Interaction of Pro-matrix Metalloproteinase-9/Proteoglycan Heteromer with Gelatin and Collagen*

Previously we have shown that THP-1 cells synthesize matrix metalloproteinase-9 (MMP-9) where a fraction of the enzyme is strongly linked to a proteoglycan (PG) core protein. In the present work we show that these pro-MMP-9·PG heteromers have different biochemical properties compared with the monomeric form of pro-MMP-9. In these heteromers, the fibronectin II-like domain in the catalytic site of the enzyme is hidden, and the fibronectin II-like-mediated binding to gelatin and collagen is prevented. However, a fraction of the pro-MMP-9·PG heteromers interacted with gelatin and collagen. This interaction was not through the chondroitin sulfate (CS) part of the PG molecule but, rather, through a region in the PG core protein, a new site induced by the interaction of pro-MMP-9 and the PG core protein, or a non-CS glycosaminoglycan part of the PG molecule. The interaction between pro-MMP-9·PG heteromers and gelatin was weaker than the interaction between pro-MMP-9 and gelatin. In contrast, collagen I bound to pro-MMP-9·PG heteromers and pro-MMP-9 with approximately the same affinity. Removal of CS chains from the PG part of the heteromers did not affect the binding to gelatin and collagen. Although the identity of the PG core protein is not known, this does not have any impact on the described biochemical properties of the heteromer or its pro-MMP-9 component. It is also shown that a small fraction of the PG, which is not a part of the pro-MMP-9·PG heteromer, can bind gelatin. As for the pro-MMP-9·PG heteromers, this was independent of the CS chains. The structure that mediates the binding of free PG to gelatin is different from the corresponding structure in the pro-MMP-9·PG heteromer, because they were eluted from gelatin-Sepharose columns under totally different conditions. Although only a small amount of pro-MMP-9·PG heteromer is formed, the heteromer may have fundamental physiological importance, because only catalytic amounts of the enzyme are required to digest physiological targets.

A large number of genetically unrelated proteins are known to contain highly negatively charged glycosaminoglycan (GAG) 2 chains. Such core proteins, substituted with GAG chains, constitute an entity of glycoproteins called proteoglycans (PGs). There are several types of GAG chains, where chondroitin sulfate (CS) and heparin/heparan sulfate (HS) are two major types (1). All GAG chains are unbranched, and they contain a variable number of negatively charged sulfate groups that are important for their function (2). Some PGs are associated with cells, whereas others are secreted and are a part of the extracellular matrix (ECM). Almost all mammalian cells synthesize PGs. Monocytes and macrophages synthesize PGs, which are mainly substituted with CS chains (CSPG) and only a minor proportion of HS (3,4). In resting monocytes most of the CSPG is not released but sorted to the endocytic pathway and degraded (4). However, when monocytes are stimulated and differentiated to macrophages, both the biosynthesis and the secretion of CSPG are increased (4). The human monocyte cell line THP-1 secretes PGs such as versican, perlecan, and serglycin (5,6). The biological role of the secreted PGs such as serglycin from macrophages is not clear, but it has been shown that it binds to other molecules released from the cells through interaction with the GAG chains (7,8), suggesting that this and other PGs may act as a kind of carrier molecule. It has also been shown that serglycin is constitutively produced by multiple myeloma plasma cells and can inhibit the bone mineralization process (9).
The family of matrix metalloproteinases (MMPs) consists of more than 20 different secreted and membrane-bound mammalian enzymes that are zinc-and calcium-dependent (10 -12). Together, the MMPs are able to degrade most ECM proteins, as well as regulating the activity of serine proteinases, growth factors, cytokines, chemokines, and cell receptors (10,(12)(13)(14). Thus, MMPs have complicated biological functions playing a role in normal and pathological conditions (10,15,16).
All MMPs are composed of various modules, including a proand catalytic domain. In addition, all the secreted MMPs with the exception of the two matrilysins (MMP-7/-26) also contain a C-terminal hemopexin-like domain (10,12). Typically, the secreted MMPs bind to the ECM proteins and proteoglycans (17). The two gelatinases MMP-9 and MMP-2 contain a unique inserted domain in their catalytic region, i.e. a module containing three fibronectin II-like repeats (FnII) (10,12). This domain is similar but not identical in the two gelatinases and is involved in the binding of denatured collagens, elastin, and native collagen (18 -23). The three FnII-like repeats in the catalytic site of MMP-2 and MMP-9 may facilitate the localization of these enzymes to connective tissue matrices. They also appear to be of importance for the degradation of macromolecules such as elastin, gelatin, and collagens IV, V, and XI but do not influence the degradation of chromogenic substrates (23)(24)(25)(26)(27).
MMP-9 (92-kDa gelatinase) is produced by a variety of cell lines, including monocytes and macrophages (28). MMP-9 is produced as a monomer as well as various dimer forms (29 -34). The homo-and several of the hetero-dimer forms are reduction-sensitive. Hence, the proteins are either covalently linked to each other through disulfide bonds or a very strong reversible interaction where intramolecular disulfide bonds are essential. Recently, we discovered that THP-1 cells produce a new type of reduction-sensitive heteromer, where pro-MMP-9 is linked to the core protein of one or several PGs (34). In the present investigation we have studied the binding properties of this pro-MMP-9⅐PG heteromers. We have especially focused on the binding to gelatin (denatured collagen) and collagen I.

Biosynthesis of PGs
The human leukemic monocyte cell line THP-1 was a kind gift from Dr. K. Nilsson, Dept. of Pathology, University of Uppsala, Sweden. The cells were cultured in RPMI 1640 medium with 10% fetal bovine serum, 50 g/ml streptomycin, and 100 units/ml penicillin. To isolate secreted cell-synthesized PGs and pro-MMP-9⅐PG heteromers, the cells were washed three times in serum-free medium and then cultured for 72 h in serum-free RPMI 1640 medium with 0.1 M PMA as described earlier (34). Conditioned medium was harvested, and loose cells were pelleted by centrifugation at 1000 rpm for 10 min. Pro-MMP-9 and pro-MMP-9⅐PG heteromers were thereafter isolated as described below.

