Abstract
Tumor cells interact with stromal cells via soluble or cell-bound factors stimulating the production of matrix metalloproteinases (MMPs), a group of enzymes largely involved in the extracellular matrix (ECM) remodeling in tumor invasion. Among these factors, extracellular matrix metalloproteinase inducer (EMMPRIN) has been shown to stimulate in vitro the fibroblast production of various MMPs such as interstitial collagenase (MMP-1), stromelysin-1 (MMP-3), and gelatinase A (MMP-2). In this study, the EMMPRIN protein was detected by immunohistochemistry prominently in malignant proliferations of the breast and the lung. It was present at the surface of both tumor epithelial and peritumor stromal cells. Because previous studies have reported that stromal cells do not express EMMPRIN mRNAs, it is very likely that EMMPRIN is bound to stromal cells via a specific receptor. Moreover, our observations also demonstrated that the same peritumor stromal cells strongly express MMP-2. Our results show that EMMPRIN is an important factor in tumor progression by causing tumor-associated stromal cells to increase their MMP-2 production, thus facilitating tumor invasion and neoangiogenesis.
Tumor cells may interact with stromal cells via soluble or cell-bound factors, stimulating MMP production. The best characterized of these factors, a tumor collagenase stimulatory factor (TCSF), recently renamed extracellular matrix metalloproteinase inducer (EMMPRIN), was originally isolated from the LX-1 human pulmonary carcinoma cell line (Ellis et al. 1989; Nabeshima et al. 1991). EMMPRIN is a transmembrane glycoprotein of 57 kD and has been identified as a member of the immunoglobulin superfamily (Biswas et al. 1995). EMMPRIN has been shown to stimulate fibroblast production of interstitial collagenase (MMP-1), stromelysin-1 (MMP-3), and gelatinase A (MMP-2) (Kataoka et al. 1993; Guo et al. 1997). EMMPRIN is present in normal tissues such as epidermis, retinal pigment epithelium, breast lobules, and ductules, suggesting that EMMPRIN may have a physiological role in tissue remodeling by causing induction of stromal MMPs (De Castro et al. 1996; Marmorstein et al. 1996; Finneman et al. 1997). However, EMMPRIN expression is more prominently found in tumor proliferations. EMMPRIN protein has been detected by immunohistochemistry in malignant tumor cells of skin, bladder, and breast carcinomas (Muraoka et al. 1993; Zucker and Biswas 1994; Van den Oord et al. 1997). Furthermore, EMMPRIN transcripts are expressed by tumor cells of lung and breast carcinomas, whereas expression of these mRNAs is undetectable or very weak in normal tissues and benign lesions (Polette et al. 1997). Taken together, all these data suggest that EMMPRIN participates in tumor progression by stimulating the synthesis of specific MMPs by peritumor fibroblasts. Because previous immunohistochemical studies employing a monoclonal antibody against EMMPRIN (E11F4) have shown that this factor is localized at the outer cell surface of several cancer cell lines (Ellis et al. 1989; Van den Oord et al. 1997), it appears likely that tumor cell-stromal cell contact is necessary for EMMPRIN-mediated regulation of MMPs. However, the mechanism of action of EMMPRIN in vivo is still unclear.
All these data prompted us to examine by immunohistochemistry the distribution of EMMPRIN and MMP-2 with regard to the persistence of intact basement membrane, which could be an obstacle for the MMP induction by cell-cell contact in various normal, benign, and malignant proliferations of the breast and lung.
Materials and Methods
Source of Tissue
Tissues were obtained from four normal human bronchi, from 20 human lungs resected for squamous cell carcinoma (14 cases) and adenocarcinoma (six cases), from eight fibroadenomas of the breast, from four intraductal carcinomas of the breast, and from 20 invasive ductal breast carcinomas (four of grade 1, 12 of grade 2, and four of grade 3, according to the Scarf and Bloom classification).
