Osteopontin, a Novel Substrate for Matrix Metalloproteinase-3 (Stromelysin-1) and Matrix Metalloproteinase-7 (Matrilysin)*

Osteopontin (OPN) is a secreted phosphoprotein shown to function in wound healing, inflammation, and tumor progression. Expression of OPN is often co-localized with members of the matrix metalloproteinase (MMP) family. We report that OPN is a novel substrate for two MMPs, MMP-3 (stromelysin-1) and MMP-7 (matrilysin). Three cleavage sites were identified for MMP-3 in human OPN, and two of those sites were also cleaved by MMP-7. These include hydrolysis of the human Gly 166 -Leu 167 , Ala 201 -Tyr 202 (MMP-3 only), and Asp 210 -Leu 211 peptide bonds. Only the N-terminal Gly-Leu cleavage site is conserved in rat OPN (Gly 151 -Leu 152 ). These sites are distinct from previously re-ported cleavage sites in OPN for the proteases thrombin or enterokinase. We found evidence for the predicted MMP cleavage fragments of OPN in vitro in tumor cell lines, and in vivo in remodeling tissues such as the postpartum uterus, where OPN and MMPs are co-expressed. Further-more, cleavage of OPN by MMP-3 or MMP-7 potentiated the function of OPN as an adhesive and migratory stimulus in vitro through cell surface integrins. We predict that inter-actionofMMPswithOPNattumorandwoundhealingsites a 9 b 1 integrin clone Y9A2, anti- a 4 integrin clone AV1, and anti- a 5 integrin clone NK1-SAM-1 were obtained from Chemicon. The peptides GRGDSP and GRGESP were ob- tained from Life Technologies, Inc. In some experiments, recombinant murine tumor necrosis factor a (TNF a , R&D Systems) was used at a concentration of 1 ng/ml. Cell Culture— AsPC-1 and HeLa cells were obtained from ATCC and of modified bovine and m Primary chondrocytes as ously and Co-cultures using cell dishes and macrophages M21 cells by Dr. H. cell Enzyme Cleavage Assays and Sequencing— and human OPN were cleaved by MMPs (enzyme/substrate ratio varied from 1:5 to 1:20 to maximize yield of cleavage fragments) and thrombin (0.25 units/ m g of OPN) in equal volume of cleavage buffer (200 m M NaCl, 50 m M Tris-HCl, pH 7.6, 5 m M CaCl 2 ) for 5–15 min at 37 °C. The amount of OPN was kept constant at 200 ng, and the enzyme amount varied according to the enzyme/substrate ratio used. For most biological assays, 10 ng of MMP was used with 200 ng of OPN for cleavage assays. The mixture of cleaved OPN fragments was separated on a 12.5% polyacrylamide gel under reducing conditions and then transferred to a 0.2- m m immunoblot modified surface toxylin.

role during tissue injury and stress. Interestingly, several members of the matrix metalloproteinase (MMP) family are also induced during injury/disease processes in patterns overlapping that of OPN (13). In particular, we have found that during squamous cell carcinoma progression, OPN and MMP-3 expression correspond both in a temporal and cell-specific fashion (9,14). We have also identified overlapping expression patterns of OPN and MMP-3 in the stroma during skin incisional wound healing (5) and OPN and MMP-7 during involution of the postpartum uterus (15).
In the present study, we have identified and characterized novel cleavage sites in OPN for two members of the MMP family, MMP-3 and MMP-7. Furthermore, we show that lower molecular weight forms of OPN corresponding to predicted MMP cleavage fragments are present in cell lines in vitro and in tissues in vivo. Biological assays demonstrate that the MMPcleaved OPN has increased activity in promoting both cell adhesion and migration compared with full-length OPN. In addition, using inhibitory reagents, we have determined that the same receptors that interact with OPN also mediate interaction of MMP-cleaved OPN with tumor cells. These data suggest that active forms of OPN at sites of tissue injury may be regulated by the activity of proteases including MMPs and that the differences in activity of modified OPN may be explained by differences in binding affinity of integrins or distinct downstream signaling events. maintained in RPMI (Life Technologies) with 20% fetal bovine serum (AsPC-1) and minimum essential medium (Life Technologies, Inc.) with 10% fetal bovine serum. Primary mouse macrophages were obtained by peritoneal injection of 3% thioglycollate medium, with collection of exudate after 4 days. Cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and 50 g/ml gentamycin. Primary chondrocytes were prepared from intervertebral discs as previously described (26) and maintained in a 1:1 mixture of Dulbecco's modified Eagle's medium and Ham's F-12 medium with 10% fetal bovine serum, 25 g/ml ascorbic acid, and 50 g/ml gentamycin. Cocultures of macrophages with chondrocytes were performed using cell culture inserts in six-well dishes with chondrocytes in the insert and macrophages in the wells as described (26). Mo and Mo␣ v subclones of M21 melanoma cells were kindly provided by Dr. Mark H. Ginsberg (Scripps Research Institute) and cultured as described. While Mo does not express ␣ v integrins, Mo␣ v expresses high levels of ␣ v integrins at the cell surface (27).
