¹Molecular Pathology Unit and MGH Cancer Center, Massachusetts General Hospital, Boston, Massachusetts, MA 02129, USA

²Institut Universitaire de Pathologie, 25 Rue du Bugnon, CH-1011 Lausanne, Switzerland

³Ludwig Institute for Cancer Research, Lausanne Branch, Chemin des Boveresses 155, CH-1066 Epalinges, Switzerland

Correspondence to: I Stamenkovic, E-mail: Ivan.Stamenkovic@chuv.hospvd.ch

Shedding of intercellular adhesion molecule 1 (ICAM-1) is believed to play a role in tumor cell resistance to cell-mediated cytotoxicity. However, the mechanism whereby ICAM-1 is shed from the surface of tumor cells remains unclear. In this study, we have addressed the possibility that matrix metalloproteinases are implicated in ICAM-1 shedding. Our observations suggest a functional relationship between ICAM-1 and matrix metalloproteinase 9 (MMP-9) whereby ICAM-1 provides a cell surface docking mechanism for proMMP-9, which, upon activation, proteolytically cleaves the extracellular domain of ICAM-1 leading to its release from the cell surface. MMP-9-dependent shedding of ICAM-1 is found to augment tumor cell resistance to natural killer (NK) cell-mediated cytotoxicity. Taken together, our observations propose a mechanism for ICAM-1 shedding from the cell surface and provide support for MMP involvement in tumor cell evasion of immune surveillance.

Oncogene (2002) 21, 5213-5223. doi:10.1038/sj.onc.1205684

ICAM-1; MMP-9; tumor; proteases; cytotoxicity

Introduction

Cell-cell adhesion is a dynamic process mediated by several classes of adhesion receptors, including lectins, immunoglobulin superfamily members, integrins, cadherins and a variety of proteoglycans (Hynes and Lander, 1992). Physiological cell-cell interactions are tightly regulated, such that events which promote adhesion are counter balanced by those that underlie release. Migration, invasion and interactions between antigen presenting cells (APCs) and effector cells of the immune system are all examples of events where adhesion is followed by detachment once appropriate signal exchange among the interacting cells or between migrating cells and host tissue stroma has occurred. However, the molecular events that signal the cessation of adhesion and the mechanisms that underlie cell detachment from each other or from the extracellular matrix (ECM) are still poorly understood.

There is increasing evidence to suggest that shedding of cell surface receptors as a result of proteolytic cleavage of a portion of their extracellular domain plays an important role in regulating their function (Peschon et al., 1998). By the same token, shedding from the cell surface may provide a regulatory mechanism for the function of adhesion receptors that helps terminate cell-cell and cell-ECM interactions (Hafezi-Moghadam et al., 2001). Such a mechanism seems plausible because it relies on the function of specific proteases whose activity may be regulated according to the physiological requirements of the adhesive process itself and the function it fulfills. Protease activation and substrate cleavage can occur more rapidly than down regulation of cell surface receptor expression, providing a potentially tighter control over specific adhesion events.

Recent work suggests that two subclasses of the metzincin protease family, matrix metalloproteinases (MMPs) and their close relatives a disintegrin and metalloproteinase (ADAMs), are implicated in the shedding of cell surface receptors (Werb, 1997). MMPs are zinc atom dependent endopeptidases that have been proposed to play a primary role in the degradation of extracellular matrix (ECM) proteins. There are currently 28 MMPs, which can be subdivided into two distinct structural categories: secreted MMPs and membrane bound MMPs, also known as MT-MMPs, which contain a transmembrane and intracellular domain (Kleiner and Stetler-Stevenson, 1999; Nagase and Woessner, 1999; Stamenkovic, 2000). Both secreted and membrane-bound MMPs are expressed in an inactive zymogen form and require proteolytic removal of a portion of their N-terminal domain in order to become activated. The combined proteolytic action of known MMPs is believed to have the ability to degrade all of the ECM proteins. More recent evidence suggests that MMPs also cleave growth factor precursors (Bergers and Coussens, 2000) and a variety of cell surface molecules, including adhesion receptors, thereby regulating growth factor and cytokine activity as well as cellular interactions with adjacent cells and the ECM. Specifically, MMP-3 has been observed to cleave E-cadherin (Perl et al., 1998), and metalloproteinase activity has been implicated in the shedding of L-selectin (Gu et al., 1998; Zhao et al., 2001) and the hyaluronan proteoglycan receptor CD44 (Okamoto et al., 1999).

Inter-cellular adhesion molecule 1 (ICAM-1) is implicated in a broad range of transient cellular interactions, which regulate leukocyte homing, activation and effector functions (Springer, 1990, 1994). Thus, the interaction between ICAM-1 and its physiological ligand LFA-1 is implicated in leukocyte arrest on endothelial cells, stabilization of cognate interactions between antigen presenting cells and T lymphocytes and adhesion of cytotoxic T cells and natural killer (NK) cells to target cells (Springer, 1994). Each of these interactions is followed by detachment of the adhering cells, suggesting the existence of mechanisms that signal their termination. Similar to their normal hematopoietic and epithelial counterparts, leukemic cells and carcinomas express ICAM-1, which may be predicted to facilitate targeting by cytotoxic cells. However, ICAM-1 is shed from the surface of tumor cells, and serum levels of soluble ICAM-1 have even been proposed to provide an indicator of tumor burden (Grothey et al., 1998; Harning et al., 1991; Pizzolo et al., 1993). More to the point, shedding of ICAM-1 has been observed to inhibit cell mediated cytotoxicity and provide primary tumor cells as well as tumor cell lines with a mechanism of defense against cytolytic T cells and NK cells (Altomonte et al., 1993; Becker and Brocker, 1995; Becker et al., 1991; Fonsatti et al., 1997; Sanchez-Rovira et al., 1998). Recently, ICAM-1 release from the cell surface has been attributed to MMP activity (Lyons and Benveniste, 1998).

Stimulation of ICAM-1 by its physiological ligand LFA-1 has been observed to induce expression of MMP-9/gelatinase B (Aoudjit et al., 1998). We therefore addressed the possibility that ICAM-1 and MMP-9 might have a functional relationship on the cell surface and that MMP-9 might be responsible for ICAM-1 release. Our observations indicate that MMP-9/gelatinase B proteolytically cleaves ICAM-1 within a membrane-proximal region of the extracellular domain and participates in tumor cell resistance to NK cell-mediated cytotoxicity.

Results

Interaction between ICAM-1 and MMP-9 and MMP-9 activity in HL-60 cells

Because ICAM-1 engagement by ligand or antibody stimulates MMP-9 expression (Aoudjit et al., 1998), we reasoned that MMP-9 might provide a candidate protease for the proteolytic cleavage and shedding of ICAM-1. MMP-9 is typically expressed at low levels or not at all in tumor cells, but its expression can be rapidly induced by a variety of cytokines (Huang et al., 1998; Ismair et al., 1998; Kondapaka et al., 1997), phorbolester (Toth et al., 1997) and cell-cell interactions (Aoudjit et al., 1998). Moreover, MMP-9 has been widely implicated in tumor invasion and metastasis (Bernhard et al., 1990, 1994), and in the promotion of tumor cell survival (Yu and Stamenkovic, 1999; Yu and Stamenkovic, 2000).