S Labeling of GAG Chains
To label the GAG chains with [ 35 S]sulfate, 15 ϫ 10 6 THP-1 cells were incubated for 72 h in 20 ml of serum-free RPMI 1640 medium containing 50 g/ml streptomycin, 100 units/ml penicillin, 0.1 M PMA, and 50 Ci/ml of [ 35 S]sulfate. Conditioned medium was thereafter harvested, loose cells removed through centrifugation at 1000 rpm for 10 min. The cell-free conditioned medium was then applied to a Sephadex G-50 column to separate free [ 35 S]sulfate from PG-labeled [ 35 S]sulfate. The radioactively labeled PG was thereafter isolated as described below.

Biosynthesis and Isolation of Free GAG Chains
To obtain biosynthesis of free GAG chains, 0.067-0.67 mM hexylxyloside was added to the cell cultures (35). In some of the cultures, [ 35 S]sulfate was added. To isolate the free GAG chains and the small amount of intact PG formed, the harvested conditioned medium was first subjected to Q-Sepharose chromatography as described for the isolation of PG. Because this method does not separate the free GAG chains from the intact PG, the pooled and desalted fractions from the second Q-Sepharose column containing GAG and PG was applied to a Superose 6 (HR10/30) column. This column was run in 0.05 M sodium acetate, 0.05 M Tris/HCl, and 0.25% Chaps, pH 8.0. The flow rate was 0.4 ml/min, and fractions of 1 min were collected. Aliquots of the collected fractions were analyzed either by the Safranin O method (see below) or by scintillation counting.
Free 35 S-labeled GAG chains were also obtained by treating isolated 35 [S]PG by NaOH (final concentration, 0.5 M) overnight at room temperature. Prior to neutralization with HCl, 1.0 M Tris-HCl, pH 8.0, was added to the NaOH-treated PG to give a buffer concentration of 0.3 M.

Detection of Free and PG-bound CS Chains
Free and PG-bound CS chains were quantified spectrophotometrically by the Safranin O method (36) as described previously (34).

Isolation of Secreted PG and Pro-MMP-9⅐PG Heteromers
Secreted PG and the pro-MMP-9⅐PG heteromers were isolated by Q-Sepharose anion-exchange chromatography as described previously (34).

Purification of Pro-MMP-9 from the THP-1 Cells
The pro-MMP-9 in conditioned medium from the THP-1 cells was partly purified by subjecting the culture medium to a gelatin-Sepharose column (pre-equilibrated with 0.1 M Hepes buffer, pH 7.5), after PGs and pro-MMP-9⅐PG heteromers had been removed by Q-Sepharose anion exchange chromatography (without Urea present). Both the pro-MMP-9 monomer and homodimer forms bound to the gelatin-Sepharose column. The column was thoroughly washed with 0.1 M Hepes buffer, pH 7.5, containing 1.0 M NaCl and 5 mM EDTA, and then the bound pro-MMP-9 was eluted from the column with 10% of DMSO in 0.1 M Hepes/5 mM EDTA buffer, pH 7.5. The eluted and pooled MMP-9 fractions were passed over a Sephadex G-50 (fine) column, run in 10 mM Hepes, pH 7.5, followed by concentration in a SpeedVac (Savant). SDS-electrophoresis under reducing conditions, followed by either silver or Coomassie Blue staining, showed two bands, a major at 92 kDa and a minor band at 28 kDa. Western blotting revealed that the 92-kDa band was pro-MMP-9, and the 28-kDa band was TIMP-1. The amount of pro-MMP-9 was estimated spectrophotometrically at 280 nm using a ⑀ 280 nm ϭ 114.360 M Ϫ1 cm Ϫ1 (37), ignoring the contribution of TIMP-1.

cABC Lyase Treatment
The PG-bound CS chains were removed by digestion for 2 h at 37°C with 0.2-1.0 unit of cABC/ml in 0.05 M Tris-HCl, pH 8.0, containing 0.05 M sodium acetate.

Gelatin Zymography
SDS-substrate PAGE was done as described previously (34) with gels (7.5 cm ϫ 8.5 cm ϫ 0.75 mm) containing 0.1% (w/v) gelatin in both the stacking and the separating gel, 4 and 7.5% (w/v) of polyacrylamide, respectively. Gelatinase activity was evident as cleared (unstained) regions, and the area of these regions was quantified by the GelBase/GelBlot TM Pro computer program from Ultra Violet Products.

H Labeling of Calf Skin Collagen
Acid-soluble calf skin collagen was labeled with tritium by reductive methylation of the amino groups as described previously (38). Collagen denatured for 5-10 min at 90°C resulted in gelatin.

