Expression of Matrix Metalloproteinases in the Microenvironment of Spontaneous and Experimental Melanoma Metastases Reflects the Requirements for Tumor Formation

Remodeling of the ECM¹ is an essential step for cell migration (1). MMPs belong to a rapidly growing family of zinc-dependent endopeptidases that are capable of degrading a variety of ECMs (2). On the basis of their structure and substrate specificity, they have been subgrouped into soluble and MT-MMPs (3, 4). MT-MMPs are localized to the cell surface via a transmembrane domain or a glycosylphosphatidylinositol anchor (4, 5). In general, MMPs are secreted in latent forms (pro-MMPs) and require specific proteolytic activation (6). MT1-MMP was first discovered as a physiological activator of MMP-2 on the cell surface; in addition, it facilitates clustering of active MMP-2 at invadopodia of tumor cells (7). MMP activity is additionally regulated by four structurally related tissue inhibitors (TIMP-1 to TIMP-4; Ref. 8).

Recent studies have shown that expression of MT1-MMP and activation of MMP-2 in melanoma cells is correlated with tumor progression (9). Both molecules are found predominantly at the tumor invasion front (10). However, their expression is not restricted to tumor cells, but surrounding and/or infiltrating stromal as well as immune cells also express MMPs and TIMPs. Recent in vitro experiments suggest that tumor cells may use the MMP-2 produced by adjacent normal stromal cells to facilitate their own invasion (11). However, until now the complex interactions between tumor cells and tumor surrounding stromal cells in vivo, as well as the differential expression of MMPs in relation to their microenvironment during progression are not fully understood (12). Therefore, we took advantage of a syngeneic murine melanoma model (13), which is characterized by the possibility to induce experimental and spontaneous metastases in various tissues, to scrutinize the expression of MMP-2, MMP-9, MT1-MMP, TIMP-1, and TIMP-2 in vivo. This analysis revealed differential patterns of MMP expression during different steps of metastasis formation.

The spontaneous metastatic clones of the B16 melanoma cell line, designated B16G3.26, B16G3.12II, and B16G3.12BM2, were obtained originally from Christopher W. Stackpole (New York Medical College, Valhalla, NY). Of these cell lines B16G3.12BM2 showed the most aggressive metastatic behavior in vivo (13). Human melanoma cell line MV3 has been described previously and was used as a positive control for TIMP-1 (14) . Human fibrosarcoma cell line HT1080 was obtained from the American Type Culture Collection (Rockville, MD) and served as a positive control. HT23 cells were derived from HT1080 cells by stable transfection with MT1-MMP cDNA (kindly provided from Dr. Alex Y. Strongin, The Burnham Institute, La Jolla, CA) were used as positive controls. Cells were cultured in RPMI 1640 (PAN Biotech GmbH, Aidenbach, Germany), supplemented with 10% FCS, penicillin (100 units/ml), and streptomycin (0.1 mg/ml) in an atmosphere of 95% humidified air and 5% CO₂ at 37°C. For passaging, cells were detached with 0.05% trypsin and 0.02% EDTA in PBS for 5 min at 37°C.

Female C57BL/6 mice were obtained from The Jackson Laboratory (Charles River, Sulzfeld, Germany) at the age of 4–6 weeks and housed under specific pathogen-free conditions. Five × 10⁴ cells in 100 μl PBS were injected s.c. into the flanks of the mice. s.c. tumors (n = 8 per cell line) were surgically removed when they reached a diameter of ∼1 cm and snap-frozen in liquid nitrogen. After an additional 3 months the animals were sacrificed and examined for spontaneous metastases. Spontaneous lymph node (n = 4 per cell line) and lung metastases (n = 4 per cell line) were isolated, snap-frozen, and stored at −80°C. For experimental lung colonization (n = 4 per cell line) 5 × 10⁴ tumor cells in 100 μl PBS were injected into the lateral tail vein. The animals were sacrificed after 11 and 15 days, and the lungs were snap-frozen and stored at −80°C.

For MMP-2 a rabbit polyclonal antibody (#AB808; Chemicon International Inc., Temecula, CA) was used in Western blot analysis and a rabbit polyclonal antibody (#AB809; Chemicon International Inc.) in immunohistochemistry. For Western blot analysis a mouse monoclonal antibody (IM37L; Oncogene Research Products, Cambridge, MA) and in immunohistochemistry a rabbit polyclonal antibody (#AB19047; Chemicon International Inc.) were used against MMP-9. For MT1-MMP a rabbit polyclonal antibody (#AB815; Chemicon International Inc.) was used in immunohistochemistry, Western blot analysis, and flow cytometry. For Western blot analysis monoclonal antibodies against TIMP-1 (#IM32L; Oncogene Research Products) and TIMP-2 (#IM11L; Oncogene Research Products), and in immunohistochemistry rabbit polyclonal antibodies against TIMP-1 (#AB800; Chemicon International Inc.) and TIMP-2 (#AB19029; Chemicon International Inc.) were used. Polyclonal rabbit antibody αPEP8 (generously provided by Dr. Vincent J. Hearing, National Cancer Institute, Bethesda, MD) was used as a control for mouse TRP2-specific staining on tumor cells.

