The Matrix Metalloproteinase 9 (MMP-9) Hemopexin Domain Is a Novel Gelatin Binding Domain and Acts as an Antagonist*

Matrix metalloproteinases (MMPs) are involved in the remodeling processes of the extracellular matrix and the basement membrane. Most MMPs are composed of a regulatory, a catalytic, and a hemopexin subunit. In many tumors the expression of MMP-9 correlates with local tumor growth, invasion, and metastasis. To ana-lyze the role of the hemopexin domain in these processes, the MMP-9 hemopexin domain (MMP-9-PEX) was expressed as a glutathione S -transferase fusion protein in Escherichia coli . After proteolytic cleavage, the iso-lated PEX domain was purified by size exclusion chromatography. In a zymography assay, MMP-9-PEX was able to inhibit MMP-9 activity. The association and dissociation rates for the interaction of MMP-9-PEX with gelatin were determined by plasmon resonance. From the measured rate constants, the dissociation constant was calculated to be Construction of Murine MMP-9-PEX-GST Fusion Proteins— GST constructed using primers on

Matrix metalloproteinases (MMPs) 1 are a family of zinc metallo-endopeptidases secreted by cells. They are responsible for most of the turnover of matrix components. The MMPs are produced as zymogens with a signal sequence and propeptide segment that has to be removed during activation. The propeptide domain contains a conserved cysteine that chelates the zinc in the active site. The gelatinases MMP-2 and MMP-9 contain fibronectin type II domains that are inserted in the middle of the catalytic domain, presumably to enhance substrate binding (1). MMP-9 also has a collagen type V-like domain located between the catalytic and the C-terminal hemopexin domain (Fig. 1). All but two MMPs (MMP-7 and MMP-26) contain a regulatory subunit, the hemopexin domain, separated from the catalytic domain by a variable hinge region (2). This domain is thought to confer much of the substrate specificity to the MMPs (3). It is involved in activation as well as inhibition of MMPs (3,4) and may enhance substrate binding and specificity (5). The hinge region also confers specificity to the MMPs either by direct binding of the substrate or by setting the orientation of the hemopexin domain and the catalytic domain (6). The hemopexin domain of MMP-2 is known to bind heparin (7). Heparin has been shown to potentiate the activities of some MMPs, and MMPs are often found associated with heparin sulfate glycosaminoglycans on the cell surface (8). The overall three-dimensional structure of the hemopexin domain is a four-bladed propeller with a calcium binding site nestled in the folds (3). A fragment of MMP-2, which comprises the C-terminal hemopexin domain, has been shown to inhibit MMP-2 activity by preventing enzyme binding to ␣ v ␤ 3 integrin and blocks cell surface collagenolytic activity (9).
Since its discovery in 1974 (10), MMP-9 has been implicated in the degradation of the extracellular matrix in a variety of physiological and pathological processes. MMP-9 is important for the migration of different cell types (e.g. leukocytes and cancer cells) because of its ability to degrade basement membranes and components of the extracellular matrix such as collagens, elastin, and aggrecan. MMP-9 is present in large amounts in the granules of neutrophils and plays an important role in inflammation diseases (11). Tumor cells themselves induce MMP-9 production in the neighboring cells and use MMP-9 for their migration when becoming invasive and metastatic (12,13). In addition, MMP-9 is important for the formation of new blood vessels essential for tumor growth and triggers the angiogenic switch during carcinogenesis (14). MMP-9 is also responsible for the processing of cytokines, e.g. pro-interleukin-1␤ and pro-tumor necrosis factor-␣ into their active forms (15,16). In hepatocellular carcinomas the expression of MMP-9 is correlated with local tumor growth, invasion, and intrahepatic metastases (17).
Furthermore, MMPs participate in cell migration (18) and may be expressed on the surface of cells, thus allowing for localized proteolysis (19,20). MMP-9 is necessary for the migration of Langerhans cells, on which it is expressed on the cell surface, and can be inhibited by a broad spectrum inhibitor of MMPs (21).
