FIGURE6 - Correlations of MMP-2 activation with wall penetration (a) and lymphatic permeation (b) of carcinoma cells. Activation rate of pro-MMP-2 was calculated by computer-assisted image analysis of the data from gelatin zymography, as described in “Material and Methods”. Data on early carcinomas (early) and lymphatic permeation-negative (ly-) cases were compared with advanced (advanced) and ly-positive (ly+) cases, respectively. *p < 0.05 (Mann-Whitney U test). ized occasionally, but exclusively to stromal fibroblasts, and TIMP-2 was detected in carcinoma cells as well as fibroblasts without any definite correlation of expression with invasion and metastasis. To examine the quantitative balance between MMPs and TIMPs, we have assayed enzymic activities. Gelatinolytic activities were detected in culture media of carcinoma and normal tissues, and the levels were significantly higher in carcinoma samples than in controls. Our data indicate that the amounts of gelatinases (MMP-2 and MMP-9) exceed those of their inhibitors in the media, in agreement with previous results showing predominance of MMPs over TIMPs in prostatic cancer (Lokeshwar et af., 1993). Actually, our EIA data illustrate the imbalance between the total amounts of MMPs and of TIMPs in favor of the proteinases. Our studies on EIA and gelatin zymography demonstrated significant elevation of MMP-2 and MMP-9 in carcinoma tissues. The enhanced gelatinolytic activity is thus ascribed to over-production of these MMPs. This is also supported by our immunohistochemical findings that many carcinoma cells were positively immunostained. However, discrepancies between EIA data and imrnunostaining were noted with MMP-1, TIMP-1 and TIMP-2. Low levels of MMP-1 in culture media and high immunochemical expression may be due to intracellular storage of MMP-1 in carcinoma cells and/or binding of the enzyme to collagen fibers in tissues. When COS-1 cells were transfected with cDNA for TIMP-1 or TIMP-2, TIMP-1 was recovered mainly in culture media but very little in cell lysates. However, TIMP-2 was present in cell lysates, with only small amounts in media (data not shown). It is known that TIMP-2 can bind to tumor cell membranes, probably via receptors (Hayakawa et al., 1994). The discrepancies of TIMP results may thus be explained by different secretion pathways and binding properties to carcinoma cells. Previous studies have indicated that appearance of active MMP-2 correlates with tumor grade and/or spreading of breast (Davies et al., 1993) and lung (Brown et al., 1993) carcinomas. Our zymographical analyses also demonstrated correlation between activation of pro-MMP-2 and tumor cell invasion, including wall penetration and lymphatic permeation. Although no correlation was observed between activation in culture media and vascular invasion, the degree of carcinoma cell membrane localization of MMP-2 directly correlated with vascular invasion. Our data further support the notion that activation of pro-MMP-2, in addition to its production, in local cancer tissues contributes to local and vascular invasion by malignant tumor cells (Brown et af., 1993; Davies et af., 1993). Our studies on pro-MMP-2 activation in gastric cancer tissues have shown that activation is achieved in carcinoma cell nests via the action of MT-MMP (Nomura et af., 1995), stressing the importance of MT-MMP in the progression of gastric carcinoma through the activation of proMMP-2. Activation of pro-MMP-9 in some lung carcinoma tissue samples was also observed in the present and previous studies (Brown et af., 1993). However, our and other studies (Brown et al., 1993) showed no correlations between proMMP-9 activation and clinicopathological factors. ACKNOWLEDGEMENTS We are grateful to Dr. I. Nakanishi, Department of Pathology, School of Medicine, Kanazawa University, for his useful advice and critical reading of the manuscript. This work was supported by the Hokkoku Foundation for Cancer Research and a Grant-in-Aid (06281107) from the Ministry of Education, Science and Culture of Japan (Y.O.). 