To determine cell surface expression of MT1-MMP on melanoma cell lines, cells were detached from the tissue-culture flask nonenzymatically with EDTA 2 mm for 10 min. Two × 10⁵ cells were incubated with antibody against MT1-MMP in 0.1% BSA in PBS for 45 min on ice, followed by centrifugation for 5 min at 1500 × g at 4°C. After incubation with FITC-labeled antirabbit immunoglobulin antibody (#111–096-045; Jackson ImmunoResearch Laboratory) for 45 min on ice, cells were centrifuged for 5 min at 1500 × g at 4°C. Upon washing, FITC-labeled cells were analyzed with a FACScan flow cytometer (Becton Dickinson, San Jose, CA).

Western blot analysis was performed as described (10). Briefly, the protein amount in SFCM was measured using BCA Protein Assay Reagent (Pierce, Rockford, IL). Samples of 12 μl of 10 × concentrated (Centricon 10 microconcentrator; Amicon, Beverly, MA) SFCM (10 μg total protein per lane) were electrophoresed on 7.5% SDS-PAGE and transferred to nitrocellulose membranes. Membranes were blocked and were probed with antibodies to the indicated MMPs. After incubation with peroxidase-labeled swine antirabbit immunoglobulin or rabbit antimouse immunoglobulin (Dako, Hamburg, Germany) secondary antibodies, bands were visualized using by chemiluminescence (Boehringer Mannheim, Penzberg, Germany). Western blot analysis was performed three times with different cell extracts giving comparable results.

To study the expression of MMP-9, MMP-2, MT1-MMP, TIMP-1, and TIMP-2 in tissue, 6-μm serial cryostat sections were stained. Immunohistochemistry was performed as described elsewhere (10). Sections were scored independently by two observers (U. B. H. and J. C. B.). Discrepancies were found in only four cases that were re-evaluated jointly until agreement was reached. For each section the percentage of positive melanoma cells was estimated by counting 500 consecutive malignant cells. A lesion was considered as positive if minimally 1% of tumor cells stained moderately or markedly. Weak staining was disregarded. For each section, the pattern of tumor cell staining was judged as diffuse or focal positivity, and the involvement of tumor center and/or tumor border was noted. For stromal cells, the extent of staining and the type of positive cells (fibroblast-like cells, macrophages, and endothelial cells) were recorded. For each cell line, three consecutive sections per tumor were stained, and they showed comparable results.

Activation status of MMP-2 and MMP-9 was studied by zymography analysis and was performed as described before (10). Briefly, 5-μl aliquots of SFCM were diluted 1:1 in sample-buffer. Samples were electrophoresed at 4°C on SDS-PAGE gels containing 50 mg/ml gelatin. After electrophoresis, gels were washed three times in 2.5% Triton X-100 for 10 min to remove SDS. After rinsing twice in substrate buffer [50 mm Tris-HCl (pH 7.8), containing 5 mm CaCl₂ and 0.1% Triton X-100], gels were incubated at 37°C for 18 h in the same buffer under gentle agitation. Gels were stained for 45 min in 40% methanol/10% glacial acetic acid containing 0.1% (w/v) Coomassie Brilliant Blue R 250 and destained in the same solution without Coomassie Brilliant Blue. MMP-9 (M_r 105,000, murine, and M_r 92,000, human), latent pro-MMP-2 (M_r 66,000), and activated MMP-2 (M_r 62,000) were identified by using the peqGOLD protein marker (peqlab, Erlangen, Germany) and SFCM of HT1080 cells as a control. For zymography analysis of tumor tissue, eight serial tissue cryostat sections (6 μm) were cut from total tumor tissue, tumor interior, and periphery, weighed, and homogenized in 4 × sample buffer, centrifuged for 1 min at 10,000 × g, and pellets were incubated at 37°C for 30 min. Samples were mixed with distilled water (1:1) and heated at 55°C for 20 min. After centrifugation (1 min; 10,000 × g) pellets were discarded, and supernatants were subjected to zymographic analysis. Zymography was performed at least three times for each sample.