Western Blotting Analyses
Tissue samples were homogenized with a turrax mixer in a lysis buffer (8 g/liter NaCl, 0.01% leupeptin, 1% pefablock, 1% aprotinin). The extracts were centrifuged at 14000 × g for 10 min and the supernatants were kept at −20C until use. Amounts of protein were evaluated with the DC Protein Assay kit (BioRad; Hercules, CA). Ten μg of protein from tissue extracts and 2 μg of recombinant EMMPRIN were mixed with Laemmli buffer (BioRad) containing 5% β-mercaptoethanol (v/v), boiled for 5 min, separated on a 12% SDS-PAGE gel, and transferred to a nitrocellulose filter (Hybond-C Extra; Amersham, Aylesbury, UK). Transfer was monitored with Ponceau red reversible staining. The membrane was blocked with 5% milk overnight at 4C before exposure with the G6.2 antibody (10 μg/ml; Chemicon, Temecula, CA) for 1 hr at room temperature (RT). The membranes were incubated with a secondary biotinylated sheep anti-mouse antibody (1:1500; Amersham, Arlington Heights, IL) for 1 hr at RT and then with the streptavidin-peroxidase complex (1:1500; Amersham) for 30 min. Immunoreactive protein bands were detected with ECL Western blotting reagents (Amersham).
Immunohistochemistry
Fresh samples were frozen in liquid nitrogen, cut at −20C in a Reichert-Jung 2800 Frigocut cryostat at a thickness of 5–8 μm, and transferred onto gelatin-coated slides. Sections were incubated overnight at 4C in a moist chamber with the monoclonal antibody against EMMPRIN (G6.2 at a concentration of 100 μg/ml; Chemicon) and then with anti-mouse IgG biotinylated complex (Amersham) at a 1:50 dilution in PBS-BSA solution for 60 min. Finally, sections were treated with streptavidin-fluorescein isothiocyanate (FITC) (Amersham) at a 1:50 dilution in PBS for 30 min.
Double immunostainings were performed for the simultaneous localization of EMMPRIN/Type IV collagen and EMMPRIN/MMP-2. For these double immunostainings, two sucessive labeling reactions for EMMPRIN/Type IV collagen and EMMPRIN/MMP-2 were done sequentially as follows: (a) detection of EMMPRIN using the monoclonal antibody G6.2 at a concentration of 100 μg/ml in PBS-BSA solution; (b) F(ab')2-fragments anti-mouse IgG-digoxigenin (Boehringer; Mannheim, Germany) at a concentration of 4 μg/ml in PBS; (c) anti-digoxigenin fluorescein Fab fragments (Boehringer) at a concentration of 1.3 μg/ml in PBS-BSA; (d) detection of Type IV collagen using a rabbit biotinylated polyclonal antibody (Institut Pasteur; Paris, France) diluted 1:1000 in PBS-BSA; (e) detection of MMP-2 using the rabbit polyclonal antibody AB809 (Chemicon) then using an anti-rabbit IgG biotinylated complex (Amersham) at a 1:50 dilution in PBS-BSA for 60 min; and (f) streptavidin-Texas Red conjugate (Amersham) at a 1:50 dilution in PBS.
We tested the absence of crossreactivity either by omitting the incubation step with the primary antibody or by replacing the primary antibody by a nonimmune IgG. The sections were counterstained with Harris' hematoxylin solution for 10 sec, mounted in Citifluor antifading solution, and observed with an Axiophot microscope (Zeiss; Oberkochen, Germany) using epifluorescence for conventional microscopy or under a confocal laser scanning microscope (BioRad MRC 600).
Results
EMMPRIN Detection in Lung and Breast Tissues
In Western blotting analyses, the antibody directed against EMMPRIN recognized two major bands of 57 kD (glycosylated form) and 30 kD (non glycosylated form) when evaluated vs normal or tumor tissue extracts (Figure 1). We observed an identical electrophoretic profile between EMMPRIN extracted from tumors and from normal tissues. The detection of the glycosylated form in normal and tumor tissues proves EMMPRIN functionality in both conditions.
By immunohistochemistry, EMMPRIN was detected in all normal and tumor tissues of the breast and lung (Table 1). In the normal ducts and lobules and in fibroadenomas of the breast, EMMPRIN was frequently confined to the apical membrane of epithelial cells (Figure 2a). In the same way, in the normal bronchi there was an accumulation of EMMPRIN at the apical pole of the ciliated cells, as confirmed by confocal microscopic examination (Figure 2b). In these normal or benign conditions, EMMPRIN was also detected with weak labeling at the basolateral pole of epithelial cells. In tumor clusters, EMMPRIN was highly expressed in all malignant cells (Figure 2c), with more intense staining on the cells located at the periphery of the well-differentiated nests (Figure 2d). Positive staining was distributed at the outer cell membrane of these cancer cells with a punctiform pattern (Figure 2e). Furthermore, EMMPRIN was also found in some stromal cells in breast and lung carcinomas close to tumor cells (Figure 2g). There was positivity at the cell surface of isolated elongated cells considered to be fibroblasts. Moreover, some endothelial cells displayed spotty labeling on their cell membranes (Figure 2c). Because our previous data have shown no EMMPRIN mRNAs in peritumor stromal cells (Polette et al. 1997), these observations support the hypothesis that stromal cells would bind EMMPRIN produced by epithelial tumor cells.