In the present work we demonstrate that MMP-9 activity (in gelatin zymography) can be inhibited by a recombinant MMP-9 hemopexin (PEX) domain. Because this construct lacks the known gelatin binding region of MMP-9, i.e. the fibronectin type II domain T2HU-2 within the catalytic domain, we assumed a second gelatin binding domain to be present within the MMP-9 hemopexin domain. We demonstrate a high affinity binding between MMP-9-PEX and gelatin. Furthermore, we show that MMP-9-PEX inhibits the migration of MMP-9-expressing melanoma cells.
Construction of Murine MMP-9-PEX-GST Fusion Proteins-GST fusion proteins were constructed using primers based on the published murine MMP-9 sequence (23). An antisense primer specific for the C-terminal nucleotides 1558 -2187 was engineered to contain an internal XbaI site (5Ј-GTGGTCTCTAGAGACTTGCACTGCACGG-3Ј). A sense primer was engineered with an internal HindIII site (5Ј-ACTGC-GAAGCTTATGGAGGCCTCTACAGAG-3Ј). A full-length MMP-9 cDNA was a generous gift from S. Masure (Leuven, Belgium) and was used as a PCR-template. Products were digested and ligated into pGEX-5X-1 plasmid (Amersham Biosciences).
Expression and Purification-The GST-MMP-9-PEX fusion protein was expressed in bacteria (BL21). Isopropyl-1-thio-␤-D-galactopyranoside (IPTG)-induced log phase bacterial cultures containing these constructs were shock frozen and sonicated several times. After lysis, the insoluble parts were removed by centrifugation. GST fusion proteins were purified on Sepharose 4B-coupled glutathione beads (Amersham Biosciences). Beads were washed extensively with PBS, immobilized fusion proteins were digested with 5 mM factor Xa in 50 mM Tris/HCl, pH8, and the cleavage product was eluted with PBS according to the supplier's instructions. The amino acids GIPETKKLM at the neo-N terminus after digestion resulted from cloning procedures and the factor Xa recognition site. Fractions were pooled, concentrated, and subjected to size exclusion chromatography, which was performed with an equilibrated Superdex 75 (16/60) column (Amersham Biosciences) at a constant flow rate of 1 ml/min. The column was calibrated using a mixture of four proteins of known molecular mass, i.e. bovine serum albumin (69 kDa), ovalbumin (46 kDa), myoglobin (17 kDa), and aprotinin (6 kDa). The column was equilibrated with PBS and loaded with 1.0 ml of protein solution. 3-ml fractions were collected and analyzed by SDS-PAGE.
Western Blot-For antibody-based detection, buffer samples were separated by SDS-polyacrylamide gel electrophoresis utilizing separating gels of 10% polyacrylamide and stacking gels of 3% polyacrylamide. Lanes were loaded with 2 g of total protein each. Following electrophoresis at 30 V, the proteins were transferred to nitrocellulose membrane. Membrane transfer was monitored by staining with Ponceau S. Blots were blocked with TBS-N (pH 7.6) containing 10% BSA, 20 mM Tris, 137 mM NaCl, and 0.1% Nonidet P40, and were washed and incubated with antibodies against the C terminus of MMP-9 or GST (dilution 1:1000). Signals were visualized by enhanced chemiluminescence according to the manufacturer's instructions (Amersham Biosciences). Experiments were done in triplicate.
Circular Dichroism Spectroscopy-Circular dichroism (CD) measurements were carried out on an AVIV (Lakewood, NJ) 62DS CD spectrometer equipped with a temperature control unit calibrated with a 0.1% aqueous solution of D-10-camphersulfonic acid according to Chen and Yang (24). The spectral bandwidth was 1.5 nm. The time constant ranged between 1 and 4 s, and the cell path length was between 0.1 and 10 mm.
Calculation of Protein Concentrations-Protein concentrations were calculated from absorption spectra in the range of 240 -320 nm using the method of Waxman et al. (25).