Gelatin and Collagen I Binding Assays
Gelatin-Sepharose Binding Assay-250 l of [ 35 S]PG (3 ϫ 10 6 to 9 ϫ 10 6 cpm) was applied to a 1.0-ml column of gelatin-Sepharose. As gelatin is known to bind to gelatinases through hydrophobic interactions, and bound enzyme is eluted from gelatin-Sepharose columns with DMSO (21,37), various hydrophobic compounds such as DMSO (5%), Brij-35 (0.05%), and Triton X-100 (0.5%) in 0.1 M Hepes buffer, pH 7.5, were used to elute the bound radioactive material. Both the material that passed through the column and the eluted material were analyzed for radioactivity.
To determine if the pro-MMP-9⅐PG and pro-MMP-9⅐PG core protein complexes bind to gelatin-Sepharose, 150 g of PG (untreated and cABC-treated) in 75 l of 0.1 M Hepes buffer, pH 7.5, was applied to a 300 l column. Both the fractions that passed through the column and the fractions eluted with DMSO (5%), Brij-35 (0.05%) and Triton X-100 (0.5%) in 0.1 M Hepes buffer, pH 7.5, were analyzed on gelatin zymography.
Gelatin and Collagen I Micro-well Binding Assay-To covalently link collagen type I or gelatin to micro-wells, equal amounts (3.0 g/well) of macromolecule in 0.1 M Na 2 CO 3 /HCl buffer, pH 9.6, were added to 384-well Nunc Immobilizer Amino plates and incubated for 2 h at room temperature. After wash in the same carbonate buffer, nonspecific binding sites were blocked by incubating the wells with 10 mM ethanolamine in the same carbonate buffer for 1 h. The wells were thereafter washed three times in 0.1 M Hepes buffer, pH 7.5.
Various amounts of pro-MMP-9 in 0.1 M Hepes buffer, pH 7.5, containing 5.0 mM EDTA were added to the wells for 1 h. The wells were then rinsed three times in washing buffer (0.1 M Hepes/5.0 mM EDTA, pH 7.5). The bound pro-MMP-9 was eluted with washing buffer containing 10% DMSO/0.05% Brij-35 and then subjected to gelatin zymography. The amount of eluted enzyme was estimated by using the GelBase/ GelBlot TM Pro computer program, and dissociation curves were obtained using a four-parameter non-linear curve-fitting algorithm (SigmaPlot, SPSS Corp., Chicago, IL) to calculate the amount of enzyme required to give a 50% saturation.
Because the exact amount of protein in the pro-MMP-9⅐PG heteromers was not possible to determine, the affinity of collagen I or gelatin to the pro-MMP-9 and its heteromers was determined by estimating the concentration of DMSO needed to release 50% of the bound enzyme and enzyme-heteromers from the respective ECM-coated wells. In these experiments, a suitable concentration of pro-MMP-9, pro-MMP-9⅐PG, or pro-MMP-9⅐PG core protein (in 0.1 M Hepes buffer, pH 7.5, containing 5.0 mM EDTA) was added to the well for 1 h. Then the wells were rinsed three times in washing buffer. Bound enzyme and enzyme-heteromers were eluted with increasing concentrations of DMSO in washing buffer containing 0.05% Brij-35. Released enzyme and enzyme-heteromers were detected by gelatin zymography. The percentage of DMSO required to release 50% enzyme/enzyme-heteromers was calculated from dissociation curves using a four-parameter non-linear curvefitting algorithm (SigmaPlot). All experiments were performed in duplicate and repeated at least four times.
To determine whether pro-MMP-9 and pro-MMP-9⅐PG heteromers competed for the same site on gelatin, competitive binding studies were performed. In these studies, micro-wells with bound gelatin were first incubated for 1 h at room temperature with 12 l of 70 nM of purified pro-MMP-9. The wells were then washed three times with washing buffer and again incubated for 1 h with 12 l of 0.7 g/l pro-MMP-9⅐PG heteromers (based on GAG determination using the Safranin O method) at room temperature. As controls, the gelatin-coated micro-wells were incubated with pro-MMP-9 or pro-MMP-9⅐PG heteromer alone. The wells were then rinsed three times with washing buffer. In some experiments the wells were first rinsed three times with a washing buffer containing 1 M NaCl followed by three times rinse in washing buffer. The latter procedure was used to test if the heteromers could bind to the micro-well bound pro-MMP-9 through its GAG chains. The bound pro-MMP-9 and pro-MMP-9⅐PG heteromers were eluted with 10% DMSO (in washing buffer) and subjected to gelatin zymography.
To determine whether gelatin and collagen I bind to the same site in pro-MMP-9⅐PG heteromers, another competitive binding study was performed. In this study, micro-wells with bound collagen were incubated for 1 h at room temperature with 12 l of either 40 nM of purified pro-MMP-9 or 4.75 g/l of pro-MMP-9⅐PG heteromers (based on GAG determination using the Safranin O method) that has been premixed with different amount of gelatin (167-8333 g/ml). Pro-MMP-9 was here used as a control of the assay, because it has been shown that both collagen and gelatin binds to the FnII repeats in the active site of the enzyme (19,22). As controls, the collagen-coated micro-wells were incubated with pro-MMP-9 or pro-MMP-9⅐PG heteromers alone. The wells were then washed three times with washing buffer. The bound pro-MMP-9 and pro-MMP-9⅐PG heteromers were eluted with 10% DMSO (in washing buffer) and subjected to gelatin zymography.
Trichloroacetic Acid Precipitation Assay-Because trichloroacetic acid does not precipitate PG (39), binding of gelatin to PG and proteins associated with PG was determined with a slightly modified form of the previously described soluble gelatinase assay method (38,40). Briefly, 100 l of purified PG in 0.1 M Hepes, pH 7.5, was mixed with 50 l of the 3 H-labeled gelatin solution (2.3 mg/ml or 6.6 ϫ 10 6 cpm/mg). Trichloroacetic acid (60%) was added to give a final concentration of 20%. The assay tubes were centrifuged for 4 min at 20,000 ϫ g using an Eppendorf Minifuge. Then 80 l of the supernatant was added to 3 ml of Ultima Gold XR, and the radioactivity was quantified in a scintillation counter.
In another type of experiment, 80 l of the above mentioned supernatant was neutralized with NaOH, and then 20 l of the neutralized supernatant was subjected to SDS-PAGE. Thereafter, the gel was soaked in Amplify and dried. Undegraded and degraded [ 3 H]gelatin was then detected with autoradiography.

Statistical Analysis
All assays were performed in at least triplicate with data presented as mean Ϯ S.D., using the Student t test.

Isolation of PG and Pro-MMP-9⅐PG
Heteromer-The conditioned serum-free media from PMA-treated THP-1 cells incubated for 72 h in the presence or absence of [ 35 S]sulfate was applied to Q-Sepharose (ion exchange) chromatography as described previously (34,41). This column separates PG and pro-MMP-9⅐PG heteromers from free pro-MMP-9 (Fig. 1a). 10 -15% of the synthesized pro-MMP-9 was strongly linked to PG, as also shown previously (34). The PG fraction that was eluted from the Q-Sepharose column thus contains a mixture of various PGs and pro-MMP-9⅐PG heteromers. This mixture was used in the different experiments to study the interaction between the pro-MMP-9⅐PG heteromers with gelatin and collagen I, as well as PG and gelatin.