Enzyme Cleavage Assays and Protein Sequencing-Rat and human OPN were cleaved by MMPs (enzyme/substrate ratio varied from 1:5 to 1:20 to maximize yield of cleavage fragments) and thrombin (0.25 units/g of OPN) in equal volume of cleavage buffer (200 mM NaCl, 50 mM Tris-HCl, pH 7.6, 5 mM CaCl 2 ) for 5-15 min at 37°C. The amount of OPN was kept constant at 200 ng, and the enzyme amount varied according to the enzyme/substrate ratio used. For most biological assays, 10 ng of MMP was used with 200 ng of OPN for cleavage assays. The mixture of cleaved OPN fragments was separated on a 12.5% polyacrylamide gel under reducing conditions and then transferred to a 0.2-m immunoblot polyvinylidene difluoride membrane (Bio-Rad). The membrane was washed with 50% methanol and stained with fresh Coomassie Brilliant Blue for less than 1 min and then destained with 50% methanol. Visible protein bands were excised by a razor-sharp blade and allowed to dry for about 1 h at room temperature and then shipped for sequencing. The N-terminal amino-terminal sequence was determined by modified automated Edman degradation (ProSeq Inc., Boxford, MA).
Immunoblotting and in Situ Hybridization-For immunodetection, proteins were transferred to polyvinylidene difluoride membrane as above. For samples of chondrocyte/macrophage co-cultures, 22.5 g of protein were present in each lane. For tumor cells and tissue lysates, 200 g of protein were run per lane. The nonspecific binding was blocked with 10% nonfat dry milk in TBS-T buffer (10 mM Tris base, pH 8.0, 150 mM NaCl, and 0.1% Tween 20) at room temperature for 1 h. The membrane was incubated in a combination of three primary anti-OPN antibodies, OPN199 goat IgG (1:1000) and LP209 and LP210 rabbit IgG (1:2000), in 10% milk TBS-T for 1 h and then in horseradish peroxidase goat anti-rabbit and horseradish peroxidase rabbit anti-goat (1:2000) for another 1 h. The OPN bands were visualized using chemiluminescent reagent containing 250 mM 3-aminophthalydrazide and 90 mM p-coumaric acid. In situ hybridization on tissue sections was performed as previously described (9) using sense and antisense riboprobes to the mouse OPN cDNA 2ar (28).
Functional Assays-Adhesion and migration assays were performed as previously described (25). Briefly, for adhesion assays, test substrates were coated onto wells of Maxisorp 96-well plates and incubated overnight at 4°C. After blocking with 10 mg/ml bovine serum albumin/ Dulbecco's modified Eagle's medium for 1 h at 37°C, wells were rinsed, and detached cells were plated for 1 h at 37°C (AsPC-1 and HeLa cells plated at 30,000 cells/well and melanoma cell strains plated at 80,000 cells/well). Nonadherent cells were washed off, and attached cells were fixed with 4% paraformaldehyde, stained, and quantitated. Migration assays were performed using test substrates in the lower chamber of a modified Boyden chamber apparatus, separated from cells with a polycarbonate filter with 8-m pores. After the migration period, cells on the upper surface of the filter (cells that did not migrate) were scraped off, and cells that had migrated through the pores to the lower surface of the membrane were fixed with 100% MeOH and stained with hematoxylin. Migrated cells were quantitated by cell counts of three random fields/well for six wells per test condition. For biological assays with protease cleaved OPN, cleavage of the substrate was verified by Western blot in all instances with an aliquot of the cleaved material before it was used for adhesion or migration assays. Inhibitory anti-integrin antibodies were used at a concentration of 25 g/ml, and peptides were used at a concentration of 200 g/ml. These reagents were incubated with cells for 15 min at 37°C before cells were plated in adhesion assays.