We began by addressing the relationship between ICAM-1 and MMP-9 in human promyelocytic leukemia HL-60 cells, in which ICAM-1 is constitutively expressed whereas MMP-9 expression is induced in response to phorbolester stimulation. Gelatin zymography of the conditioned culture media of unstimulated cells revealed a 72 kDa band, corresponding to MMP-2, and a weak 92 kDa band, consistent with pro-MMP-9 (Figure 1, lane a). PMA stimulation of HL-60 cells resulted in a significant increase in pro-MMP-9 secretion, as well as the appearance of an 84 kDa band that corresponds to active MMP-9 (Figure 1, lane b (Mazzieri et al., 1997)). In addition, pro- as well as active MMP-9 were detected in the crude membrane fraction of PMA-stimulated cells (Figure 1, lane d). None of the gelatinolytic bands was observed when the broad spectrum MMP inhibitor 1-10 phenanthroline at a 5 mM concentration was added to the substrate buffer (data not shown).

Western blot (Figure 2a) and gelatin zymography (Figure 2b) analysis of anti-ICAM-1 antibody immunoprecipitates from lysates of HL-60 cells stimulated with PMA revealed the presence of a dominant 92 kDa species corresponding to pro-MMP-9 (Figure 2a,b, lane b), suggesting that the two molecules interact or form part of the same complex. Association of ICAM-1 and MMP-9 was found not to be limited to HL-60 cells because two molecules were observed to co-immunoprecipitate from the lysates of other cell types, including the immortalized breast epithelial cell line MCF10A and the colon carcinoma cell line SW620 (data not shown).

To determine whether ICAM-1 might provide a cell surface docking mechanism for MMP-9, we assessed cell surface localization of MMP-9 in HL-60 cells stably transfected with a cDNA encoding ICAM-1 in anti-sense orientation (Table 1). FACS analysis showed that independent isolates of HL-60 cells transfected with anti-sense ICAM-1 (AS-HL-60 cells) lacked cell surface ICAM-1 expression in the resting state and displayed a reduction of ICAM-1 expression after stimulation with PMA (Table 1). Secretion of proMMP-9 into the cell culture medium by cells transfected with vector only and antisense ICAM-1 was comparable (Figure 3a). In contrast, PMA-stimulated AS-HL-60 cells showed a significant decrease in the intensity of 92, 84 and 79 kDa MMP-9 species in cell membrane extracts compared to HL-60 cells transfected with expression vector only, as assessed by gelatin zymogram analysis (Figure 3b) and confirmed by FACS (data not shown). The gelatinolytic bands of 92, 84 and 79 kDa have been shown to correspond to pro-, active and deglycosylated MMP-9 species respectively (Mazzieri et al., 1997). To address the functional effect of the reduced cell membrane localization of MMP-9 in the AS-HL-60 cells, a cell-mediated collagen degradation assay was performed, in which release into the cell culture medium of ³H-labeled type IV human collagen embedded in Matrigel and incubated with vector only and antisense ICAM-1 transfectants was compared. Phorbolester stimulation of HL-60 cells transfected with vector only increased their ability to degrade the radiolabeled type IV collagen, and the degradation was partially blocked by an MMP blocking peptide and a blocking anti-MMP-9 antibody (Figure 4a), but not by an MMP-3 blocking peptide. In contrast, the supernatant of HL-60 cells stimulated with PMA did not augment the degradation of collagen compared to that of unstimulated cells (Figure 4b). These observations suggest that the collagen degradation is predominantly due to the proteolytic activity of MMP-9 localized to the cell membrane. AS-HL-60 cells displayed a significant reduction in their ability to degrade type IV collagen (Figure 4c), suggesting that ICAM-1 may play an important role in regulating cell surface MMP-9 localization and activity in HL-60 cells.

MMPs are involved in the shedding of ICAM-1 from HL-60 cells

Phorbolester mediated increase in ICAM-1 expression was observed to be associated with significant ICAM-1 shedding into the cell culture medium (Figure 5a). Shedding of ICAM-1 was observed to be partially or completely blocked by MMP inhibitors (Figure 5b). 1,10 phenanthroline, the cyclic peptide MMP-2/MMP-9 inhibitor III (Calbiochem) and a broad spectrum MMP inhibitor (Sigma) induced a significant reduction of the shedding (Figure 5b, lanes b, c and d), whereas a variety of non-MMP protease inhibitors had no detectable effect (data not shown). Because ICAM-1 and MMP-9 co-immunoprecipitate, and because the MMP-2/MMP-9 cyclic peptide inhibitor III had a strong inhibitory effect on ICAM-1 shedding, we tested the possibility that MMP-9 might be implicated in the proteolytic cleavage of cell surface ICAM-1 (mICAM-1).

In vitro cleavage of ICAM-1 by MMP-9

In vitro, purified, active MMP-9 did not cleave ICAM-1 immunoprecipitated from COS cells transfected with the ICAM-1 cDNA (data not shown). In contrast, when the active MMP-9 was added to a crude membrane preparation of COS cells transfected with the ICAM-1 cDNA, an increase in soluble ICAM-1 (sICAM-1) was detected by Western blot analysis of the supernatants after centrifugation of the membranes, as described in Materials and methods (Figure 6, lane b). A cocktail of broad-spectrum protease inhibitors (mini EDTA-free tablets, Roche), excluding MMP inhibitors, had no effect on the release of sICAM-1 (Figure 6, lane c), whereas the broad spectrum MMP inhibitor 1-10 phenanthroline abrogated the effect of MMP-9 (Figure 6, lane d). Because the size of the sICAM-1 of about 80 kDa suggests a cleavage site close to the transmembrane region, it is conceivable that solubilization and immunoprecipitation of ICAM-1 from COS cells resulted in a conformational change that prevented its cleavage in vitro. We therefore developed a soluble form of ICAM-1 composed of the entire extracellular domain tagged with the viral v5 peptide (sICAM-1v5). We also used a fusion protein composed of the extracellular domain of ICAM-1 and the constant region of human IgG1 (ICAMRg, (Aruffo et al., 1990)). Serum free conditioned culture media from COS cells transiently transfected with sICAM-1v5 were collected and incubated with purified, active MMP-9, MMP-3 and MMP-1. Only when sICAM-1v5 was incubated with MMP-9, did the anti-v5 mAb fail to recognize the fusion protein (Figure 7b, lane b). The other MMPs tested had no effect, suggesting that the cleavage of ICAM-1 was specific to MMP-9. Furthermore, when sICAM-1v5 was incubated with MMP-9 and 1-10 phenanthroline, cleavage was completely inhibited (Figure 7b, lane e). Earlier work has shown that MMP-9 does not cleave the v5 peptide (Yu and Stamenkovic, 2000), suggesting that the loss of antibody reactivity is most likely due to the cleavage of ICAM-1 at a site close to that of the fusion itself. To provide further support for this observation, COS cells were transiently transfected with a cDNA encoding the ICAMRg fusion or a cDNA encoding the CD5Rg fusion protein (Aruffo et al., 1990) as a control. Conditioned culture media were collected 72 h later and incubated with p-aminophenylmercuric acid (APMA)-activated MMP-9, MMP-3 and MMP-1 in the appropriate reaction buffer. Only in the presence of active MMP-9 did ICAMRg undergo proteolytic cleavage, as shown by the presence of 110 and 55 kDa bands that correspond to dimeric and monomeric cleaved ICAM-1-Ig moieties (Figure 7b, lane b). The other MMPs had no effect and CD5Rg was not cleaved by MMP-9, providing support to the notion that the proteolytic cleavage site of ICAMRg was not in the IgG portion of the protein.