PG Isolated from THP-1 Cells Contains Mostly CS Chains-
The PMA-differentiated cells were incubated in the presence of [ 35 S]sulfate. Analysis of the conditioned medium showed that the [ 35 S]sulfate was almost exclusively incorporated in the synthesized GAG chains (not shown). To determine which types of GAG chains were synthesized by the THP-1 cells, [ 35 S]PG was treated with cABC, which degrades CS but not HS chains. Approximately 98% of the radioactivity was removed from the [ 35 S]PG after cABC treatment, clearly indicating that the PG was almost exclusively substituted with CS chains. Hence, no more than ϳ2% of the GAG chains can be of the HS type. However, because cABC treatment has been shown to leave small stubs of CS on the PG core protein that may contain a sulfate group (42,43), it can be assumed that the small amount of radioactivity remaining linked to the purified proteins can at least partly be ascribed to these stubs. If the radioactivity left on the PG after cABC treatment is due to the presence of PG that contains only HS chains, this would be detected in SDS-PAGE as bands that do not change in size after the cABC treatment. Further, if any PGs contained only CS chains or both CS and HS chains, these should appear with a smaller molecular weight and less bound radioactivity. Autoradiography of gels after SDS-electrophoresis showed that the main 35 S-labeled PG was located in the stacking gel and in a zone that just entered the proMMM-9 PG proMMP-9/PG proMMM-9 PG proMMP-9/PG proMMM-9 PG proMMP-9/PG Bound Unbound Unbound FIGURE 1. Summary chart for separation of pro-MMP-9 (monomer and homodimer), PGs, and pro-MMP-9⅐PG heteromers. The charts show the fate of the PGs and pro-MMP-9⅐PG heteromers that is referred to in various parts of the text. The PGs were labeled with [ 35 S]sulfate and detected by scintillation counting, whereas the pro-MMP-9, including the heteromers, was detected by gelatin zymography. a, summarizes the amount of the pro-MMP-9, pro-MMP-9⅐PG, and PGs in the conditioned media from PMA-stimulated THP-1 cells that binds or not to Q-Sepharose. When 35 S-labeled material was analyzed, ϳ3% of the labeled macromolecules did not bind to the Q-Sepharose column. This unbound material consisted of sulfated glycoproteins and not PGs (data not shown). PGs and pro-MMP-9⅐PG heteromers bound to the column and were eluted in the same fractions (see "Experimental Procedures"). In contrast, the monomer and homodimer forms of pro-MMP-9 did not bind to the column. Of the total pro-MMP-9 synthesized, between 10 and 15% are linked to a PG core protein (pro-MMP-9⅐PG heteromer) (34). b, describes experiments where the isolated PGs (including pro-MMP-9⅐PG heteromers) are tested for the ability to bind to a gelatin-Sepharose column. The results reveal that the PGs and pro-MMP-9⅐PG heteromers are heterogeneous. The fractions of bound and unbound PGs include the pro-MMP-9⅐PG heteromers and hence reflect the entire amount of PGs. Shown also is the relative amount of bound PGs eluted with DMSO, Brij-35/ Triton X-100, and NaCl compared with the entire amount of PGs. The quantitative estimation of unbound and bound pro-MMP-9⅐PG heteromers is based on gelatin zymography and, hence, only reflects relative quantities of the enzyme complexes. Here, the amount of pro-MMP-9⅐PG eluted with either DMSO, Brij-35, or Triton X-100 is presented relative to the amount of bound complex to the column. Based on the quantitative estimation of bound and unbound PGs as well as bound and unbound pro-MMP-9⅐PG heteromers to the gelatin-Sepharose column, it was possible to estimate that ϳ0.1% of the total amount of PGs are pro-MMP-9⅐PG heteromers.
separating gel (Fig. 2). After cABC treatment, only weak bands could be detected even after 7 days of exposure of the film to the dried gel. These bands occurred both at the position of the untreated control [ 35 S]PG as well as at positions with reduced molecular weight. The majority of the radioactivity appeared at the same position as the tracking marker dye and is most likely due to the cABC-produced radioactive disaccharides. This suggests that only a very small amount of the radioactivity left on the PG after cABC treatment can be ascribed to PGs that contain only HS chains.
PG Isolated from THP-1 Cells Binds Gelatin-In contrast to other proteins, trichloroacetic acid does not precipitate PGs due to the presence of GAG chains (39). Thus, if a protein such as gelatin binds to purified PG, it can be expected that the bound protein would remain in solution along with the PG in the presence of trichloroacetic acid, while unbound proteins are precipitated. In the present work, 3 H-labeled gelatin was incubated with the PG isolated from THP-1 cells. Trichloroacetic acid was added to give a final concentration of 20%, which precipitates peptides larger than 5 kDa, followed by centrifugation. The amount of radioactivity left in the supernatant increased with increasing concentrations of intact PG (Fig. 3a).
As controls, 3 H-labeled gelatin was incubated with cABCtreated PG, buffer alone, or THP-1 culture medium where PGs had been removed by Q-Sepharose chromatography. In isolated PG where the CS chains have been removed by cABC degradation, the amount of radioactivity in the supernatant was the same as in the other two controls (data not shown). (a) FIGURE 3. Isolated PGs bind gelatin. a, different amounts of the isolated PGs were mixed with 115 g of 3 H-labeled gelatin (6.6 ϫ 10 6 cpm/mg). The amount of radioactivity in the supernatant after trichloroacetic acid precipitation was determined and presented as percent relative to the total amount of gelatin used in the assay. Corrections for background counts in the controls that lack PGs have not been done. The controls show that ϳ8% of the 3 H-labeled gelatin is not precipitated. In this representative experiment each point shows the mean Ϯ S.D. where n ϭ 4. The quantification of PGs was done by the Safranin O method. b, different amounts of PGs were mixed with gelatin as described in a. After the trichloroacetic acid precipitation, the supernatant was neutralized with NaOH, and an aliquot was applied to 7.5% SDS-PAGE. After electrophoresis the gel was treated as described under "Experimental Procedures" and analyzed with autoradiography. To verify that the radioactivity in the supernatant after trichloroacetic acid precipitation was due to intact gelatin, and not degraded gelatin, 3 H-labeled gelatin was added to increasing concentrations of PG. After trichloroacetic acid precipitation, supernatants were neutralized with NaOH, applied to SDS-PAGE electrophoresis, and analyzed with autoradiography. Increasing concentrations of PG resulted in increasing amounts of intact gelatin in the supernatant, whereas no gelatin was detected in the supernatant from the controls without PG (Fig. 3b). Also when 3 H-labeled gelatin was mixed with cABCtreated PG, no intact gelatin could be detected in the supernatant (data not shown). Thus the radioactivity left in the supernatant after trichloroacetic acid precipitation was due to intact gelatin bound to the PG and not due to degraded gelatin.
Intact PG but Not Free CS Chains Prevents Soluble Gelatin from Being Precipitated by Trichloroacetic Acid-To gain insight into the mechanism behind the interaction between gelatin and PG, we first investigated whether the binding was to the CS chains or a protein component of the isolated PG. Previously it has been reported that heparin binds to gelatin (44), whereas others did not detect interaction between these two molecules (45). To explore whether the CS chains are involved in the binding of soluble gelatin to PG, we isolated free CS chains from hexylxyloside-treated THP-1 cells. The intact PG and free CS chains were first purified on a Q-Sepharose column. To separate free CS chains from intact PG a Superose-6 column was used. Experiments with [ 35 S]sulfate-labeled material showed that ϳ80% of the CS chains were free (peak P2) and 20% were bound to PG core protein (peak P1) as shown in Fig.  4a. In experiments with unlabeled material, fractions from the two peaks were pooled. The material in the two pools was mixed with 3 H-labeled gelatin, followed by trichloroacetic acid precipitation. The P2 pool containing mainly free CS chains prevented only negligible amount of the 3 H-labeled gelatin from being precipitated compared with the effect produced by an equal amount of intact PG, i.e. the P1 pool (Fig. 4b). Thus, it can be concluded that 3 H-labeled gelatin did not bind to the CS chains in the PG, but to either the PG core protein or a protein linked to the PG, such as pro-MMP-9. Alternatively, free CS chains cannot prevent bound gelatin from trichloroacetic acid precipitation.