Mass Spectrometry-The purity and correct molecular mass of the protein were proven by ESI and MALDI-TOF MS. ϳ80 pmol of MMP-9-PEX was injected into an LC-MS-coupled electrospray single quadruple mass spectrometer (Ettan ESI ToF, Amersham Biosciences). The protein was first desalted on a MicroTrap column (Michrom Bio-Resources) before spraying into the ionization chamber. For determining the mass by MALDI-TOF (Ettan Maldi ToF Pro, Amersham Biosciences), ϳ3 pmol of MMP-9-PEX was crystallized in an excess of 3,5-dimethoxy-4-hydroxycinnamic acid. Deconvolution of mass spectra was performed using MagTran software (www.ionsource.com/links/ programs.htm).
Zymography-MMP activity was assessed by gelatin zymography following the methods described previously (26,27). Lanes were loaded with a definite amount of recombinant protein or serum-free supernatant as indicated in Figs. 3 and 5. Samples were preincubated with 4-aminophenyl mercuric acetate (APMA) at a final concentration of 2 mM for 2 h at 37°C. Proteins were run on non-reducing SDS/polyacrylamide gel (10%) containing 1 mg/ml gelatin either with or without 0.41 g of recombinant hemopexin domain (or albumin or peptide GIPET-KKLM as indicated in the Fig. 3 legend) per milliliter of gel. After electrophoresis in 25 mM Tris, 250 mM glycine, and 1% SDS, the gel was washed at room temperature with 2.5% Triton X-100, 5 mM CaCl 2 , 50 mM Tris/HCL, pH 7.5, and incubated again in the same buffer twice for 1 h. After rinsing the gel extensively with six changes of distilled water, it was incubated overnight at 37°C in 5 mM CaCl 2 , 50 mM Tris/HCL, pH 7.5, followed by Coomassie Blue staining (0.5% w/v) and destaining in methanol/acetic acid/water (10:10:80). Gelatin zymography depicts MMPs as negatively staining bands of gelatinolytic activity. Zymographic bands were scanned, and the optical density was determined using the Bio-Rad Gel-Doc system. The activity data of all samples analyzed by zymography were normalized by setting the activity of a defined sample to 1.0 as indicated in the Fig. 3 legend.
Surface Plasmon Resonance Studies-Gelatin was covalently immobilized to a carboxymethyl dextran matrix (Fisons, Loughborough, UK) at 10 g/ml for 2 min in 10 mM sodium acetate buffer, pH 3.9, as recommended by the manufacturer. Binding experiments were performed at controlled temperature (15°C) with 12 different concentrations of the purified hemopexin domain using the IASYSTEM (Fisons, Loughborough, UK) optical biosensor. Association was monitored for at least 2 min, the sample was replaced by 10 mM sodium acetate buffer, pH 3.9, and dissociation was monitored accordingly before the cuvette was regenerated with phosphate buffer/1 M sodium chloride, pH 7.4, and equilibrated again with 10 mM acetate buffer, pH 3.9. Association and dissociation affinograms were analyzed by nonlinear regression with the FASTfit software (Fisons, Loughborough, UK), which uses the Marquardt-Levenburg algorithm for iterative data fitting.
Cells and Cell Treatment-Human malignant melanoma A375 cells were obtained from ATCC (CRL-1619), and rat mesangial cells were a generous gift from P. Mertens (RWTH, Aachen, Germany). Cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 5% (v/v) fetal-calf serum (FCS), 100 mg/liter streptomycin, and 60 mg/liter penicillin. Cells were grown at 37°C in a water-saturated atmosphere in air/CO 2 (19:1). Cells were assessed for the expression of MMP-2 and MMP-9 by zymography after culturing under serumfree conditions for 24 h.