In a recent study, a consensus MMP-9 cleavage site has been proposed to consist of the P-X-X-Hy-T motif, where X can be any and Hy a hydrophobic residue (Kridel et al., 2001). A sequence consistent with this motif, P-G-N-W-T, is present in the membrane proximal portion of the extracellular domain of ICAM-1 (Figure 8a). To determine whether this sequence may indeed provide the putative proteolytic cleavage site in MMP-9-mediated ICAM-1 shedding, we substituted the proline in position 404 with glutamic acid (P404E). When COS cells were co-transfected with the wild type or v5-tagged cell surface ICAM-1 and MMP-9, sICAM-1 was detected in the culture supernatant (Figure 8b, lanes d and f). In contrast, when the P404E mutated ICAM-1 was co-transfected with MMP-9, no ICAM-1 shedding was detected (Figure 8b, lane d). Soluble ICAM-1 was absent from supernatants of COS cells transfected with either the wild type or P404E mutant ICAM-1 cDNA alone (Figure 8b, lanes a-c). The P404E mutation had no effect on ICAM-1 cell surface expression (data not shown).

ICAM-1 shedding in MDA-MB435 cells transfected with MMP-9 is associated with resistance to NK cell-mediated cytotoxicity

Several studies have shown that sICAM-1 released in the culture medium by cell lines can inhibit the cytotoxic effect of NK and LAK cells (Altomonte et al., 1993; Becker and Brocker, 1995; Becker et al., 1991; Fonsatti et al., 1997; Sanchez-Rovira et al., 1998). To address the effect of MMP-9 in promoting tumor cell resistance to NK-mediated cytotoxicity, we selected the breast cancer cell line MDA-MB435, which expresses mICAM-1, but no MMP-9 and no detectable sICAM-1 in cell culture supernatants. MDA-MB435 cells were stably transfected with cDNAs encoding the human MMP-9 and an MMP-9/CD44 fusion protein in which the whole coding region of MMP-9 has been fused in frame to sequences encoding the transmembrane and intracellular domains of CD44 (Yu and Stamenkovic, 2000). Wild type cells did not express MMP-9 as detected by Western blot and gelatin zymogram analysis, whereas in the transfectants, soluble and fusion protein forms of MMP-9 were detected by both gelatin zymography (data not shown) and Western blot analysis (Figure 9a). Comparable amounts of ICAM-1 were present in the lysates of wild type and transfectant MDA-MB435 cells (Figure 9b, lane d, e and f), but only the cells transfected with the soluble MMP-9 or the transmembrane fusion protein MMP-9/CD44 displayed detectable shedding of ICAM-1 (Figure 9b, lanes b and c). ICAM-1 was not detected in the supernatants of wild type cells (Figure 9b, lane a).

The NK-92MI cells are able to induce lysis of a number of cell lines (Gong et al., 1994; Tam et al., 1999) and were tested for their ability to kill MDA-MB435 cells and the corresponding MMP-9 transfectants. In cytotoxicity assays, depending on the effector/target cell ratio, 50-75% of wt MDA-MB435 cells were lysed by the NK-92MI cells. At an effector/target ratio of 40 : 1, 65% specific lysis was observed (Figure 10). The NK cell line was much less effective (10-28% lysis depending on the level of expression MMP-9 and MMP-9/CD44) in killing MDA-MB435 cells transfected with the soluble MMP-9 or the MMP-9/CD44 fusion protein (Figure 10 and data not shown). To determine that the reduced sensitivity of MDA MB435 transfectant cells was due to MMP-9 activity, the transfectants were incubated with NK-92MI cells in the presence or absence of a variety of MMP inhibitors, including phenanthroline, a non-specific MMP inhibitor peptide GM6001 and its scrambled control derivative (Calbiochem), and the more selective MMP2/MMP-9 cyclic inhibitor peptide III (Calbiochem). Inhibition of MMP activity augmented the NK-mediated cytotoxicity to levels comparable to that observed in wt MDA-MB435 cells (Figure 10). None of these inhibitors altered the NK-mediated cytotoxicity of wt (vector only-transfected) MDA-MB435 cells (data not shown). Addition of soluble ICAM-Rg reduced NK-mediated cytotoxicity of the MMP-9 transfectants a little further, whereas CD5-Rg had no effect. Vector-only transfected MDA-MB435 cells displayed strongly reduced sensitivity to NK-mediated killing in the presence of soluble ICAM-Rg but not in the presence of soluble CD5-Rg (Figure 10). The observation that ICAM-1Rg induced somewhat greater resistance to NK-mediated killing in MDA-MB435 cells than did expression of MMP-9 in the transfectants used may be due to the relatively high concentration (10 g/ml) of soluble ICAM-1 administered.

Discussion

Based on the observation that stimulation of ICAM-1 induces MMP-9 expression (Aoudjit et al., 1998), we investigated the functional relationship between ICAM-1 and MMP-9 and found that ICAM-1 can provide a cell surface docking mechanism for MMP-9 whereas proteolytically active MMP-9 cleaves ICAM-1 and augments tumor cell resistance to NK-mediated cytotoxicity.

The observation that ICAM-1 and MMP-9 co-immunoprecipitate does not in and of itself provide evidence that the two molecules interact on the cell surface. It is conceivable, for example, that another cell surface molecule recruits ICAM-1 and MMP-9 into the same complex, and that anti-ICAM-1 antibody immunoprecipitates contain the entire complex or a portion thereof, MMP-9 being but one component. However, it seems clear that expression of ICAM-1 plays a role in cell surface localization of MMP-9. Reduction of cell surface ICAM-1 expression in HL-60 cells by transfection of ICAM-1 cDNA in antisense orientation resulted in a parallel reduction in cell surface MMP-9 localization, as assessed by membrane extract gelatin zymography, whereas MMP-9 release into the cell culture supernatant remained unaltered. Moreover, collagen IV degradation by antisense ICAM-1 transfectants was decreased, as might be expected from the reduction of the amount of MMP-9 retained on the cell surface. Since ICAM-1 is not known to mediate cell attachment to collagen IV, a plausible explanation is that ICAM-1 either directly interacts with MMP-9 or provides a key stabilizing component of the putative cell surface complex. It is noteworthy that most of the MMP-9 that co-immunoprecipitates with ICAM-1 appears to be in the 92 kDa proenzyme form, even though gelatin zymograms of crude membrane extracts contain both pro- and active MMP-9 species. This observation suggests that ICAM-1 may serve to recruit pro-MMP-9 to the cell surface, rendering it available for activation by other proteases whenever its function is required to fulfill the physiological demands. Consistent with this view, recent evidence suggests that activation of proMMP-9 by components of the urokinase-plasmin system, believed to provide a physiological activation mechanism for several MMPs, requires that the reactants be localized to the cell surface (Mazzieri et al., 1997). Proteolytic cleavage of ICAM-1 by active MMP-9 may explain why little or no processed 84 kDa MMP-9 is found in the anti-ICAM antibody immunoprecipitates.