16 NOMURA ETAL. REFERENCES T., HOYHTYA, AUTIO-HARMAINEN, H., KARTNNEN,T.. HURSKAINEN, M., KAUPPILA, A. and TRYGGVASON, K., Expression of 72 kilodalton type IV collagenase (gelatinase A) in benign and malignant ovarian tumors. Lab. Invest., 69,312-321 (1993). BIRKEDAL-HANSEN, H., MOORE,W.G.I., BODDEN,M.K., WINDSOR, L.J., BIRKEDAL-HANSEN, B., DECARLO,A. and ENGLER, J.A., Matrix metalloproteinases: a review. Crit. Rev. oral biol. Med., 4, 197-250 (1993). BROWN,P.D., BLOXIDGE, R.E., STUART,N.S.A., GAPER, K.C. and CARMICHAEL, J., Association between expression of activated 72kilodalton gelatinase and tumor spread in non-small-cell lung carcinoma. J. nut. Cancerhist.. 85,574-578 (1993). J.M., HOLM,R. and SOBRINHO-SIMOES, M., DAVID,L., NESLAND, Expression of laminin, collagen IV, fibronectin, and type IV collagenase in gastric carcinoma. Cancer, 73,518-527 (1994). DAVIES,B., MILES,D.W., HAPPERFIELD, L.C., NAYLOR,M.S., BoBROW, L.G., RUBENS, R.D. and BALKWILL, F.R., Activity of type IV collagenases in benign and malignant breast disease. Brit. J. Cancer, 67, 1126-1131 (1993). EMONARD, H.P., REMACLE,A.G., NOEL, A.C., GRIMAUD,J.-A., STETLER-STEVENSON, W.G. and FOIDART,J.-M., Tumor cell surfaceassociated binding site for the M,72,000 type IV collagenase. Cancer Res., 52,5845-5848 (1992). FREIJE, J.M.P., DiEZ-ITZA. I., BALB~N, M., S ~ C H E L.M., Z , BLASCO, R., TOLIVIA, J. and LOPEZ-OT~N, C.: Molecular cloning and expression of collagenase-3, a novel human matrix metalloproteinase produced by breast carcinomas. J. biol. Chem., 269,16766-16773 (1994). N., IWATA,K., SHINYA, T., OKADA,Y. and FUJIMOTO, N., HOSOKAWA, HAYAKAWA, T., A one-step sandwich enzyme immunoassay for inactive precursor and complexed forms of human matrix metalloproteinase 9 (92-kDa gelatinaseitype IV collagenase, gelatinase B) using monoclonal antibodies. Clin. chim.Acfa, 231,79-88 (1994). FUJIMOTO, N., MOURI,N., IWATA, K., OHUCHI, E., OKADA,Y. and HAYAKAWA, T., A one-step sandwich enzyme immunoassay for human matrix metalloproteinase 2 (72-kDa gelatinaseitype IV collagenase) using monoclonal antibodies. Clzn. chim. Acfa, 221,91-103 (19930). T., OKADA, Y. and FUJIMOTO, N., ZHANG,J., IWATA,K., SHINYA, HAYAKAWA, T., A one-step sandwich enzyme immunoassay for tissue inhibitor of metalloproteinases-2 using monoclonal antibodies. Clin. chim. Acta, 220,31-45 (1993b). GRIGIONI, W.F., D'ERRICO, A,. FORTUNATO, C., FIORENTINO, M., MANCINI,A.M., STETLER-STEVENSON, W.G., SOBEL,M.E., LIOTTA, L.A.. ONISTO,M. and GARBISA,S., Prognosis of gastric carcinoma revealed by interactions between tumor cells and basement membrane. Mod. Pathol., 7,220-225 (1994). T., YAMASHITA, K., OHUCHI, E. and SHINAGAWA, A,, Cell HAYAKAWA, growth-promoting activity of tissue inhibitor of rnetalloproteinases-2 (TIMP-2). J. CeIlSci., 107,2373-2379 (1994). HEWITT,R E , LEACH,I.H., POWE,D.G., CLARK, I.M., CAWSTON, T.E. and TURVER. D.R.. Distribution of collagenase and tissue inhibitor of metalloproteinases (TIMP) in colorecta tumors. Znt. J. Cancer, 49, 666-672 (1991). JAPANESE