Boyden Chamber Experiments-Transwell 0.64-cm 2 filter inserts (8-m pores) coated with a thin layer of Matrigel basement membrane matrix (125 g/cm 2 ) were rehydrated according to the customer's instructions either with or without recombinant MMP-9-PEX (1-500 g/ ml) and placed in wells of 24-well plates containing 0.5 ml of serum-free DMEM. Prior to seeding into the Transwell inserts, cells were released using trypsin-EDTA and sequentially rinsed with DMEM containing 10% FCS and serum-free DMEM. The rinsed cells were resuspended in DMEM (5.0 ϫ 10 4 cells/ml) and 250 l were added to the upper chambers. DMEM containing 5% FCS was used as chemoattractant. Chambers were incubated for 22-24 h at 37°C, 5% CO 2 in a humidified tissue culture incubator. The cells on the upper surface of the filters were then removed using a cotton swab, and those remaining on the lower surface of the filter were fixed with 10% methanol and stained with hematoxylin and eosin.
The filters were rinsed with deionized water, dried, and examined using light microscopy. The number of cells in five random optical fields (400ϫ magnification) from triplicate filters were averaged to obtain the number of migrating cells.

Molecular
Cloning, Purification, and Characterization of MMP-9-PEX-To obtain a soluble MMP-9 PEX domain we generated a GST fusion protein. The cDNA encoding the PEX domain (MMP-9 amino acids Glu-20 to Pro-729) was inserted into pGEX downstream of the sequence encoding GST and a recognition site for factor Xa (Fig. 1).
The GST fusion protein was expressed in Escherichia coli. Aliquots of bacterial lysates were subjected to SDS-PAGE. In Western blots the protein was detected with two different polyclonal antisera against the C terminus of murine MMP-9 ( Fig.  2A, right, and not shown) or GST ( Fig. 2A, left). A protein band representing the GST-MMP-9-PEX fusion protein was detectable with both antisera and corresponds to a molecular mass of about 52 kDa. Bacterial lysates were then applied to a glutathione-Sepharose column. After digestion with factor Xa, a protein of 25 kDa could be detected using the MMP-9 antiserum ( Fig. 2A), and a protein of 27 kDa was detectable with a polyclonal antiserum against GST (open arrow).
The recombinant protein was further purified by size exclusion chromatography (Fig. 2B). The three fractions containing detectable amounts of proteins (fractions 9, 20, and 34) were analyzed by SDS-PAGE and Western blotting using a sheep polyclonal antiserum directed against the C-terminal part of MMP-9. Only fraction 34 contained considerable amounts of a protein with a molecular mass of 25 kDa (Fig. 2C).
We then identified and characterized MMP-9-PEX by mass spectrometry and circular dichroism, respectively. The purified protein was automatically desalted and analyzed on an Ettan LC-MS (Fig. 2D). The MS run presented in the lower panel of Fig. 2D corresponds to the main peak in the ion chromatogram (not shown). The determined experimental molecular mass for PEX is 25,034 Da (MALDI and ESI MS) and exhibits an offset of 42.58 Da compared with theoretical mass (24,991.42 Da, M ϩ H ϩ ), which could be in line with an additional acetyl group (ϩ 42.03 Da). The CD spectrum of the purified recombinant protein in the far UV is indicative of a protein in a folded state (Fig.  2E). Secondary structure analysis (28) of the far UV CD spectrum reflects the ␤-sheet character of the protein (␣-helix ϭ 21%, ␤-sheet ϭ 35%, turn ϭ 17%), which is typical for a hemopexin domain.
Inhibition of MMP-9 Activity by MMP-9-PEX-MMP-9-PEX itself did not have any gelatinolytic activity in the zymography assay or the MMP-9 ELISA (data not shown). Next, we examined whether MMP-9-PEX had any influence on the enzymatic activity of MMP-9 toward gelatin. Gelatin zymography was used to determine MMP-9 activity in a semiquantitative assay. Zymography bands (Fig. 3A) were quantified by densitometry (Fig. 3B). The activity of 0.2 ng of recombinant murine MMP-9 was given the arbitrary value of 1 in standard gelatin zymography (Fig. 3B). In the presence of 0.41 g of MMP-9-PEX per milliliter of gelatin gel, the activity of MMP-9 (0.2 ng and 1 ng) was reduced by 71 and 61%, respectively (Fig. 3, A, second  panel from the left, and B). As negative controls, albumin and the peptide GIPETKKLM, respectively, were polymerized into the gel instead of MMP-9-PEX, and no effect on the MMP-9 activity was observed (Fig. 3A, far right panel and second panel  from the right). No effect of MMP-9-PEX could be detected on the activity of MMP-2 by zymography (data not shown).