Only a Small Fraction of the PG Secreted from THP-1 Cells Binds to Gelatin-To determine how much of the secreted PG that binds to gelatin, [ 35 S]PG was applied to gelatin-Sepharose chromatography (Fig. 1b). Most of the radioactive material passed through the column, and only 0.5-1.5% was bound (Figs. 1b and 5a). Because gelatin is known to bind to gelatinases through hydrophobic interactions, and bound enzyme is eluted from gelatin-Sepharose columns with DMSO (21,37), various hydrophobic compounds such as DMSO, Brij-35, and Triton X-100 were used to elute the bound radioactive material. Only a minor fraction (1-2%) of the bound PG was eluted with 5.0% DMSO (Figs. 1b and 5a). Further elution with 0.05% Brij-35 resulted in additional release of 26 -51% of the bound PG (Fig.  5a). Most of the remaining PG (29 -40%) was eluted with 0.5% Triton X-100 (Fig. 5a). To assure that all the radioactive material bound to the column had been eluted, the entire column material was dissolved in scintillation fluid after Triton X-100 elution. This showed that 19 -32% of the total radioactive material was left on the column in the various experiments. This material could be eluted with 1 M NaCl (Fig. 1b). Approximately the same values for the eluted [ 35 S]PG was obtained when the elution started with Brij-35, followed by DMSO, and ended with Triton X-100 (data not shown). When 0.5% Triton X-100 was used as the starting eluant, all of the bound radioactive material was eluted, except for the peak eluted with DMSO (data not shown). Our results show that there is a heterogeneity in the nature of the PG, because most of the PG secreted from THP-1 cells did not bind to gelatin, and the small amount that bound was eluted in different fractions depending on the hydrophobic eluant.
A Large Part of the Pro-MMP-9⅐PG Heteromer Binds to the Gelatin-Previously we have shown that 10 -15% of the pro-MMP-9 secreted by THP-1 cells is strongly linked to a PG core protein forming a pro-MMP-9⅐PG heteromer (34). This heteromer was separated from the free pro-MMP-9 (monomer and homodimer) during the first Q-Sepharose purification step of PG (Fig. 1a). To investigate to what extent the pro-MMP-9⅐PG heteromer binds to gelatin, the bound fraction eluted from Q-Sepharose chromatography, containing unlabeled PG and MMP-9⅐PG heteromers, was subjected to gelatin-Sepharose chromatography. The bound material was eluted under the same conditions as in Fig. 5a. To determine in which fractions the pro-MMP-9⅐PG heteromer was found, each fraction was  3 H-labeled gelatin (115 g; 6.6 ϫ 10 6 cpm/mg) was mixed with 4 g of the pooled fractions P1 and P2 from unlabeled CS/CSPG followed by trichloroacetic acid precipitation as described in Fig. 3. Shown is the amount of radioactivity in the supernatant (corrected for the background counts in the controls that lack PGs) in % relative to the total amount of gelatin used in the assay. The quantification of the GAG chains and intact PG was done with the Safranin O method as described under "Experimental Procedures." applied to gelatin zymography. This revealed that 15-35% of the pro-MMP-9⅐PG heteromers was bound to the column while the rest of the heteromers passed through. The bound heteromers were eluted with DMSO ( Figs. 1b and 5b). No further heteromers appeared when DMSO elution was followed by either Brij-35, Triton X-100, or NaCl (Figs. 1b and 5b). If Brij-35 or Triton X-100 were used prior to DMSO, only minor amounts of the bound pro-MMP-9⅐PG was eluted with the two former compounds. The vast majority of the heteromers was obtained in the following DMSO fractions (data not shown). In contrast to the pro-MMP-9⅐PG-bound fraction, the flow through fraction of the heteromers did not bind to a second gelatin-Sepharose column. These experiments were repeated several times with almost identical results. In all experiments, the bound fraction contained mainly the 300-kDa pro-MMP-9⅐PG heteromer but in some experiments also a heteromer with an M r Ͼ 10 6 . To summarize, 15-35% of the pro-MMP-9⅐PG heteromers were bound to gelatin-Sepharose. Of the [ 35 S]PG bound to the gelatin-Sepharose column, 1-2% was eluted with DMSO. This is ϳ0.013% of the total amount of [ 35 S]PG subjected to the column. This indicates that the pro-MMP-9⅐PG heteromers constitutes ϳ0.1% of the total PGs produced (Fig. 1).
As shown by cABC treatment of 35 S-labeled PG, Ͼ98% of the GAG chains were of the CS type. Further, we have previously shown that the pro-MMP-9⅐PG heteromers mainly contain CS chains (34), as was also confirmed in material used in the present study (Fig. 6a). To determine if there were any differences concerning the amount of CS chains in the pro-MMP-9⅐PG heteromers that passed through the gelatin-Sepharose column versus the bound heteromers, material from these two fractions was treated with cABC. This showed that, in both the passthrough fraction and the bound fraction, most of the GAG chains were of the CS type (Fig. 6, b and c).
Previously we showed that the pro-MMP-9⅐PG heteromers could be activated by Ca 2ϩ ions, but not with the classic organomercurial activator of MMPs, p-aminophenylmercury acetate (41). As shown in Fig. 6 (b and c), the pro-MMP-9⅐PG heteromers in both the pass-through and the bound fractions could be activated by Ca 2ϩ ions.
The CS Chains Do Not Take Part in the Binding of Pro-MMP-9⅐PG Heteromers to Gelatin-The CS chains in the PGs were removed by cABC, and the obtained PG core proteins were applied to gelatin-Sepharose chromatography. The eluted fractions were analyzed by gelatin zymography. ϳ20% of the pro-MMP-9⅐PG core protein heteromers bound to the gelatin-Sepharose column, whereas the rest passed through the column (data not shown). In repeated experiments the percentage of bound heteromer varied between 15 and 30%. The amount of bound material was approximately the same for both intact and cABC-treated material, clearly indicating that the CS chains are not involved in the binding of the heteromers to gelatin. To verify that the CS chains did not prevent binding of the heteromers to gelatin, the pro-MMP-9⅐PG fraction that passed through the gelatin-Sepharose column was treated with cABC. ]PG (9.06 ϫ 10 6 cpm) was applied to a 1-ml gelatin-Sepharose affinity chromatography column, and fractions of 0.5 ml were collected and counted in a scintillation counter. Initially 0.1 M Hepes buffer, pH 7.5, was used to elute all unbound material. When the obtained radioactivity reach the background level, bound radioactive materials were eluted in three steps using first 5% DMSO, then 0.05% Brij-35, followed by 0.5% Triton X-100. All eluants were dissolved in 0.1 M Hepes, pH 7.5. b, unlabeled PG were produced and purified as described under "Experimental Procedures." 25 l of purified PG (185 g) was applied to a 200-l gelatin-Sepharose column, and fractions of 120 l were collected and applied to gelatin zymography. Unbound and bound material was obtained from the column as described in a. The arrowhead shows the border between the stacking and separating gel. As a standard (St), partly purified pro-MMP-9 was used, which contains a monomeric (92 kDa) and homodimeric (225 kDa) form. The formed pro-MMP-9⅐PG core proteins were then applied to a new gelatin-Sepharose column. All pro-MMP-9⅐PG core protein complexes passed through the column (data not shown), showing that the CS chains did not prevent the heteromers from binding to gelatin.