We and others (Mazzieri et al., 1997; Toth et al., 1997; Yu and Stamenkovic, 2000) have proposed that the proteolytic activity of MMP-9 is most effective when MMP-9 is localized to the cell surface. Cell surface localization of proteolytic activity has been observed for a variety of secreted MMPs other than MMP-9, including MMP-1 (Guo et al., 2000), MMP-2 (Brooks et al., 1996) and MMP-7 (Yu and Woessner, 2000). Moreover, cell surface localization of MT1- and MT2-MMPs which contain a transmembrane domain, has been found to be essential for the ability of the enzymes to promote cell-mediated ECM degradation (Hotary et al., 2000). There is evidence that the cell surface may provide a privileged microenvironment for the preservation of MMP-9 activity from natural inhibitors, including the tissue inhibitors of metalloproteinases (TIMPs) that are believed to rapidly block the activity of soluble MMPs (Toth et al., 1997; Yu and Stamenkovic, 1999, 2000). The mechanisms that underlie MMP-9 localization to the cell surface remain to be fully elucidated. However, there are thus far at least three candidate molecules that may provide cell surface docking sites for MMP-9, a collagen IV chain (Olson et al., 1998), the hyaluronan receptor CD44 (Yu and Stamenkovic, 1999, 2000) and ICAM-1 (present study). CD44-mediated cell surface localization of MMP-9 is restricted to keratinocytes and some carcinomas, suggesting that post-translational modifications of the highly polymorphic CD44 molecule may be key. We have found here that ICAM-1 and MMP-9 co-immunoprecipitate from the lysates of hematopoietic tumors a well as immortalized epithelial cells and carcinomas suggesting a potentially broader association. However, the domains and putative corresponding modifications of both ICAM-1 and MMP-9 that are required for the interaction or complex formation remain to be defined.

The second major observation of the present work is that MMP-9 proteolytically cleaves ICAM-1. Shedding alone of ICAM-1 from the cell surface upon expression of MMP-9 does not provide evidence that MMP-9 is directly implicated in the proteolytic cleavage. MMP-9 could conceivably play an indirect role by activating another protease that is responsible for ICAM-1 shedding. However, the observations that purified active MMP-9 cleaves soluble Ig- and v5-tagged ICAM-1 provide strong support to the notion that MMP-9 is directly responsible for the proteolytic cleavage of ICAM-1. The membrane proximal domain of ICAM-1 contains the P-G-N-W-T motif that is consistent with a recently identified consensus sequence for MMP-9 substrates (Kridel et al., 2001). Mutation of a single residue of this sequence, as well as its entire deletion (data not shown), resulted in the abrogation of MMP-9-mediated cleavage further supporting the view that ICAM-1 is an MMP-9 substrate. However, it remains to be clarified whether this sequence represents the cleavage site itself or whether it plays an essential role in maintaining the site in a conformation that is permissive for MMP-9-mediated cleavage. Although several MMPs used as controls failed to cleave ICAM-1, we cannot exclude the possibility that other metalloproteinases, including some of the ADAMs might share the ability of MMP-9 to cleave ICAM-1, given the substantial overlap in substrate specificity among metzincin family members (Bergers and Coussens, 2000; Birkedal-Hansen et al., 1993; Nagase and Woessner, 1999; Stamenkovic, 2000).

The relationship between ICAM-1 and MMP-9 raises an interesting hypothetical regulatory mechanism for the function of both molecules. Ligand-mediated stimulation of ICAM-1 may help localize the MMP-9 precursor to the cell surface, which may then be poised for autoactivation or activation by other proteases (Mazzieri et al., 1997). Following activation, at a moment specified by signals derived from cell-cell interactions, MMP-9 may cleave ICAM-1 providing a termination signal to ICAM-mediated adhesion. By the same token, MMP-9 may limit its own activity by eliminating its cell surface docking receptor and cell membrane associated protection from the action of inhibitors.

A potentially important pathophysiological effect of MMP-9 expression in tumor cells is the augmentation of tumor cell resistance to cell-mediated cytotoxicity. Although we cannot exclude the possibility that MMP-9 proteolytically cleaves additional cell surface molecules that may play an important part in NK-target cells interaction, shedding of ICAM-1 provides a reasonable mechanism whereby MMP-9 protects tumor cells from NK cell-mediated death. ICAM-1 is expressed on the surface of a broad range of tumor types and its shedding has long been reported to confer partial protection to tumor cells from NK and cytotoxic T cell-mediated killing (18-22). Consistent with these observations, we have shown here that resistance of MDA-MD435 cells to NK cell-mediated cytotoxicity induced by MMP-9 expression and activity correlates with ICAM-1 shedding, and that soluble ICAM-1 Rg can strongly reduce MDA-MB435 cell sensitivity to NK-mediated killing. MMPs have been proposed to promote tumor invasion, survival and growth by a variety of mechanisms, including the degradation of ECM proteins (Kleiner and Stetler-Stevenson, 1999; Shapiro, 1998), activation of latent growth factors (Bergers and Coussens, 2000), inhibition of death receptor engagement by ligand (Mitsiades et al., 2000) and stimulation of angiogenesis (Vu et al., 1998). The present observations provide evidence that MMPs may also play a role in tumor evasion of immune surveillance.

Materials and methods

Cells and reagents

All cell lines were obtained from American Type Culture Collection (ATCC, Rockville, MD, USA). HL-60 cells were cultured in RPMI with 10% FBS (both from GIBCO-BRL, Gaithersburg, MD, USA). MDA-MB435 breast cancer cells were grown in DMEM with 10% FBS. MCF-10A cells were grown in DMEM/F12 medium (1 : 1) supplemented with 5% horse serum, 10 g/ml insulin, and 20 ng/ml epidermal growth factor (all from GIBCO-BRL) and 0.5 g/ml hydrocortisone (Sigma St Luis, MO, USA). NK-92MI cells were cultured in DMEM/F12 medium supplemented with 12.5% FBS and 12.5% horse serum. Phorbol 12-Myristate 13-acetate (PMA), the MMPs inhibitor, 1,10 phenanthroline and the inhibitor peptide (2S,3R)-3-amino-2-hydroxy-4-(4-nitrophenyl)butanoyl-L-leucine were from Sigma. MMP inhibitor I, MMP-2/MMP-9 inhibitor III, MMP-3 inhibitor I, GM6001 and GM6001 Negative Control were from Calbiochem (San Diego, CA, USA). Anti MMP-9 antibody with MMP blocking activity was from NeoMarkers (Fremont, CA, USA).