RESULTS
We have previously characterized a model of herniated disc resorption, allowing the study of the interaction of inflammatory cells with chondrocytes (26). In this system, both MMP-3 and MMP-7 are activated (26). MMP-3 is necessary for the generation of a macrophage chemoattractant (26), while MMP-7 cleaves TNF␣ (29), and both events contribute to disc resorption. Because of the prevalence of OPN during inflammation, we were interested in possible interactions of these MMPs with OPN in this system. When we examined chondrocyte cultures in the presence or absence of macrophages, we found high levels of OPN in chondrocytes alone, but additional lower molecular weight forms detectable following co-culture with macrophages (Fig. 1A). Since the co-culture system has been previously shown to induce the production of MMP-3 (26), we tested whether MMP-3 might be responsible for the presence of the lower molecular weight forms of OPN. Using purified native rat OPN (25) and active MMP-3, cleavage assays were performed, and similar OPN bands at apparent molecular masses of 40 and 32 kDa were seen as a result of this cleavage (Fig. 1B). In order to confirm the direct effect of MMP-3 in the co-cultures, two approaches were taken. Chondrocytes were cultured alone or in the presence of TNF␣, a cytokine known to stimulate MMP-3 production (29). In the absence of cytokines or macrophages, only full-length OPN was observed (Fig. 1C,  lane 2), while a predominant 40-kDa band was additionally found in chondrocyte cultures treated with TNF␣ (lane 3). Furthermore, we utilized macrophages derived from wild type or MMP-3 null animals (30) to confirm that MMP-3 was necessary for the generation of OPN fragments in the co-culture system. As shown in Fig. 1C, chondrocytes co-cultured with macrophages from MMP-3 null (Ϫ/Ϫ) animals produced only the full-length OPN (lane 4; compare with chondrocytes cultured alone, (Fig. 1A)), while co-cultures of chondrocytes and macrophages from wild type animals generated OPN cleavage fragments (lane 5). These data show that purified MMP-3 can cleave native OPN and that in the co-culture system, cleavage of endogenously produced OPN is dependent on the presence of endogenous MMP-3.
To further characterize the proteolytic cleavage of OPN by MMPs, we studied recombinant OPN and active MMP-3 and MMP-7. We used recombinant human and rat OPN and found that both were substrates for MMP cleavage, although the cleavage pattern was slightly different (Fig. 2). Similar to our results using native rat OPN (Fig. 1A), recombinant rat OPN was cleaved by MMP-3 to generate two fragments at apparent molecular masses of 40 and 32 kDa ( Fig. 2A). However, the pattern of cleavage of recombinant human OPN (huOPN) differed (Fig. 2B), and in addition to 40-and 32-kDa bands, MMP-3-cleaved huOPN additionally generated a 25-kDa band. Following MMP-7 cleavage of huOPN, 25-, 20-, and 15-kDa bands were generated. All MMP-cleaved OPN fragments were distinct from those generated by thrombin cleavage (Fig. 2C), which generated 30-and 28-kDa bands in both rat and human recombinant OPN. Titration of enzyme/substrate ratios and cleavage times showed that with limiting amounts of enzyme and/or shorter cleavage times with MMP-7, huOPN also generated the 40-and 32-kDa bands, suggesting that these are intermediate forms in the reaction (Fig. 2D). All cleavage products of both MMP-3 and MMP-7-cleaved rat and huOPN were analyzed by protein sequencing for determination of cleavage sites. The determined cleavage recognition sites are shown in Fig. 3. The predominant cleavage site found for both MMP-3 and MMP-7 in rat and huOPN was a Gly-Leu bond (Gly 166 -Leu 167 in huOPN, Gly 151 -Leu 152 in rat OPN) just five amino acid residues C-terminal to the RGD sequence. Additionally, both MMP-3 and MMP-7 were found to cleave huOPN at the Asp 210 -Leu 211 bond, but only MMP-3 displayed a minor cleavage site at Ala 201 -Tyr 202 in huOPN. These cleavage sites corresponded with the apparent size of the fragments on SDSpolyacrylamide gel electrophoresis. A schematic showing cleavage patterns and OPN fragments generated is depicted in Fig.  4. At the present time, the only unidentified fragments that we expect should be generated are the small molecular weight fragments derived from MMP-3 cleavage of the 32-kDa band giving rise to the 25-kDa band and the MMP-7 cleavage of the 25-kDa band to generate the 20-kDa band (Fig. 4, question   FIG. 3. Identified MMP-3 and MMP-7 cleavage sites in rat and human OPN. Amino acid sequences of human and rat OPN are aligned, with cleavage sites for MMP-3, MMP-7, and thrombin indicated. Underlined sequences indicate ␣ 9 ␤ 1 and ␣ 4 ␤ 1 recognition sites revealed following thrombin cleavage of OPN. The determined N-terminal sequence of each fragment is shown with the apparent molecular weight and corresponding fragments a-e as shown in Fig. 2. marks). These predicted fragments have not been detected even in high percentage gels, suggesting that they may be further degraded.