Co-localization of EMMPRIN and Type IV Collagen
We next investigated whether basement membrane integrity could be an obstacle to the diffusion of EMMPRIN from tumor cells. Using double immunofluorescence labeling, we found that the stromal positivity of EMMPRIN was not necessarily associated with the absence of Type IV collagen around tumor clusters, suggesting that basement membrane integrity is not a limiting factor to EMMPRIN translocation (Figure 2f).
Colocalization of EMMPRIN and MMP-2
Because EMMPRIN has been shown to stimulate fibroblast production of several MMPs, we looked at MMP-2 localization in relation to that of EMMPRIN. As expected, MMP-2 protein was detected in the benign neoplasms and malignant lesions, whereas this enzyme was rarely detected in normal tissues adjacent to tumors. In the carcinomas, the MMP-2 was largely found both in cancer and peritumor stromal cells, whereas in fibroadenomas it was present only in some sparse fibroblasts at a distance from the proliferating ducts (Table 1). In all carcinomas, we observed that the presence of EMMPRIN in/on stromal cells coincided with the detection of MMP-2 in the same cells (Figure 2g and 2h).
Discussion
This immunohistochemical study demonstrates epithelial expression of EMMPRIN in normal tissues and in various benign and malignant proliferations of the breast and the lung. These data are in agreement with previous reports using a different antibody (E11F4) in the mammary gland (Zucker and Biswas 1994). With the E11F4 antibody, De Castro et al. (1996) have also found that normal keratinocytes express EMMPRIN in vitro and in vivo. In contrast, Muraoka et al. (1993) have detected the presence of EMMPRIN mostly in urothelial malignant cells and considered that this cell surface protein could be a tumor marker for bladder cancers. There is an apparent discrepancy between our immunohistochemical data and our first observations with in situ hybridization on the same kind of tumors. EMMPRIN mRNA transcripts have been detected only in malignant tumor cells but never in normal tissues and benign proliferations (Polette et al. 1997). Nevertheless, in these latter samples, Northern blot analysis has revealed some weak positivity that might reflect low transcriptional activity. Therefore, the immunohistochemical detection of the protein EMMPRIN in normal epithelial cells may correspond to a low basal level of expression and/or to a short half-life of EMMPRIN mRNAs in normal tissues and benign lesions undetectable with in situ hybridization. Even though the EMMPRIN expression in normal epithelial cells was weak, it might be representative of a physiological role in tissue remodeling under particular conditions (De Castro et al. 1996). In addition, an aspect of the biochemical properties of EMMPRIN that we have demonstrated and which needs emphasis is that the EMMPRIN molecule must be glycosylated to be functional (Biswas et al. 1995; Guo et al. 1997). This functionality is retained in normal and tumor tissues because EMMPRIN is detected in its glycosylated form (57 kD) under both conditions. Extensive expression of EMMPRIN in invasive tumor cells of breast and lung carcinomas clearly supports its role in tumor invasion. Moreover, the strong immunoreactivity for EMMPRIN in intraepithelial carcinomas emphasizes that EMMPRIN is implicated in the early stages of tumor progression.
Detection of EMMPRIN in various tissue extracts by Western blotting analyses. The anti-EMMPRIN antibody recognized two major bands of 57 kD and 30 kD in normal and tumor breast extracts (Lanes 1 and 2) and in normal and tumor lung extracts (Lanes 3 and 4). This antibody also detected recombinant EMMPRIN purified from CHO cells transfected with EMMPRIN cDNA (Lane 5).