Pro-MMP-9⅐PG Heteromers Have Weaker Affinity to Gelatin Compared with Pro-MMP-9
Monomer-To determine the affinity of pro-MMP-9, pro-MMP-9⅐PG, and pro-MMP-9⅐PG core protein to gelatin and collagen I, these pro-MMP-9 variants were added to gelatin and collagen I-coated micro-wells. The bound pro-MMP-9 variants were eluted with DMSO, because this compound has been shown to release both pro-MMP-9 (21, 37) and its variants (see Fig. 5) from gelatin-Sepharose. The amount of pro-MMP-9 and its heteromers eluted from the micro-wells was determined by gelatin zymography. Normally, the affinity between various MMPs, recombinant collagen binding domains from MMP-2/-9, and various matrix molecules has been determined by adding increasing amounts of the former compounds to micro-wells coated with a certain matrix molecule (18,22). However, it was not possible to determine the exact amount of the pro-MMP-9⅐PG heteromers as well as its molecular size because of the heterogeneity of the material. Therefore it was not possible to compare the affinity based on the concentration of the pro-MMP-9 variants. Instead, the affinity was based on the concentrations of DMSO needed to dissociate 50% of the pro-MMP-9 variants from the gelatin and collagen I matrix.
To establish the method, increasing concentrations of purified pro-MMP-9 were added to gelatin and collagen I-coated micro-wells. After thorough washes, the bound enzyme was dissociated from the ECM molecules using 10% DMSO. Typical zymography gels are shown in Fig. 7 (a and b). Saturation curves occurred in all cases for the binding to gelatin (Fig. 7c). In contrast, binding of pro-MMP-9 to collagen I gave saturation curves only in ϳ50% of the experiments. A 50% saturation occurred at 24 Ϯ 3 nM (n ϭ 4) pro-MMP-9 for gelatin, whereas 33 Ϯ 7 nM (n ϭ 5) pro-MMP-9 when saturation curves were obtained with collagen I. This shows that pro-MMP-9 has a weaker affinity to collagen I than to gelatin as also shown by others (22). Based on these experiments, we decided to use 20 -25 nM pro-MMP-9 in the following affinity experiments. Similar experiments were performed to establish the amount of pro-MMP-9⅐PG and PG core protein heteromers needed to obtain enough bound material to study the affinity to gelatin and collagen I (data not shown). No pro-MMP-9 or pro-MMP-9 heteromers were bound to wells coated with either bovine serum albumin or 0.01 M ethanolamine.
To determine the affinity to gelatin, the three MMP-9 variants were added to gelatin-coated micro-wells. After incubation and rinsing, different concentrations of DMSO were used to elute the bound MMP-9 variants. As shown in Fig. 8, larger concentrations of DMSO were needed to release pro-MMP-9 than pro-MMP-9⅐PG heteromers from the gelatin, clearly indicating that pro-MMP-9⅐PG and pro-MMP-9⅐PG core protein bind weaker to gelatin than pro-MMP-9.
Identical experiments were performed to determine the affinity to collagen I. Fig. 8 shows that the three MMP-9 variants have a similar affinity, in contrast to what was observed for gelatin. All these experiments were repeated several times, and Fig. 8i summarizes the results. As can be seen, a higher concentration of DMSO was needed to elute pro-MMP-9 from gelatin than from collagen I. This confirms the results from the saturation experiments (Fig. 7) as well as what has been reported by other investigators that pro- Curve fitting and hence the amount of pro-MMP-9 required to give a 50% saturation (binding) was achieved using a four-parameter non-linear curve-fitting algorithm (Sig-maPlot). In these two representative experiments a 50% saturation was calculated to be 22 nM for gelatin and 28 nM for collagen I. MMP-9 binds stronger to gelatin than to collagen I (22). Further, our results show that the pro-MMP-9⅐PG and the pro-MMP-9⅐PG core protein complexes have an almost identical affinity to both ECM molecules (Fig. 8).