To determine if other cell lines in addition to the chondrocyte and macrophage co-culture generated cleavage fragments of OPN, we screened a variety of tumor cell lines, which are known to have high expression of MMPs (31)(32)(33)(34). In addition, CHO cells that were either stably transfected with MMP-3 or vector alone were compared. Both conditioned medium and cell lysates were analyzed, and although OPN was detected in both compartments, there was consistently more protein detectable in the cell lysates. This was probably due to the fact that secreted OPN binds to cell surface proteoglycans and is found in the extracellular space (5), sometimes bound to other extracellular matrix proteins (35). Thus, a large proportion of secreted OPN remained associated with the cell layer and extracellular matrix. The observation that a member of the MMP family, MMP-2, may associate with cell surface receptors (36) indicates a potential interaction of MMP with substrates at the cell surface or in the extracellular milieu. As shown in Fig. 5A, vector-transfected CHO cells did not produce detectable amounts of OPN protein by immunoblotting (lane 2). However, in CHO cells stably expressing MMP-3, OPN was detected in three forms, a faint species corresponding to full-length and two lower molecular weight forms running at 40 and 32 kDa (lanes 3 and 4). In addition, in the tumor lines DU-145, a human prostate cancer cell line, and AsPC-1, a human pancreatic cancer line, multiple forms of OPN were detected, including bands corresponding to fulllength OPN, as well as species with apparent molecular masses of 40, 32, 25, and 15 kDa (lanes 5-8). In vivo, the epithelial cells of the mouse postpartum uterus have been shown to be a region of high MMP-7 expression (15), in a pattern overlapping that of OPN expression (Fig. 5, B and C). MMP-7 and OPN expression overlap precisely in the epithelial cell layer of the remodeling uterus. Tissue extracts from 24-h mouse postpartum uterus (PPU) also showed strong lower molecular weight OPN bands in addition to the fulllength protein (Fig. 5A).
OPN is a well characterized adhesive protein and migratory stimulus. To test the regulation of bioactivity of OPN by MMP cleavage, adhesion and migration assays were performed using either recombinant full-length protein or OPN cleaved by MMP-3, MMP-7, or thrombin as a comparison. Full-length OPN was cleaved with MMPs or thrombin and used in adhesion assays in comparison with the same concentrations of OPN treated similarly but in the absence of enzyme. As a control, reactions were prepared with enzyme but in the absence of OPN (Fig. 6A). HPLC-purified 40-kDa fragment was also used and gave similar results as the unpurified MMP-cleaved OPN. We found that the cleavage of OPN (either human or rat OPN gave similar results) with MMP-3, MMP-7, the catalytic domain of MMP-3 (cat), or thrombin significantly increased adhesion of tumor cells in comparison with full-length OPN alone. Interestingly, the sensitivity of the cells to protease-cleaved OPN was greatly enhanced compared with full-length OPN, as seen in dose comparison experiments (Fig. 6A). Using 200 ng of OPN as a substrate, cleavage by MMP-3 and MMP-7 increased adhesion by ϳ2-fold. However, with 50 ng of OPN as a substrate, MMP-3 and MMP-7 cleavage enhanced adhesion by 10-and 18-fold, respectively. Several cell lines were screened and found to display enhanced adhesion to MMP-cleaved OPN, including A5 and B9 murine squamous cell carcinoma lines, rat smooth muscle cells, human aortic endothelial cells, and NIH3T3 fibroblasts (p Ͻ 0.01 for all). Similarly, migration was tested using full-length OPN or MMP-3-cleaved OPN as a stimulus, and we found that macrophage migration toward MMP-3-cleaved OPN was significantly enhanced compared with full-length OPN (Fig. 6B). Again, several cell types showed the same increased migratory response to MMP-3cleaved OPN compared with full-length OPN, including A5 cells and B9 cells (p Ͻ 0.01), rat smooth muscle cells, human aortic endothelial cells, and NIH3T3 fibroblasts. These data show that the bioactivity of OPN toward cells can be regulated by proteolysis by MMPs.