EMMPRIN localization in breast and lung tissues. EMMPRIN is present on the outer membrane of epithelial cell with a dense labeling at the apical cell membrane (arrows) in lobule normal breast (a). Bar = 70 μm. By confocal microscopic analysis, EMMPRIN is concentrated at the apical compartment of the epithelial cells in the normal airways (arrows) (b). Bar = 40 μm. EMMPRIN is localized in cancer cells (T) and is also found in endothelial cells (arrow) close to tumor nests in lung carcinomas (c). Bar = 70 μm. By confocal microscopic analysis, the EMMPRIN labeling is preferentially distributed at the periphery of well-differentiated clusters (T) (d) and is detected around cancer cells (e). Bars = 40 μm, 4 μm. Stromal positivity for EMMPRIN (arrow) is detected near tumor clusters where Type IV collagen (orange) is present (f). Bar = 70 μm. In the same section, EMMPRIN (g) and MMP-2 (h) are detected in both tumor (T) and peritumor stromal cells (arrows). Bars = 40 μm.
Distribution of EMMPRIN and MMP-2 in bronchopulmonary and breast lesions a
a+++, extensive; ++, moderate; +, faint; -, negative; grades 1,2,3, Scarf and Bloom classification.
In addition to this distribution of EMMPRIN on cancer cells, we also found this protein in peritumor stromal cells in malignant lesions. Because the biosynthesis of EMMPRIN by fibroblasts has never been reported in previous in vitro and in vivo studies (Ellis et al. 1989; Muraoka et al. 1993; Zucker and Biswas 1994; Polette et al. 1997; Van den Oord et al. 1997), it is very likely that these peritumor stimulated cells are able to bind the EMMPRIN secreted by epithelial cancer cells. The absence of EMMPRIN protein both in fibroblasts and in endothelial cells of normal mammary and lung tissues suggests that peritumor stromal cells may have acquired the ability to express specific receptors for EMMPRIN during tumor progression. The binding of EMMPRIN to stromal cells could then induce their MMP production. Indeed, we and others have previously showed that stromal cells are the principal source of several MMPs, such as MMP-2 (Poulson et al. 1992,1993; Polette et al. 1993; Pyke et al. 1993). In this study we observed that the presence of EMMPRIN is associated with the expression of MMP-2 in all the cases of carcinomas. By double immunostaining, we detected a concomitant expression of EMMPRIN and MMP-2 in tumor cells and fibroblasts. Furthermore, we have clearly observed a colocalization of EMMPRIN protein and MMP-2 in tumor-associated fibroblasts. These observations are consistent with the hypothesis that EMMPRIN expressed by cancer cells stimulates stromal MMP production via specific receptors.
Even though the cellular mechanism of action of EMMPRIN is not yet well understood, several hypotheses can be drawn from our results. In general, it has been suggested that the plasma membrane localization of EMMPRIN at the periphery of tumor clusters serves to restrict its bioactivity to cells in close proximity. EMMPRIN attached to the plasma membrane via a transmembrane domain could then interact with a cell surface receptor present on peritumor stromal cells via an extracellular domain (Nabeshima et al. 1991; Biswas et al. 1995). In this instance, the presence and/or the persistence of an intact basement membrane would represent an obstacle to the direct stimulating effect of EMMPRIN on fibroblasts adjacent to epithelial cells. This may partly explain the absence of concomitant stromal detection of MMP-2 in normal tissues and its limited expression in benign neoplasms. This limited MMP-2 expression could also reflect the apparent inability of stromal cells in normal tissues to express specific receptors and consequently to respond to EMMPRIN by increased MMP-2 production. Independent of the presence of an intact basement membrane, some invasive tumor cells appear to acquire the capacity to shed EMMPRIN, thereby enabling EMMPRIN activity to extend beyond the direct cell contact effect. We found in our malignant tumors the presence of EMMPRIN in the connective tissue close to tumor cells, even in the presence of Type IV collagen deposited between tumor clusters and the stroma. The cancer-generated basement membranes may also exhibit altered physiochemical properties that might allow easy diffusion of tumor proteins into adjacent stroma. This shed form of EMMPRIN could therefore bind to specific receptors present only in peritumor stromal cells and stimulate them to produce MMPs in carcinomas.
In conclusion, our results describing intense expression of EMMPRIN in cancer tissue clearly show the implication of EMMPRIN in tumor invasion. The detection of EMMPRIN protein on peritumor stromal cells also supports the hypothesis that this factor could be shed from tumor cells and bound to tumor-associated stromal cells. EMMPRIN would further stimulate the stromal production of MMP-2, facilitating tumor invasion and neoangiogenesis. Taken together, these results improve our understanding of the cooperation between cancer and host cells during the process of invasion and metastatis.
Footnotes
Acknowledgements
Supported by the Lions Club of Soissons. We thank Chemicon (Temecula, CA) for generously providing the antibody to EMMPRIN.