Binding of MMP-9-PEX to Gelatin-To specifically address the ability of MMP-9-PEX to directly bind gelatin, we performed surface plasmon resonance studies. Gelatin was immobilized to a carboxymethyl dextran matrix as described under "Experimental Procedures." Binding experiments were performed at a controlled temperature with 12 different concentrations of purified hemopexin protein using the IASYSTEM optical biosensor. Association and dissociation affinograms (Fig. 4) were analyzed by nonlinear regression with FASTfit software (Marquardt-Levenburg algorithm for iterative data). From these data, an association rate of k on ϭ 3.83 ϫ 10 5 M Ϫ1 s Ϫ1 and a dissociation rate of k off ϭ 1.01 ϫ 10 Ϫ2 s Ϫ1 could be deduced. With these values, the dissociation constant of K d ϭ 2,64 ϫ 10 Ϫ8 M was calculated (Fig. 4, and Table I). Evidently the MMP-9-PEX domain is able to bind to gelatin.
Invasion of MMP-9-expressing Cells in the Presence of MMP-9-PEX-These results prompted us to examine whether the recombinant MMP-9-PEX is able to interfere with the MMPdependent invasion of mammalian tumor cells. Human malignant melanoma A375 cells expressing MMP-9 and MMP-2 and rat mesangial cells expressing MMP-2 as demonstrated by gelatin zymography (Fig. 5A) were cultured on Matrigel-coated membranes in Boyden chambers under serum-free conditions. FCS-containing medium was used as a chemoattractant. Half of the chambers were preincubated with recombinant MMP-9-PEX domain in a concentration of 1, 10, 100, and 500 g/ml (0.04 -20 M) media as indicated in Fig. 5B. A375 incubated with MMP-9-PEX showed decreased invasion and migration through the coated filters compared with control cells. Incubation with GIPETKKLM at 20 M as an additional control did not alter the migration pattern of A375 melanoma cells (Fig.  5B). Rat mesangial cells, expressing MMP-2 exclusively, however, showed no difference in migration in the presence of MMP-9-PEX (data not shown). Thus, with respect to gelatinases, recombinant MMP-9-PEX inhibited migration of predominantly MMP-9 expressing cells but not the migration of exclusively MMP-2 expressing cells. DISCUSSION Reported here is the gelatin binding property of the C-terminal hemopexin domain of murine gelatinase B, the first such report for any of the MMPs. In 1995, Li et al. (3) described the crystal structure of porcine synovial collagenase (MMP-1) consisting of a catalytic domain and a second domain of ϳ200 amino acids homologous to hemopexin, a heme-binding glycoprotein. The MMP-1 hemopexin domain contains four units of four-stranded antiparallel ␤-sheets stabilized on its 4-fold axis by a cation (3). The domain constitutes a four-bladed ␤-propeller structure and was assumed to control the specificity of MMPs, affecting both substrate and inhibitor binding (3). Two cysteines at either end are conserved in all hemopexin domains of the MMP family.
Since 1992 it has been known that the gelatin binding of MMP-9 is mediated by the second fibronectin-type II domain (T2HU-2), although the presence of another gelatin binding site could not be excluded. Although T2HU-2 mediates binding, it is not the rate-limiting step in the hydrolysis of gelatin by the enzyme (29). Because Me 2 SO inhibits gelatin binding but not its degradation (29), and the hemopexin structure depends on the joining of the two cystein residues susceptible to Me 2 SO, this would suggest that Me 2 SO inhibits gelatin binding by destroying the hemopexin structure. Therefore, we postulated a second gelatin binding site within the MMP-9 hemopexin domain.