Protein concentrations of different MMPs and TIMPs were determined by Western blot analysis in 10 × concentrated SFCM derived from 1 × 10⁷ cells of three subclones of murine B16 melanoma, i.e., B16G3.26, B16G3.12II, and B16G3.12BM2, characterized by different metastatic behavior in vivo (13). MMP-2, MMP-9, MT1-MMP, TIMP-1, and TIMP-2 were released by all of the cell lines; however, an increased expression of MMP-2, MMP-9, and TIMP-1 was restricted to the most aggressive cell line, i.e., B16G3.12BM2. Additionally, only in this cell line active MMP-2 (M_r 62,000) was present as determined by an antibody recognizing active MMP-2 (Fig. 1). TIMP-2 protein was secreted in equal amounts by all of the cell lines. Flow cytometry and Western blot analysis showed comparable levels of membrane-bound and soluble MT1-MMP (Fig. 2, A and B). As measured by flow cytometry the mean fluorescence per cell of B16G3.26 was 27.00, B16G3.12II 16.44, and B16G3.12BM2 25.78; the mean fluorescence of the positive control, i.e., HT23 cells, was 92.54.

Serial sections of s.c. tumors derived from all three of the melanoma cell lines were used to study the expression of MMP-2, MMP-9, MT1-MMP, TIMP-1, and TIMP-2 in vivo. Immunohistochemistry for MMP-2 and MT1-MMP showed moderate to strong staining of individual cells in the tumor center and a strong staining of mainly tumor surrounding stromal cells at the tumor border (Fig. 3, A, B, and M). Staining for the melanoma-associated antigen TRP-2 on serial sections allowed us to discriminate between tumor and stromal cells (Fig. 3,N). To this end, tumor surrounding stromal cells, isolated tumor cells, and tumor-infiltrating cells in the tumor center, as well as some tumor cells at the tumor border stained strongly positive for MMP-9 (Fig. 3,C). In s.c. tumors a higher number of cells were positive for MMP-2 and MT1-MMP compared with MMP-9. In all of the lesions studied, ∼50% of the tumor cells in the tumor center or at the tumor border were positive for TIMP-2 (Fig. 3 O). A moderate immunoreactivity against TIMP-1 was found in some tumor cells at the tumor border, whereas a strong staining was observed in endothelial cells, nerves and sebaceous glands adjacent to the tumors (data not shown). Notably, s.c. tumors derived from all three of the cell lines showed comparable staining results for each component.

Zymography was performed on tissue sections of the total tumors as well as of microdissected tumor interiors and peripheries to determine the activation status of MMP-9 and MMP-2. In total tumor extracts, MMP-9 (M_r 105,000), and both latent (M_r 66,000) and active form (M_r 62,000) of MMP-2 were present in similar amounts (Fig. 4,A). Higher levels of MMP-2, however, were observed at the invasive front as compared with the central part of the tumors (Fig. 4 B). Moreover, the presence of activated MMP-2 was restricted to the invasive front.

To scrutinize the effect of the surrounding tissue on the expression of MMPs, we studied the expression of MMP-2, MMP-9, and MT1-MMP in experimental lung metastases. Surprisingly, the strongest expression for MMP-9 was observed for immune cells that were localized in the lung tissue. The number of these MMP-9-positive cells increased 2-fold on day 15 compared with day 11 (Fig. 3, J and K). Notably, in these experimental lung metastases tumor cells were virtually negative for MMP-9 (Fig. 3, J and K). The same applied for MMP-2 and MT1-MMP (data not shown). The presence of experimental lung metastases was confirmed by staining for TRP-2 (Fig. 3 L). No differences between any of the three cell lines with regard to MMP expression in the tumor cells or the induction of these enzymes in the reactive inflammatory infiltrate were observed.

After ∼3 months all three of the murine melanoma cell lines spontaneously formed lymph node and lung metastases after s.c. injection into the flank of C57BL/6 mice. It should be noted, however, that the incidence and degree of metastasis varied as had been described previously with B16G3.12BM2 being the most aggressive cell line (13). Immunohistochemistry for MMP-2, MMP-9, and MT1-MMP of spontaneous lymph node and lung metastases allowed for the comparison of the differential expression of MMPs in experimental and spontaneous metastases, as well as in relation to their microenvironment. In spontaneous lymph node metastases MMP-2 and MT1-MMP display a similar distribution pattern with a positive staining of individual tumor cells within the tumor and a strong immunoreactivity at the invasive front as experimental s.c. tumors (Fig. 3, D and E). In 8 of 12 cases (66.6%) of lymph node metastases a high number (50–100%) of tumor cells stained moderately positive for MMP-9 (Fig. 3,F). Similar to experimental lung metastases, tumor cell clusters of spontaneous lung metastases were negative for MMP-2 and MT1-MMP (Fig. 3, G and H). Bronchial epithelia and lung macrophages stained positively for both components. A very striking difference, however, between experimental and lung metastases was that in 10 of 12 cases of spontaneous tumors ∼50% of the tumor cells expressed MMP-9 both in the tumor center and the tumor periphery (Fig. 3 I). Notably, metastases derived from all three of the cell lines showed comparable MMP-9 expression in the spontaneous metastases irrespective of the different expression pattern in vitro or the different metastatic potential, suggesting the in vivo selection of metastatic subclones.