Although in many instances, the adhesive and migratory activity of MMP-cleaved OPN appeared similar to thrombincleaved OPN (Fig. 6A), we found situations in which the functional consequences of these proteolytic modifications were rather distinct. For example, HeLa carcinoma cell lines displayed a poor adhesion to full-length OPN and significantly higher, but not impressive, adhesion to MMP-cleaved OPN (Fig. 6C). In contrast, high levels of adhesion were seen on a thrombin-cleaved OPN substrate. These findings imply that the activities of thrombin-cleaved versus MMP-cleaved OPN are distinct, and this may be related to receptor specificity on individual cell lines.
One possibility explaining the enhanced bioactivity of MMPcleaved OPN in comparison with full-length OPN is that different and additional receptors are activated. Indeed, this is the case with thrombin-cleaved OPN, where an ␣ 9 ␤ 1 and ␣ 4 ␤ 1 binding site is revealed after cleavage (19). Since most OPN interaction with cells is integrin-mediated, we tested if known integrins were active in mediating cell responses to MMPcleaved OPN.
Using GRGDSP or GRGESP peptides, we found that interaction of the RGD sequence of either full-length OPN or MMP-cleaved OPN could account for virtually all of the adhesion of AsPC-1 tumor cells, since the GRGDSP peptide specifically inhibited the majority of adhesion in each instance (Fig. 7A). Since RGD-mediated adhesion is through both ␣ v -containing and non-␣ v integrins, we tested the adhesion of a substrain of human melanoma cells known to be deficient in ␣ v integrins, Mo, and a corresponding clone with high ␣ v integrin levels, Mo␣ v (27). As shown in Fig. 7B, we consistently found very poor adhesion of ␣ v -deficient melanoma cells (Mo) to any form of OPN. However, the high ␣ v -expressing strain (Mo␣ v ) displayed high levels of adhesion to either MMP-3 or MMP-7-cleaved OPN at low coating concentrations (50 ng). Similar to our previous findings in AsPC-1 cells (Fig. 6A), the ability of the Mo␣ v cells to adhere to lower concentrations of MMP-cleaved OPN was enhanced compared with the same concentration of full-length OPN. Although a small amount of Mo cell adhesion was seen to MMP-cleaved OPN, we found that the variability between experiments gave inconsistent results when we tried to inhibit the small amount of adhesion with specific antibodies.
Using the AsPC-1 tumor cells, we further addressed the requirements of the ␣ v integrin in cell adhesion to OPN forms (Fig. 7C). Adhesion of cells to full-length OPN was significantly inhibited in the presence of anti-␣ v ␤ 3 and to a lesser extent with antibodies against ␣ v ␤ 5 and ␣ 9 ␤ 1 . This pattern was the same with MMP-3-and MMP-7-cleaved OPN, where anti-␣ v ␤ 3 inhibited ϳ75% adhesion, and anti-␣ v ␤ 5 and anti-␣ 9 ␤ 1 inhibited 20 -40% adhesion. Antibodies specifically blocking the ␣ 4 integrin had no effect on the adhesion of ␣ 4 -containing tumor cells under any condition tested (data not shown). DISCUSSION We have made the novel observation that OPN is a substrate for cleavage by two MMPs, MMP-3 and MMP-7. OPN has previously been shown to be proteolytically modified by two other enzymes, thrombin (17) and enterokinase (18). 2 Interestingly, a major cleavage site for both MMP-3 and -7 resides just 2 amino acid residues N-terminal to the thrombin cleavage site. Cleavage by either of these MMPs generates an N-terminal fragment containing the RGD sequence. Of interest, this N-terminal fragment also contains a truncated version (SVVYG) of a motif (SVVYGLR) that has been shown to be recognized by both ␣ 4 ␤ 1 (20) and ␣ 9 ␤ 1 (19) in thrombin-2 P.-L. Chang and C. Prince, personal communication. cleaved OPN. Structural studies have suggested that the C-terminal LR residues are necessary for recognition of both ␣ 9 ␤ 1 (16) and ␣ 4 ␤ 1 3 through the SVVYGLR motif. Consistent with this, we found that antibodies that block the ␣ 4 integrin did not affect binding of ␣ 4 -containing tumor cells to MMPcleaved OPN, suggesting that neither the SVVYG sequence nor the second identified ␣ 4 ␤ 1 (20) site are active in adhesion to the MMP-cleaved OPN. On the other hand, we found a small but consistent inhibition of cell adhesion to all forms of OPN with a blocking anti-␣ 9 ␤ 1 antibody. Although this was an unexpected result for adhesion to the full-length OPN, it is important to note that the anti-␣ 9 ␤ 1 antibody did not additionally inhibit cell binding to MMP-cleaved OPN. These results suggest that the cleavage of OPN by MMPs to generate the SVVYG sequence does not reveal an additional ␣ 9 ␤ 1 binding site. However, it is still unclear why some inhibition of cell adhesion to full-length OPN occurred in the presence of this antibody, but it may involve the activation state of the integrin and functional modulation by the presence of other cell surface integrins. In addition to the N-terminal 40-kDa fragment containing several integrin binding motifs, four other fragments are generated by MMP cleavage, which contain no known consensus receptor binding sequences. We are currently investigating whether purified fragments may be active in promoting cell adhesion and migration or acting in an inhibitory manner (22).