Cloning of MMP-9 and ICAM-1 genes

Human MMP-9 cDNA was cloned by RT-PCR from HL-60 cells incubated with PMA 100 nM for 18 h. Total RNA was isolated from the cells using the TRIzol reagent (GIBCO-BRL) according to the manufacturer's instruction and cDNA was synthesized from 5 g of total RNA using Superscript II Rnase H- reverse transcriptase (GIBCO-BRL). PCR was performed using synthetic oligonucleotide primers complementary to the sequences 5' and 3' of the coding region of the MMP-9 gene and designed to include a KpnI site at the 5' and an EcoRV site at the 3' end. Primer sequences were: forward primer 5'-CAC GAT GGT ACC ATG AGC CTC TGG CAG CCC CTG GTC-3'; reverse primer 5'-CAC GAC GAT ATC GTC CTC AGG GCA CTG CAG GAT GTC-3'. PCR conditions, using Taq DNA polymerase (Fisher Scientific, Pittsburgh, PA, USA) were: 94°C/30 s, 57°C/30 s and 72°C/60 s for 30 cycles. The PCR product was digested with KpnI and EcoRV (Promega, Madison, WI, USA) and inserted into the pcDNA6/v5/His (Invitrogen Corp., Carlsbad, CA, USA) expression vector digested with the same enzymes. To develop the MMP-9/CD44 fusion, sequences encoding the transmembrane and intracellular domains of CD44 (CD44Tm) were amplified by RT-PCR from HL-60 cells as described above using the following primers (the restriction sites used are indicated in parentheses): TmCD44 forward (EcoRV) 5'-TCT GCA GAT ATC CCA GAA TGC CTG ATC ATC TTG GCA-3'; TmCD44 reverse (XhoI) 5'-CAC GAT CTC GAG CAC CCC AAT CTT CAT GTC CAC ATT-3'. The PCR product was digested with the restriction enzymes EcoRV and XhoI and inserted in the pcDNA6v5/His vector containing the MMP-9 gene and digested with the same enzymes. The cloning of human ICAM-1 cDNA was previously described (Makgoba et al., 1988). A cDNA encoding soluble ICAM-1, composed of sequences encoding the extracellular domain of ICAM-1, was PCR-amplified and inserted into the pcDNA 3.1 V5/His (Invitrogen Corp) expression vector. Full length ICAM-1 cDNA was used as template along with the following oligonucleotide primers: ICAM-1sol forward (BamHI) 5'-CAG GAC GGA TCC ATG GCT CCC AGC AGC CCC CGG CCC-3'; ICAM-1sol Rev (EcoRV) 5'-CAC GAC GAT ATC CCG GGG GGA GAG CAC ATT CAC GGT-3'. The v5-tagged ICAM-1 was inserted into the same vector using the same template and forward primer and the following reverse primer (NotI) 5'-TAG ACT CGA GCG GCC GCC GGG AGG CGT GGC TTG TGT GTT-3'. Anti-sense ICAM-1 was cloned using the full length ICAM-1 as template and the following primers: ICAM AS forward (XbaI) CAC GAC TCT AGA ATG GCT CCC AGC AGC CCC CGG CCC-3'; ICAM AS reverse (HindIII) 5'-CAG GAC AAG CTT TCA GGG AGG CGT GGC TTG TGT GTT-3'. The P404E mutation in the ICAM-1 cDNA was generated by PCR using the following primers (the mutated codon is underlined): P404E forward: 5'-G GAC GAG AGG GAT TGT GAG GGA AAC TGG ACG TGG CCA GAA-3'; P 404 E reverse 5'-GG CCA CGT CCA GTT TCC CTC ACA ATC CCT CTC GTC CAG TCG-3'. The correct sequence and orientation of all the cloned genes was verified by direct sequencing. Development of fusion constructs composed of the extracellular domain of ICAM-1 or CD5 and the constant region of the human IgG1 (encoding ICAMRg and CD5Rg respectively) has been previously described (Aruffo et al., 1990).

Transient and stable transfection

COS cells were transfected with the DEAE-dextran method as described (Stamenkovic and Seed, 1990) and the supernatant or the cell lysate was used 72 h after the transfection. HL60 and MDA MB435 cells were transfected using Superfect (Qiagen, Valencia, CA, USA), according to the manufacturer's instruction and clones resistant to G418 1 mg/ml (Calbiochem) were selected. Culture supernatants and lysates of transfectants were tested by Western blot analysis for expression of the appropriate gene product. HL-60 cells transfected with anti-sense ICAM-I were selected by FACS analysis using an anti-ICAM-1 antibody conjugated with FITC (Pharmingen, San Diego, CA, USA).

Soluble ICAM-1 detection

Twenty ml of serum free supernatants of the cells cultured as indicated in the text were concentrated 100-fold using Centricon Plus-20 centrifugal filters (Millipore, Bedford, MA, USA), resuspended in non-reducing Laemmli sample buffer, resolved by 8% SDS-PAGE and transferred to a nitrocellulose membrane (Hybond ECL, Amersham Corp., Arlington Heights, IL, USA). The membranes were blotted with anti ICAM-1 (clone 15.2 Santa Cruz Biotechnology, Santa Cruz, CA, USA) or anti-V5 tag antibody (Invitrogen Corp).

Crude membrane preparation, gelatin zymography and Western blot analysis

Cells were incubated in serum free medium as indicated in the text. After 18 h the supernatant was collected and the cells were resuspended in TRIS buffer 50 mM pH 7.5 containing NaCl 150 mM and a protease inhibitor cocktail that did not contain MMP inhibitors (Complete mini EDTA free tablets, Roche, Basel, Switzerland). The cells were homogenized in a Dounce homogenizer and centrifuged at 6300 g to remove nuclei and large cell debris. The supernatant was then centrifuged at 10⁵ g for 1 h and the pellet was resuspended in 50 mM TRIS pH 8 containing NP-40 1%, NaCl 150 mM and the protease inhibitor cocktail and incubated on ice for 30 min. The samples were centrifuged again at 10⁵ g for 1 h to remove the detergent-insoluble material and the supernatant was considered to represent the detergent soluble fraction of the crude membrane preparation. Gelatin zymography was performed as described previously (Herron et al., 1986). Briefly, 50 l of serum free cell culture supernatant or 50 g of protein from the detergent soluble fraction of the crude membrane preparation were resuspended in non-reducing Laemmli sample buffer and resolved by 8% SDS-PAGE containing 1 mg/ml gelatin (Fisher, Columbia, MD, USA). Following electrophoresis, gels were washed with 2.5% Triton X-100 to remove SDS and incubated in substrate buffer (50 mM pH 8 TRIS buffer containing CaCl₂ 5 mM) for 18 h at 37°C. Gelatinase activity was visualized by staining the gels with 0.5% Comassie blue for 30 min followed by incubation in the destaining solution (methanol 30%, acetic acid 7%). For Western blot analysis, gels subjected to electrophoresis were transferred to a nitrocellulose membrane and incubated with anti MMP-9 Ab (Ab 3 Oncogene Research Product, Boston, MA, USA).

Immunoprecipitation

HL-60 cells were cultured in serum free medium with or without PMA 100 nM for 18 h and lysed with NP-40 1% in TRIS 50 mM pH 7.4 containing NaCl 150 mM and the protease inhibitor cocktail (lysis buffer). The samples were pre-cleared using a mouse anti IgG1 (Sigma) and G protein beads (Amersham), incubated with anti ICAM-1 (Santa Cruz) or control mouse IgG1 (5 g/mg of protein) for 18 h and with G protein beads (50 l/sample) for 2 h and finally washed five times with the lysis buffer. All the procedures were performed at 4°C. The samples were analysed by Western blot or gelatin zymography as described above.