To obtain high amounts of protein, the MMP-9 hemopexin domain was expressed as a GST fusion protein in E. coli. A short sequence of buffer amino acids at the N terminus of the protein (GIPETKKLM) was attributed to the cloning procedure and most probably does not influence the binding properties in a significant way. MMP-9 activity in gelatin zymography and the migration of MMP-9-producing tumor cells was not influenced by the peptide GIPETKKLM (Figs. 3 and 5). The precise mass of the hemopexin domain, including the buffer amino acids, was measured to be 25,034 Da by electrospray mass spectrometry (Fig. 2D), which is close to the estimated molecular mass. The difference of 42.58 Da compared with the theoretical mass of 24,991.42 Da (M ϩ H ϩ ) might be explained by acetylation, a rather common phenomenon also observed for proteins from E. coli (30,31). The difference between the experimental and theoretical mass of the protein is 0.55 Da, which is an acceptable error of the instrument. However there may also be differences due to oxidation reactions occurring somewhere in the protein. MMP-9-PEX contains three cysteines and two methionines, which may all be candidates to undergo oxidation. Immunoreactivity with two different polyclonal antibodies further verified the identity of the purified protein, and circular dichroism spectra indicated the correct folded protein. When the enhanced chemiluminesence (ECL) reaction was allowed to proceed, a few bands of higher molecular weight turned up on the Western blots, which might represent cross-linked multimeric forms of the recombinant protein.
Secondary structure analysis (28) of the far UV CD spectrum reflects the ␤-sheet character of the protein (␣-helix ϭ 21%, ␤-sheet ϭ 35%, turn ϭ 17%). Inspection of the x-ray structures of the hemopexin domains of MMP-2 and MMP-13 revealed ␤-sheets only and no helical secondary structure (32,33). The overestimation of the calculated ␣-helical content of the MMP-9-PEX is due to the negative band around 228 nm, which might originate from aromatic side chain contributions (34).
For MMP-2 it could already be shown that the C-terminal domain exhibits strong binding properties for fibronectin and heparin and that binding depends on the structural Ca 2ϩ but not Zn 2ϩ ion in this domain (7). In these studies, however, the influence of the hemopexin domain on enzymatic activity was not tested. By gelatin zymography, an easily practicable in vitro assay to determine gelatinase activity, we demonstrated a reduction of MMP-9 activity as a consequence of the presence of recombinant MMP-9-PEX by 60 -71%. A greater reduction in the case of smaller amounts of MMP-9 may be explained by reaching the nonlinear range of zymography and densitometry when loading high amounts like 1 ng of MMP-9 protein (Fig. 3,  A and B).
To explain a reduction of MMP-9 activity in the presence of recombinant MMP-9-PEX, we carried out a comprehensive study to determine the kinetic parameters for the binding of MMP-9-PEX to gelatin. By plasmon resonance, we measured a dissociation constant of K d ϭ 2.64 ϫ 10 Ϫ8 M. The K d values of TIMP-1 for the MMP-9 latent and active species are 35 and 23.9 nM for the high affinity site (35). That means that the binding affinity of gelatin to MMP-9-PEX is as high as the binding of MMP-9 to TIMP-1. These results indicated that the association of MMP-9-PEX and gelatin was rather rapid (k on ϳ3.8 ϫ 10 5 M Ϫ1 s Ϫ1 ). Furthermore, the dissociation of the complex was quite slow (k off ϳ10 Ϫ2 s Ϫ1 ), resulting in a very effective binding. Collectively, these findings support the importance of the C-terminal domain of MMP-9 for an efficient binding to the substrate gelatin. Others (36) have also characterized the binding parameters of pro-MMP-9 and demonstrated that the proenzyme binds with high affinity (K d ϳ20 -30 nM) to the surface of a variety of cell types, and the K d values of the ␣2(IV) chain of collagen IV for pro-MMP-9 were calculated to be ϳ45 nM. However, because a pro-MMP-9-TIMP-1 complex and MMP-9 binds to ␣2(IV), the authors suggested that neither the C-terminal (binds to TIMP-1) nor the N-terminal domain of MMP-9 (which is processed during activation) is directly involved in ␣2(IV) binding, but the domain was not identified further. From our results, we concluded that the C-terminal domain of MMP-9, although known to bind TIMP-1 with high affinity (35,37), is also able to bind to gelatin with a comparable high affinity.