The biological significance of proteolytic modification of OPN has been suggested in the case of thrombin cleavage, where in vitro studies have established regulation both at the level of receptor recognition (16,19) as well as adhesive (21) and migratory (23) activity of OPN. These findings are particularly relevant in vivo, where OPN is specifically and selectively expressed during tissue injury, inflammation, wound repair, and tumorigenesis (37,38). It is also known that MMP expression and activation is coincident in many of these processes. For example, we have found that MMP-3 and OPN expression correspond temporally and spatially during tumorigenesis of the skin (9,14) and in the stroma of skin during wound healing (9). In addition, inflammatory cells express high levels of MMPs and OPN, and remodeling tissues such as the postpartum uterus have high levels of both MMP-7 and OPN (15). We therefore predict that in vivo, is it likely that OPN may be found in MMP-modified forms in addition to the full-length protein. During tissue repair and tumor growth, the presence of modified OPN may be biologically significant, since we have found that the activity of OPN changes following MMP modification. In particular, we find that following MMP cleavage, OPN activity as an adhesion molecule and migratory stimulus 3 S. T. Barry, personal communication. is significantly increased in a variety of cell types. Our initial analysis of receptors involved in cell interaction suggests that the same integrins that bind full-length OPN also bind to cleaved OPN fragments, at least in the tumor cells tested. Therefore, we hypothesize that rather than changing receptor specificity, alternate signaling mechanisms or receptor sensitivity may be involved in the potentiation of response. We are currently focused on addressing the individual contribution of each fragment generated by MMP cleavage and the receptor binding capability of the fragments. We also cannot eliminate the possibility that some of the cleaved fragments might be functionally inhibitory (22).
The concept that proteases such as MMPs function only to degrade or inactivate matrix proteins has been challenged, and it is now clear that proteolysis can modify or even increase functions of a given protein (39). Classes of molecules whose activities are regulated by MMP proteolysis include cytokines, growth factors and growth factor receptors, and extracellular matrix proteins. Interleukin-1␤ (40), insulin growth factorbinding proteins (41), heparin-binding EGF-like growth factor (42), TNF␣ (29), fibroblast growth factor receptor 1 (43), and Fas-L (44) are among the cytokines/growth factor pathways affected. In some cases, MMPs release membrane-tethered forms of proteins to increase activity of soluble forms. For example, MMP-7 releases TNF␣ and Fas-L from the cell surface, after which they modulate macrophage migration and cell apoptosis, respectively. Proteolytic modification of the extracellular matrix components laminin-5, decorin, entactin, and fibronectin by MMPs alter the ability to stimulate cell migration, growth factor sequestration, cell apoptosis, and proliferation (45)(46)(47)(48). Transforming growth factor-␤ binds to the proteoglycan decorin in the extracellular matrix, and this may be a reservoir for the cytokine. MMP-2, -3, and -7 cleave decorin, thus releasing transforming growth factor-␤ from the complex. Our observations of the changes in OPN bioactivity following MMP cleavage are consistent with MMP roles as regulators of extracellular protein activity. Since many MMP family members often have overlapping substrate specificity, it is of interest to determine if other members of the MMP family can proteolytically modify OPN.