Collagen IV degradation assay

In the collagen degradation assay, 10 l ³H-labeled type IV human collagen (N-[propionate-2,3-³H];1.08 Ci/g; Du Pont NEN, Boston, MA, USA), were diluted in 4 ml of sterile PBS; 50 l of this solution (about 12 000 c.p.m.) were mixed with 210 l of Matrigel (Becton Dickinson Labware, Franklin Lakes, NJ, USA) and used to coat 23 mm diameter cell culture inserts with 8 m pores (Becton Dickinson Labware). The inserts were placed in 6-well plates (Becton Dickinson Labware). The lower chamber was filled with 500 l of serum free RPMI containing the chemotactic factor macrophage inflammatory protein 1- (MIP 1- Sigma). In the upper chamber, 200 l of the cell suspension (3´10⁵ cells/ml) in serum free medium were added with or without 100 nM PMA and MMP inhibitors as indicated in the text. After 24 h, 100 l aliquots were removed from the upper chamber for assessment of radioactivity by liquid scintillation counting. Chambers with unstimulated cells and no cells provided the controls. All experiments were done in triplicate and are presented as the mean values±s.d.

In vitro cleavage of ICAM-1

Crude membrane fractions of COS cells transfected with V5-tagged ICAM-1 were incubated with purified, MMP-9 (Calbiochem) for 18 h at 37°C. The samples were then centrifuged at 10⁵ g for 1 h and the supernatant subjected to Western blot analysis using anti-ICAM-1 or anti-V5 tag antibodies. Prior to use the MMPs were activated with p-aminophenylmercuricacetate (APMA) according to the manufacturer's instructions and MMP activity was checked by zymography. In separate assays, COS cells were transfected with soluble v5-tagged ICAM-1, ICAM-1-Ig or CD5-Ig cDNAs. After 72 h, the serum free culture supernatants were collected, concentrated using Centricon Plus-20 centrifugal filters and incubated with purified, APMA activated, MMP-1, MMP-3 and MMP-9 for 18 h at 37°C. The samples were then subjected to Western blot analysis using anti-ICAM-1 anti-V5 tag or anti human IgG antibodies.

Cytotoxicity assay and ICAM-Rg purification

Vector only- (wild type) and MMP-9-transfected MDA-MB435 breast cancer cells were trypsinized, resuspended at a density of 2´10⁵ cells/ml in DMEM containing 10% FBS and incubated for 2 h at 37°C with 100 Ci/ml of Chromium-51 (1 Ci/l DuPont NEN). The cells were washed three times with complete medium and 5´10⁴ cells/well were seeded onto 96 well plates. Labeled tumor cells were then incubated for 4 h at 37°C with 100 M phenatroline, 200 M GM6001, 200 M GM6001 Negative Control, 100 M MMP2/MMP9 inhibitor III or incubated without any inhibitor. Treated cells were then washed twice with PBS and incubated with NK-92MI to produce the target effector ratio (T : E) of 1 : 40, together with the MMP inhibitors, while untreated cells were incubated with NK-92MI cells in the absence of any inhibitor. After a 4 h incubation, aliquots of 100 l of supernatant were removed for liquid scintillation counting. The effect of soluble receptor globulins (Rg) was addressed in the following manner: COS cells were transfected with soluble ICAM-1-Ig or CD5-Ig cDNAs. After 96 h, the serum free culture supernatants were collected and passed through a 2 ml column of 50% protein A sepharose. After washing with 20 ml of PBS and 10 ml of water, the purified Rg proteins were released from the column with 25 mM HCl, neutralized with 1 M Tris pH 9.0 and concentrated using Centricon Plus-20 centrifugal filters. Ten g/ml solutions of ICAM Rg or CD5 Rg were used in the cytotoxicity assay. Spontaneous release was measured by incubating the tumor cells in complete medium without NK cells and maximum release was determined by lysing the cells with 10% Triton X-100 in saline phosphate buffer (PBS). The percentage of specific lysis induced by the NK cells was calculated according to the following formula: % specific cytotoxicity=(d.p.m. experimental well - d.p.m. spontaneous release)/(d.p.m. maximum release - d.p.m. spontaneous release). All cytotoxicity assays were done in triplicate.

Acknowledgements

We thank Daniel Haber for the gift of MDA-MB435 cells. This work was supported by USPHS grants CA55735 and GM48614 and grant 3100-065090 from the Swiss National Science Foundation to I Stamenkovic.

Altomonte M, Gloghini A, Bertola G, Gasparollo A, Carbone A, Ferrone S, Maio M. (1993). Cancer Res., 53: 3343-3348. MEDLINE

Aoudjit F, Potworowski EF, St-Pierre Y. (1998). J. Immunol., 160: 2967-2973. MEDLINE

Aruffo A, Stamenkovic I, Melnick M, Underhill CB, Seed B. (1990). Cell, 61: 1303-1313. MEDLINE

Becker JC, Brocker EB. (1995). Recent Results Cancer Res., 139: 205-214. MEDLINE

Becker JC, Dummer R, Hartmann AA, Burg G, Schmidt RE. (1991). J. Immunol., 147: 4398-4401. MEDLINE

Bergers G, Coussens LM. (2000). Curr. Opin. Genet. Dev., 10: 120-127. MEDLINE

Bernhard EJ, Gruber SB, Muschel RJ. (1994). Proc. Natl. Acad. Sci. USA, 91: 4293-4297. MEDLINE

Bernhard EJ, Muschel RJ, Hughes EN. (1990). Cancer Res., 50: 3872-3877. MEDLINE

Birkedal-Hansen H, Moore WG, Bodden MK, Windsor LJ, Birkedal-Hansen B, DeCarlo A, Engler JA. (1993). Crit. Rev. Oral Biol. Med., 4: 197-250. MEDLINE

Brooks PC, Stromblad S, Sanders LC, von Schalscha TL, Aimes RT, Stetler-Stevenson WG, Quigley JP, Cheresh DA. (1996). Cell, 85: 683-693. MEDLINE

Fonsatti E, Altomonte M, Coral S, Cattarossi I, Nicotra MR, Gasparollo A, Natali PG, Maio M. (1997). Br. J. Cancer, 76: 1255-1261. MEDLINE

Gong JH, Maki G, Klingemann HG. (1994). Leukemia, 8: 652-658. MEDLINE

Grothey A, Heistermann P, Philippou S, Voigtmann R. (1998). Br. J. Cancer, 77: 801-807. MEDLINE

Gu B, Bendall LJ, Wiley JS. (1998). Blood, 92: 946-951. MEDLINE

Guo H, Li R, Zucker S, Toole BP. (2000). Cancer Res., 60: 888-891. MEDLINE

Hafezi-Moghadam A, Thomas KL, Prorock AJ, Huo Y, Ley K. (2001). J. Exp. Med., 193: 863-872. MEDLINE

Harning R, Mainolfi E, Bystryn JC, Henn M, Merluzzi VJ, Rothlein R. (1991). Cancer Res., 51: 5003-5005. MEDLINE

Herron GS, Banda MJ, Clark EJ, Gavrilovic J, Werb Z. (1986). J. Biol. Chem., 261: 2814-2818. MEDLINE

Hotary K, Allen E, Punturieri A, Yana I, Weiss SJ. (2000). J. Cell. Biol., 149: 1309-1323. MEDLINE