We extended our characterization of MMP-9-PEX with respect to its impact on cell migration and found that migration of malignant human melanoma cells (A375) was reduced in the presence of recombinant MMP-9-PEX (Fig. 5). Because MMP-9-PEX was able to inhibit the migration of MMP-9-expressing FIG. 4. Kinetics of binding of MMP-9-PEX to gelatin. Association of MMP-9-PEX to immobilized gelatin led to the increase of the resonance angle as a function of time. The association phase was recorded for different nanomolar MMP-9-PEX concentrations. The decrease of the resonance angle indicates the dissociation phase resulting from replacement of the MMP-9-PEX solution from the solid phase by triplicate washing with 10 mM sodium acetate buffer, pH 3.9. The data obtained from the curves allowed the determination of the dissociation rate constant k off from the ordinate intercept and the association rate constant k on from the slope of the graph (see Table I). All kinetic experiments were performed at a controlled temperature (25°C). A375 cells but not of mesangial cells known to express MMP-2 (but no detectable amounts of MMP-9), we speculate that MMP-9-PEX interferes specifically with the MMP-9-dependent migration. MMP-9 is not crucial for only migration and metastasis but also for hematopoiesis. Recently, it has been published (38) that recruitment of stem and progenitor cells from the bone marrow requires MMP-9-mediated release of serum Kit ligand (sKitL). The relative deficiency of serum Kit ligand at baseline or after myelosuppression in MMP-9 Ϫ/Ϫ mice strongly suggests that MMP-9 plays a physiological role in releasing sKitL, setting up the stage for hematopoietic reconstitution. Thus, inhibition of MMP-9 (e.g. by MMP-9-PEX) may provide a novel mechanism for regulating hematopoiesis in sKitL-dependent myeloproliferative disorders.
We have presented MMP-9-PEX as an inhibitor of MMP-9 activity in vitro and in cell culture, which raises many questions and perspectives for further investigations. One important issue is the identification of the gelatin binding epitope within the MMP-9 hemopexin domain. In addition, the re-combinant MMP-9-PEX offers the possibility for determining other substrate specificities. Another important question relates to the effects of MMP-9-PEX in vivo. Because the catalytically inactive MMP-9-PEX binds with high affinity to gelatin, a substrate of MMP-9, we consider MMP-9-PEX as a candidate to antagonize MMP-9-activity in the context of tumor cell invasion and metastasis. Recombinant MMP-9-PEX may be used as a new approach in the treatment of diseases with high MMP-9 activity such as melanomas or colorectal carcinomas.

FIG. 5. MMP-9-PEX inhibits invasion of A375 melanoma cells.
A, zymography of the serum free supernatants of human malignant A375 melanoma cells, which secrete MMP-2 and MMP-9, and rat mesangial cells, which secrete MMP-2 only. An MMP-standard for human MMP-2 and MMP-9 is shown (left lane). B, invasion and migration of A375 melanoma cells were measured in a Boyden chamber in the presence of MMP-9-PEX at different concentrations as indicated (500 g/ml equivalent to 20 M). Cells were cultured with or without MMP-9-PEX for 24 h. For control, cells were incubated with GIPETKKLM (20 g/ml equivalent to 20 M) as indicated. Additional control cells were incubated with equal amounts of phosphate-buffered saline (0). Cells on the lower surface of the membranes were counted after staining with hematoxylin and eosin. y axis, number of cells in [%]. Each bar represents the mean ϮS.D. of three chambers. This is a representative result of three experiments. *, p Ͻ 0.01 compared with control.