RESEARCHSOCIETYFOR GASTRICCANCER:The general rules for the Gastric Cancer Study in Surgery and Pathology: 11. Histological classification of gastric cancer. Jpn. J. Surg., 11, 140-145 (1981). KODAMA, S., IWATA,K.. IWATA,H., YAMASHITA, K. and HAYAKAWA, T., Rapid one-step sandwich enzyme innmunoassay for tissue inhibitor of metalloproteinases. An application for rheumatoid arthritis serum and p1asma.J. imnunol. Mefh., 127, 103'-108 (1990). LOKESHWAR, B.L., SELZER,M.G., BLOCK, N.L. and GUNJA-SMITH, Z., Secretion of matrix metalloproteinases and their inhibitors (tissue inhibitor of metalloproteinases) by human prostate in explant cultures: reduced tissue inhibitor of metallo roteinases secretion by malignant tissues. Cancer Res., 53,4493-4498 f1993). S., NAVRE,M.: COFFEY,R.J., JR. and MATRISIAN, L.M., MCDONNELL, Expression and localization of the matrix metalloproteinase pump-1 (MMP-7) in human gastric and colon carcinomas. Mol. Carcinogenesis, 4,521-533 (1991). NOMURA, H., SATO,H., SEIKI,M., MAI,M. and OKALIA, Y., Expression of membrane type-matrix metalloproteinase in human gastric carcinomas. Cancer Res., 55,3263-3266 (1995). OBATA,K., IWATA,K., OKADA,Y., KOHRIN, Y., OHUCHI, E., YOSHIDA, S., SHINMEI, M. and HAYAKAWA, T., A one-step sandwich enzyme immunoassay for human matrix metalloproteinase 3 (stromelysin-1) using monoclonal antibodies. Clirz. chinr. Acta, 211,59-72 (1992). OTANI,Y., OKAZAKI, I., ARAI, M., KAMEYAMA, K., WADA, N., K., YOSHINO,K., KITPJIMA,M., HOSODA,Y. and MARUYAMA, TSUCHIYA, M., Gene expression of iinterstitial collagenase (matrix metalloproteinase 1) in gastrointestinal tract cancers. J. Gastroenterol., 29,391-397 (1994). POULSOM, R., PIGNATELLI, M., STETLER-STEVENSON, W.G., LIOTTA, P.A.,JEFFERY, R.E., LONGCROFT, J.M., ROGERS,L. and L.A., WRIGHT, STAMP,G.W.H., Stromal expression of 72 kda type IV collagenase (MMP-2) and TIMP-2 mRNAs in colorectal neoplasia. Anier. J. Patho/., 141,389-396 (1992). PYKE?C., RALFKIAER, E., TRYGGVASON, K. and DAN^, K.. Messenger RNA for two type IV collagenases is located in stromal cells in human colon cancer. Amer. J. Pathol., 142,359--365 (1993). SATO,H., TAKINO, T., OKADA,Y., CAO,J., SHINAGAWA, A., YAMAMOTO,E. and SEIKI,M., A matrix metalloproteinase expressed on the surface of invasive tumor cells. Nature (Lond.), 370,61-65 (1994). SHIMA,I., SASAGURI, Y., KUSUKAWA, J., YAMANA,H., FUJITA,H., KAKEGAWA, T. and MORIMATSU, M., Production of matrix metalloproteinase-2 and metalloproteinase-3 related to malignant behavior of esophageal carcinoma. Cancer, 70,2747-2753 (1992). STETLER-STEVENSON, W.G., AZNAVOORIAN, S. and LIOTTA,L.A., Tumor cell interactions with the extracellular matrix during invasion and metastasis. Annu. Rev. CellBioL, 9, 541-573 (1993). VISSCHER, D.W., HOYHTYA, M., OPOSE~N, S.K., LIANG,C.-M., SARKAR, F.H.. CRISSMAN, J.D. and FRIDMAN, R., Enhanced expression of tissue inhibitor of metalloproteinases-2 (TIhdP-2) in the stroma of breast carcinomas correlates with tumor recurrence. Int. J. Cancer, 59, 339-344 (1994). WOLF,C., ROUYER, N., LUTZ,Y . , ADIDA,C., LORIOT, M., BELLOCQ, J.-P., CHAMBON, P. and BASSET,P., Stromelysin 3 belongs to a subgroup of proteinases expressed in breast carcinoma fibroblastic cells and possibly implicated in tumor progression. Proc. nut. Acad. Sci. (Wash.), 90,1843-1847 (1993). ZHANG,J., FUJIMOTO, N., IWATA,K., SAKAI,T., OKADA,Y. and HAYAKAWA, T., A one-step sandwich enzyme immunoassay for human matrix metalloproteinase 1 (interstitial collagenase) using monoclonal antibodies. Clin. chirn. Acfa, 219, 1-14 (1993).