Huang HY, Wen Y, Irwin JC, Kruessel JS, Soong YK, Polan ML. (1998). J. Clin. Endocrinol. Metab., 83: 1721-1729. MEDLINE

Hynes RO, Lander AD. (1992). Cell, 68: 303-322. MEDLINE

Ismair MG, Ries C, Lottspeich F, Zang C, Kolb HJ, Petrides PE. (1998). Leukemia, 12: 1136-1143. MEDLINE

Kleiner DE, Stetler-Stevenson WG. (1999). Cancer Chemother. Pharmacol., 43: (Suppl) S42-S51. Article MEDLINE

Kondapaka SB, Fridman R, Reddy KB. (1997). Int. J. Cancer, 70: 722-726. Article MEDLINE

Kridel SJ, Chen E, Kotra LP, Howard EW, Mobashery S, Smith JW. (2001). J. Biol. Chem., 276: 20572-20578. MEDLINE

Lyons PD, Benveniste EN. (1998). Glia, 22: 103-112. Article MEDLINE

Makgoba MW, Sanders ME, Ginther Luce GE, Dustin ML, Springer TA, Clark EA, Mannoni P, Shaw S. (1988). Nature, 331: 86-88. MEDLINE

Mazzieri R, Masiero L, Zanetta L, Monea S, Onisto M, Garbisa S, Mignatti P. (1997). EMBO J., 16: 2319-2332. MEDLINE

Mitsiades N, Poulaki V, Tseleni-Balafouta S, Koutras DA, Stamenkovic I. (2000). Cancer Res., 60: 4122-4129. MEDLINE

Nagase H, Woessner JrJF. (1999). J. Biol. Chem., 274: 21491-21494. Article MEDLINE

Okamoto I, Kawano Y, Tsuiki H, Sasaki J, Nakao M, Matsumoto M, Suga M, Ando M, Nakajima M, Saya H. (1999). Oncogene, 18: 1435-1446. MEDLINE

Olson MW, Toth M, Gervasi DC, Sado Y, Ninomiya Y, Fridman R. (1998). J. Biol. Chem., 273: 10672-10681. MEDLINE

Perl AK, Wilgenbus P, Dahl U, Semb H, Christofori G. (1998). Nature, 392: 190-193. Article MEDLINE

Peschon JJ, Slack JL, Reddy P, Stocking KL, Sunnarborg SW, Lee DC, Russell WE, Castner BJ, Johnson RS, Fitzner JN, Boyce RW, Nelson N, Kozlosky CJ, Wolfson MF, Rauch CT, Cerretti DP, Paxton RJ, March CJ, Black RA. (1998). Science, 282: 1281-1284. Article MEDLINE

Pizzolo G, Vinante F, Nadali G, Ricetti MM, Morosato L, Marrocchella R, Vincenzi C, Semenzato G, Chilosi M. (1993). Br. J. Haematol., 84: 161-162. MEDLINE

Sanchez-Rovira P, Jimenez E, Carracedo J, Barneto IC, Ramirez R, Aranda E. (1998). Eur. J. Cancer, 34: 394-398. MEDLINE

Shapiro SD. (1998). Curr. Opin. Cell Biol., 10: 602-608. MEDLINE

Springer TA. (1990). Annu. Rev. Cell Biol., 6: 359-402. MEDLINE

Springer TA. (1994). Cell, 76: 301-314. MEDLINE

Stamenkovic I. (2000). Semin. Cancer Biol., 10: 415-433. Article MEDLINE

Stamenkovic I, Seed B. (1990). Nature, 345: 74-77. MEDLINE

Tam YK, Maki G, Miyagawa B, Hennemann B, Tonn T, Klingemann HG. (1999). Hum. Gene Ther., 10: 1359-1373. Article MEDLINE

Toth M, Gervasi DC, Fridman R. (1997). Cancer Res., 57: 3159-3167. MEDLINE

Vu TH, Shipley JM, Bergers G, Berger JE, Helms JA, Hanahan D, Shapiro SD, Senior RM, Werb Z. (1998). Cell, 93: 411-422. MEDLINE

Werb Z. (1997). Cell, 91: 439-442. MEDLINE

Yu Q, Stamenkovic I. (1999). Genes Dev., 13: 35-48. MEDLINE

Yu Q, Stamenkovic I. (2000). Genes Dev., 14: 163-176. MEDLINE

Yu WH, Woessner JrJF. (2000). J. Biol. Chem., 275: 4183-4191. MEDLINE

Zhao L, Shey M, Farnsworth M, Dailey MO. (2001). J. Biol. Chem., 276: 30631-30640. MEDLINE

Figure 1 Expression of MMP-9 in HL-60 cells. Equal aliquots of HL-60 cells were seeded onto culture plates and incubated with 100 nM PMA in serum-free medium for 18 h. The supernatants were then removed, cells were counted in a hemocytometer and tested for viability by MTT assay analysis. The viability of untreated and PMA-treated cells was found to be comparable in three separate counts (92±3%). The amount of material from the conditioned culture media loaded per lane was normalized to the total number of cells at the end of the 18 h of incubation with or without PMA. The supernatant of unstimulated HL-60 cells reveals a gelatinolytic band of about 72 kDa (corresponding to MMP-2) and a weak band of about 92 kDa, corresponding to pro-MMP-9 (lane a). PMA stimulation causes a significant increase of pro-MMP-9 in the supernatant, along with the appearance of an 84 kDa species, corresponding to active MMP-9, whereas the amount of MMP-2 remains largely unchanged (lane b). In the crude membrane fraction of unstimulated HL-60 cells, no gelatinase activity is detected (lane c), but after PMA stimulation two bands are detected, corresponding to the pro- and active MMP-9 forms (lane d)

Figure 2 Co-immunoprecipitation of MMP-9 with anti-ICAM-1 Ab from the lysates of HL-60 cells. (a) Lysates of unstimulated (lanes a, c) and PMA-stimulated (lanes b, d) HL-60 cells were incubated with anti-ICAM-1 Ab (lanes a and b) or with a control mouse IgG1 (lanes c and d) and the immunoprecipitates were subjected to Western blot analysis using an anti-MMP-9 Ab. Lane e: Concentrated (100-fold) supernatants of PMA-stimulated HL-60 cells immunoblotted with anti-MMP-9 Ab. (b) Lanes a-d, immunoprecipitates in the corresponding lanes in a were subjected to gelatin zymography. Lanes e-g, gelatin zymogram of lysates of unstimulated (lane e) and PMA-stimulated (lane f) HL-60 cells, and supernatants from PMA-stimulated cells (lane g). Molecular weight markers are indicated on the left. The arrows indicate 92 kDa bands that correspond to proMMP-9

Figure 3 Expression of secreted and cell surface-associated MMP-9 in HL-60 cells transfected with anti-sense ICAM-1 cDNA. (a) Gelatin zymogram analysis of culture supernatants of PMA-stimulated HL-60 isolates transfected with vector only (lane a) and with ICAM-1 cDNA in antisense orientation (lanes b, c). (b) Gelatin zymogram analysis of crude membrane extracts from cells corresponding to the supernatants in a. Lanes b and c represent supernatants and lysates from AS32 and AS49 isolates respectively. Equal amounts of protein were loaded onto each lane. Molecular weight markers are indicated on the left