Melanoma is a highly invasive tumor with a high capacity for both lymphatic and hematogeneous metastasis formation. By means of spontaneous and experimental metastases models we attempted to at least partially dissect this process to establish the relevance of MMPs during the different phases of tumor spread.

Several lines of evidence suggest that proteolytic enzyme systems are important for melanoma progression (15). In human and murine melanoma cell lines increased expression of MMP-2, MT1-MMP, and MMP-9 was shown to correlate with an invasive phenotype (9, 14, 16). Because activation of MMP-2 is dependent on overexpression of MT1-MMP our observation that in human melanocytic lesions the expression of both MMP-2 and MT1-MMP was increased with tumor progression and that positive tumor cells were predominantly localized at the invasive front were in line with this hypothesis (10). In situ zymography studies confirmed that activated MMP-2 is located at the tumor-stroma border (17). The present study confirms in a syngeneic mouse model that MMP-2 and MT1-MMP expression is indeed stressed at the tumor border, and that presence of functionally active MMP-2 was restricted to the invasive front.

The balance between levels of activated MMP and free inhibitors appears to be critical for tumor progression (18). Accordingly, overexpression of TIMP-1, TIMP-2, and TIMP-3 has been shown to inhibit tumor-mediated proteolysis and melanoma invasion (19, 20, 21). However, no evident correlation of TIMP-1 or -2 expression with activation of MMP-2 and MMP-9 was observed in the present in vivo model. Morrison et al. (22) have demonstrated recently the existence of a TIMP-2-independent pathway for cellular activation of MMP-2 involving MT2-MMP. However, additional studies are needed to define the role of TIMPs and MT-MMPs for MMP-2 activation.

MMP expression is not restricted to cancer cells but is also found in the stromal cells, suggesting a contribution of host-derived proteases to tumor progression (12, 23). Studies with MMP-2-deficient mice have revealed that host MMP-2 synthesis is indeed important for experimental metastasis formation (24). We observed in s.c. tumors, spontaneous lymph node, and both experimental as well as spontaneous pulmonary metastases a strong immunoreactivity for MMP-2 and MT1-MMP on stromal and immune cells. The mechanisms by which malignant cells up-regulate MMP-2 expression in normal remain elusive, but signaling via cell surface molecules such as ECM metalloproteinases inducer seem to be involved (25, 26). In this context, it should also be noted that Saad et al. (11) have shown that tumor cells induce the release of MMP-2 from normal fibroblasts via cancer cell-associated fibronectin. Thus, once displaced from the fibroblast-associated collagen, MMP-2 is activated by MT1-MMP/TIMP-2 complexes expressed by tumor cells (11).

Conflicting data have been reported on the role of MMP-9 expression for melanoma progression. In humans MMP-9 was exclusively expressed in the horizontal but not in the vertical growth phase of primary tumors (27). In murine xenograft models, however, MMP-9 was expressed by, or was inducible in, cell lines derived from advanced but not early primary melanoma lesions (28). Furthermore, in cell lines constitutively expressing MMP-9, an enhancement of lung colonization was observed in experimental lung metastasis models, and in MMP-9-deficient mice, metastases formation was suppressed (29, 30). These results suggest that MMP-9 produced either by neoplastic or stromal cells is important for metastasis formation. However, the differential expression and localization of MMPs in relation to the microenvironment and the nature of the metastasis, i.e., experimental or spontaneous, have not been addressed to date. It is important to note that these factors are likely to influence the requirements of protease activity, as outlined above. Our data indicate that these requirements seem to be more dependent on the metastatic process rather than the microenvironment. In spontaneous lymph node and pulmonary metastases the majority of tumor cells expressed MMP-9, whereas experimental metastases did not express MMP-9 at all. Thus, the in vivo selection process for metastatic subclones favors those expressing MMP-9. Because this selection is independent of the microenvironment and is not seen for experimental metastases, MMP-9 seems to be involved in the early steps of metastasis formation but not the later steps such as extravasation.

In summary, our data support the notion that expression of MMP-2 and MT1-MMP at the tumor-stromal border is important for the progression and invasion of s.c. tumors, whereas MMP-9 is essential for spontaneous metastasis formation. Thus, expression of MMPs is not only influenced by localization but also the nature of metastasis induction, suggesting that these different MMPs serve individual and specific roles during the different stages of metastasis formation.