Binding of Pro-MMP-9⅐PG to Gelatin Is Not through the FnII Domain in the Pro-MMP-9
Part of the Complex-In the fraction of the pro-MMP-9⅐PG heteromers that binds gelatin (Fig.  1b), it is possible that this interaction is mediated by the FnII domain in the proenzyme. To determine whether this is the case or not, competitive binding studies were performed. Pro-MMP-9⅐PG heteromers bound to gelatin-coated wells that first had been saturated with pro-MMP-9 (Fig. 9a). To investigate whether the binding of the pro-MMP-9⅐PG heteromers to the gelatin-coated wells saturated with pro-MMP-9 is due to an interaction between the free enzyme and the GAG chains of the PG, wells were rinsed with washing buffer containing 1 M NaCl. This washing procedure had no effect on the binding of the pro-MMP-9⅐PG heteromers to the pre-coated wells, showing that the heteromer binds to the gelatin and not to the gelatinbound pro-MMP-9. Further, these results reveal that the heteromer did not bind to gelatin through the FnII domains in the pro-MMP-9 part of heteromers. It also suggests that the two regions on the gelatin molecule that bind pro-MMP-9 and pro-MMP-9⅐PG heteromers are well physical separated, because the two different molecules can bind simultaneously.
The Same or an Overlapping Region in the Pro-MMP-9⅐PG Heteromers Binds Gelatin and Collagen I-It is known that gelatin and collagen bind to the same region (the FnII-like domain) in the catalytic site of MMP-9, and that the former molecule has a stronger interaction (22). To determine whether gelatin and collagen bind to the same or an overlapping site in the pro-MMP-9⅐PG heteromer, competition studies were performed. Micro-wells were coated with collagen I, and then pro-MMP-9 and pro-MMP-9⅐PG heteromers mixed with increasing concentrations of gelatin were added to the wells. As shown in Fig.  9b, increasing concentrations of gelatin prevented the binding of both pro-MMP-9 and pro-MMP-9⅐PG heteromer to the collagen-coated wells. The left part of Fig. 9b displays pro-MMP-9, which was used as a positive control for the method. Because gelatin prevents the binding of pro-MMP-9⅐PG to the wells, it can be concluded that gelatin and collagen bind to the same or an overlapping site in the heteromer.