Figure 4 ³H-labeled collagen degradation assay. ³H-labeled type IV human collagen was embedded in Matrigel and used to coat cell culture inserts of double chamber cell culture plates. The lower chamber was filled with 500 l of serum free RPMI containing the chemotactic factor MIP 1-. Wild type HL-60, AS HL-60 cells or the conditioned culture media of wild type HL-60 cells were added to the upper chamber and the radioactivity released into the supernatant was measured. The results are expressed as mean±s.d. of triplicate values. (a) Wild type HL-60 cells show increased ability to degrade the radiolabeled type IV collagen after stimulation with PMA which is partially blocked by an MMP blocking peptide (MMPI) and an anti-MMP-9 antibody that blocks MMP-9 proteolytic activity (anti MMP-9 Ab). An inhibitor with high affinity for MMP-3 (MMP-3I) was less effective. (b) The supernatants of PMA-stimulated HL-60 cells do not increase the degradation of collagen compared to those of unstimulated cells. (c) Two independent isolates of AS-HL-60 (AS32 and AS49) cells show a significant reduction in their ability to degrade the type IV collagen compared to the wild type (vector only-transfected) cells. Variations in maximal counts between a and c are due to variations in matrigel coating efficiency of inserts

Figure 5 Shedding of ICAM-1 from HL-60 cells. (a) 10⁷ cells were incubated with or without 100 nM PMA for 18 h and the supernatant (20 ml) was concentrated 100-fold and analysed by Western blot using an anti-ICAM-1 Ab. Soluble ICAM-1 is observed only after PMA stimulation. (b) HL-60 cells were incubated with PMA 100 nM and different MMP inhibitors and the supernatants were analysed as described above. Cell cultures contained: lanes a: no inhibitor; lanes b: 1-10 phenanthroline 10 M; lanes c: MMP-2/MMP-9 inhibitor III (Calbiochem) 100 M; lanes d: MMP inhibitor 2S,3R-3-amino-2-hydroxy-4-(4-nitrophenyl)butanoyl-L-leucine (Sigma). The experiment was done in duplicate. Molecular weight markers are indicated on the left

Figure 6 Shedding of ICAM-1 in transfected COS cells. COS cells were transfected with ICAM-1 and the crude membrane fraction of the transfected cells was incubated with or without purified, APMA activated MMP-9 in the presence or absence of protease inhibitors. ICAM-1 shedding was assessed by Western blot analysis of the supernatants after centrifugation of the membranes, as described in Materials and methods. Membrane fractions were incubated with: lane a, medium alone; lane b, 10 g/ml MMP-9; lane c, 10 g/ml MMP-9 and a cocktail of protease inhibitors exclusive of any MMP inhibitor; lane d, 10 g/ml MMP-9 and 1-10 phenanthroline 1 mM. Molecular weight markers are indicated on the left

Figure 7 In vitro cleavage of ICAM-1 by MMP-9. COS cells were transfected with v5-tagged sICAM-1 (a) and the supernatants were collected, incubated with purified, APMA activated MMPs and analysed by Western blot with an anti-v5 Ab. Supernatants were treated with: lane a, nothing; lane b, MMP-9 10 g/ml; lane c, MMP-3 10 g/ml; lane d, MMP-1 10 g/ml; lane e, MMP-9+1-10 phenanthroline 1 mM. (b) COS cells were transfected with ICAM-Ig (lanes a-d) or CD5-Ig (lanes e-h), and 72 h following transfection, the serum-free supernatants were collected, incubated with purified, APMA activated MMPs as in a and analysed by Western blot using an anti-human IgG Ab. Samples were treated with: lanes a and e, nothing; lanes b and f, MMP-9 10 g/ml; lanes c and g, MMP-3 10 g/ml; lanes d and h, MMP-1 10 g/ml. Molecular weight markers are shown on the left. Arrows indicate bands of about 110 and 55 kDa that correspond to dimers and monomers (gel electrophoresis was performed under non-reducing conditions) of ICAMRg cleavage products

Figure 8 P404E mutation of ICAM-1 abrogates MMP-9-induced shedding. (a) Schematic representation of the putative cleavage sequence and the P404E mutation. EC, TM and IC denote the ICAM-1 extracellular, transmembrane and intracellular domains, respectively. (b) COS cells were transfected with wild type ICAM-1 (lanes a and d), v5-tagged P404E ICAM-1 mutant (lanes b and e) or v5-tagged wild type ICAM-1 (lanes c and f), and ICAM-1 shedding was assessed by Western blot analysis of culture supernatants 72 h post-transfection, using an anti-ICAM-1 Ab. Without MMP-9 co-transfection, no sICAM was detected in the culture supernatants (lanes a-c). When MMP-9 cDNA was co-transfected with the ICAM-1 cDNAs, sICAM-1 was detected in the culture supernatants of cells transfected with wild type ICAM-1, with and without the v5 tag (lanes d and f). When MMP-9 cDNA was co-transfected with the P404E ICAM-1 mutant cDNA, no shedding of ICAM-1 was detected (lane e). Expression of wt and mutant ICAM-1 was comparable, as assessed by Western blot analysis (data not shown)

Figure 9 Shedding of ICAM-1 from MDA-MB435 cells. (a) Western blot analysis of wild type (vector only-transfected) MDA-MB435 cells (lane a) and cells transfected with v5-tagged MMP-9/CD44 fusion protein (lane b) or v5-tagged soluble MMP-9 (lane c). (b) 10⁷ cells were incubated in serum free medium for 18 h and the supernatant (20 ml) was concentrated 100-fold and analysed by Western blot with an anti-ICAM-1 Ab. The corresponding cells were lysed and equal amounts of lysate-derived protein were subjected to SDS-PAGE, transfer onto nitrocellulose filters and Western blot analysis using anti-ICAM-1 Ab. Supernatants (lanes a-c) and corresponding lysates (lanes e-g) were from: lanes a and e, wt MDA-MB435 cells; lanes b and f, soluble MMP-9 transfectants; lanes c and g, MMP-9/CD44 fusion transfectants. Comparable amounts of ICAM-1 were present in the lysates of wild type and MMP-9 transfected MDA-MB435 cells, but significant ICAM-1 shedding was detected in the supernatants of MMP-9 transfectants only and not in those of wild type cells

Figure 10 Cytotoxicity assay. (a) MMP-9-transfected MDA MB435 cells were incubated with NK-92MI cells adjusted to produce a target: effector ratio (T : E) of 1 : 40. Specific lysis of 65% target cells (b) was observed when the NK-92MI cells were incubated with wild type (vector only-transfected) MDA-MB435 cells. The NK cell line induced 25% lysis of the MDA-MB435 cells transfected with MMP-9 (data not shown) or the MMP-9/CD44 fusion protein. However, inhibition of MMP-9 activity restored transfectant sensitivity to NK-92MI-mediated killing to wt cell levels. (b) NK-92MI-mediated cytotoxicity of vector only-transfected MDA-MB435 wt cells was strongly reduced in the presence of ICAM-1Rg but not CD5Rg. All experiments were done in triplicate and are presented as mean values±s.d.

Table 1 ICAM-1 expression in wild type and AS HL-60 cells