DISCUSSION
PGs are known to interact with a lot of molecules, including ECM proteins like collagen and gelatin (46). The present work with the monocytic cell line THP-1 shows that pro-MMP-9⅐PG complexes are heterogeneous with respect to their ability to bind collagen and gelatin. A significant fraction of the heteromers bound gelatin and collagen as summarized in Fig. 1. Only a minor fraction of the total amount of PGs bound to gelatin. Similar to the pro-MMP-9⅐PG complexes, the PGs that were not linked to pro-MMP-9 were heterogeneous with respect to their ability to bind gelatin. The discussion will focus on the ability of these fractions to interact with the two ECM molecules, as well as on the heterogeneity within these fractions.
The Classic Gelatinase Assay Is Not Always Reliable-The classic quantitative gelatinase activity assay is based on the use of radioactive labeled gelatin (38,40). Enzyme-generated fragments Ͻ5 kDa cannot be precipitated by 15-20% of trichloroacetic acid. Hence, the amount of radioactivity left in the supernatant after trichloroacetic acid precipitation is used as a measure of the enzyme activity. As shown in the present work, this assay cannot be used when gelatin is bound to PGs, because PG prevents trichloroacetic acid precipitation of bound proteins. Therefore, when this assay is used, one should always check that the gelatin left in the supernatant is degraded. . Gelatin and collagen bind to the same site in the pro-MMP-9⅐PG heteromers, and this is not the FnII repeat in the catalytic domain of pro-MMP-9. a, Nunc micro-wells were coated with 3 g of gelatin as described under "Experimental Procedures." After thorough rinses with washing buffer, 70 nM pro-MMP-9 was added to two wells, and 700 g/ml (8.4 g) of pro-MMP-9⅐PG heteromers was added to another well and then incubated for 1 h at room temperature. After three rinses with wash buffer, 700 g/ml (8.4 g) of pro-MMP-9⅐PG complex was added to one of the wells that contained bound pro-MMP-9 and then incubated for 1 h at room temperature. After three rinses with wash buffer, bound enzymes were released with 12 l of elution buffer containing 10% DMSO, and the eluted material was thereafter mixed with 3 l of sample buffer. 6 l of eluted sample mixture was analyzed by zymography, and the amounts of bound pro-MMP-9 variants were quantified using the GelBase/GelBlot TM pro program. The results are representative of at least three independent experiments. b, Nunc micro-wells were coated with 3 g of collagen I as described under "Experimental Procedures." After thorough rinses with washing buffer, 12 l of 40 nM pro-MMP-9 containing increasing amounts of gelatin (167 to 4167 g/ml) and 4.75 g/l pro-MMP-9⅐PG heteromers (based on GAG determination using the Safranin O method) containing increasing concentrations of gelatin (167-8333 g/ml) was added to four wells. As controls, corresponding amounts of the pro-MMP-9 variants without gelatin were added to two other wells. After 1-h incubation at room temperature, wells were rinsed three times with washing buffer, and thereafter the bound enzymes were released with 12 l of elution buffer containing 10% DMSO. Eluted samples were analyzed by zymography, and the amount of bound pro-MMP-9 was quantified using the GelBase/GelBlot TM pro (UVP) program. The results are representative of at least three independent experiments.
ionic bonds. However, there is a discrepancy in the literature whether gelatin and collagen bind to GAG chains (44,45,47). Binding of gelatin to PGs secreted from the THP-1 cells as described in the present work is not likely to occur through the CS chains. This is based on the following observations: (i) only detergents could elute the bound PG from gelatin-Sepharose (Figs. 1b and 5a), hence the binding is through hydrophobic interactions. (ii) Only a minor fraction of the secreted PGs binds gelatin (Figs. 1b and 5a). If the binding had been through the CS chains, one would expect that much more PG would have bound to the gelatin-Sepharose column, because CS chains are known to be relatively homogeneous and represent Ͼ98% of the GAG chains in the isolated PGs from our PMAtreated THP-1 cells. (iii) Free CS chains did not prevent trichloroacetic acid precipitation of gelatin, in contrast to intact proteoglycan (Fig. 4). These results fits well with a previous study that showed that gelatin cannot bind heparin (45). However, it cannot be totally excluded that gelatin interacts with some other type of GAG chains linked to the PG core protein, although the CS counts form Ͼ98% of the GAG chains.
The Pro-MMP-9⅐PG Heteromer Did Not Bind Gelatin through CS Chains-The binding of gelatin to the pro-MMP-9⅐PG complex (Fig. 1b) did not appear to involve the CS chains, because enzymatic removal of the CS chains by cABC did not prevent binding. In addition, the affinity was approximately the same with or without the CS chains. Because gelatin is not bound to the heteromers through the CS chains, the interaction must be either to the PG core protein, the pro-MMP-9, or another protein associated with the heteromer. As for the free PGs, it cannot be excluded that gelatin interacts with some minor type of GAG chains linked to the PG core proteins in the heteromer.
Binding to Gelatin Involves Various Structures-PG bound to gelatin-Sepharose was eluted sequentially with the detergents Brij-35 and Triton X-100 (Fig. 5a). As the two detergents have similar hydrophobic properties and were used in different concentrations, it must be assumed that there are various structures in the PG molecules that bind gelatin. The pro-MMP-9⅐PG heteromers were only eluted with DMSO and not by the detergents (Figs. 1b and 5b), indicating that gelatin binds to a structure in this heteromer that is different from the corresponding binding structures in the PGs.
The FnII Domain of Pro-MMP-9 in the Pro-MMP-9⅐PG Heteromer Is Not Involved in the Binding of Gelatin-The pro-MMP-9⅐PG heteromer that was bound to a gelatin-Sepharose column or gelatin-coated micro-wells could be eluted with DMSO but not with Brij-35 or Triton X-100 (Fig. 1b). It is known that gelatin is bound to pro-MMP-9 through the three FnII repeats in the catalytic site of the enzyme (18,21,22). This interaction is due to hydrophobic forces, and the two molecules can be dissociated by DMSO (21,37). Because the pro-MMP-9⅐PG heteromers are also dissociated from gelatin by DMSO, although at lower DMSO concentrations than needed for pro-MMP-9 dissociation (Fig. 8), it could be expected that the heteromers are bound to gelatin through its pro-MMP-9 component. However, an excess of pro-MMP-9 could not prevent the binding of pro-MMP-9⅐PG heteromers to gelatin coated microwells (Fig. 9a). This clearly indicates that the heteromer did not interact with gelatin through the FnII domain in the catalytic site of the pro-MMP-9 part. Thus the FnII domain in the catalytic site of pro-MMP-9 is masked both in the pro-MMP-9⅐PG heteromers that did not bind gelatin, as well as in those heteromers that bound gelatin as indicated in the schematic model of a heterodimer (Fig. 10).
There are several possible models that can explain the interaction between pro-MMP-9⅐PG heteromers and gelatin. (i) The PG in the heteromer is identical with one of the PGs that are eluted with Brij-35 or Triton X-100 (Fig. 1b). In that case, the strong interaction between the PG and pro-MMP-9 must have hidden the gelatin-binding domain in both of these proteins. In addition, a new binding site in the PG core protein and/or pro-MMP-9 must have been induced. (ii) The PG itself was not able to bind gelatin. In this case, the strong interaction between the PG and pro-MMP-9 must have induced a new binding site in the PG core protein and/or pro-MMP-9. It is known from previous experiments that the interaction between pro-MMP-9 and PG core protein(s) has resulted in hiding as well as inducing new functional sites in pro-MMP-9. This was seen in activation experiments using p-aminophenylmercury acetate and calcium (41). In these experiments p-aminophenylmercury acetate, a known activator of pro-MMP-9, did not induce activation of the enzyme in the heteromer. Further, Ca 2ϩ , a known stabilizer of MMP-9, acted as an activator of the enzyme in the heteromer (41). (iii) The PG part of the heteromer is different from those PGs that containing an active site zinc ion (Zn) and the three fibronectin II-like repeats (FnII) that are known to bind gelatin and collagen. The C-terminal end of the enzyme contains the hemopexin like domain (Hpx), which has a structure of a four bladed ␤-propeller, and this domain is linked to the catalytic domain through a hinge region (H). Previously we have shown that the core protein of one or several PGs bind in a reduction-sensitive manner to the C-terminal part of the Hpx domain in pro-MMP-9 (34,41), here indicated with a disulfide bridge. The schematic presentation of the pro-MMP-9⅐PG heterodimer emphasizes that the interaction between the PG core protein and the enzyme, independently of the CS chains, prevented the binding of gelatin to the FnII repeats in the catalytic site in the enzyme (present work) as well as altered the ability of the enzyme to be activated (41).
passed through the gelatin-Sepharose column and those that were eluted with Brij-35 and Triton X-100. It is known that THP-1 cells synthesize various types of PGs (5,6). It is possible that only one of these PGs is strongly linked to pro-MMP-9 and forms the heteromer. In that case this particular PG must be synthesized only as a part of the pro-MMP-9⅐PG heteromer and have other biochemical properties than the other PGs.
The Pro-MMP-9⅐PG Heteromer Binds Collagen I and Gelatin with Approximately the Same Affinity-Previously it has been shown that collagen I, like gelatin, binds to MMP-9 through the three FnII-like repeats in the catalytic site of the enzyme (18,21,22). The interaction is stronger for gelatin than for collagen I (22). This fits well with the results in the present work where it is shown that higher concentrations of DMSO are needed to release pro-MMP-9 from gelatin-coated micro-wells than from collagen I-coated micro-wells. Gelatin had lower affinity to the binding site in the pro-MMP-9⅐PG heteromers than to the binding site (FnII) in pro-MMP-9. The binding region in the pro-MMP-9⅐PG heteromers also appears to have slightly different properties than the binding region in pro-MMP-9, because it binds gelatin and collagen I with similar affinity. Because pro-MMP-9⅐PG heteromers can bind to gelatin-coated micro-wells saturated with pro-MMP-9 (Fig. 9a), our results indicate that the heteromers and the pro-MMP-9 monomer bind to different and well separated parts of the gelatin molecule. This suggests that pro-MMP-9⅐PG heteromers bind to collagen I as well as denatured collagen in vivo, and hence can be localized to these molecules along with free pro-MMP-9.
MMP-9 Dimers and Complexes Have Altered Biochemical Characteristics Compared with the MMP-9 Monomer-MMP-9 is known to form various types of dimers, including homo-and heterodimers that involve the C-terminal hemopexin-like domain of the enzyme (29 -34). Some of these dimers are detected in SDS-PAGE under non-reducing conditions, but not under reducing conditions. Hence these dimers are reduction-sensitive and assumed to be linked through one or several S-S bridges. Recombinant MMP-9 hemopexin domain (PEX9) also formed a reduction-sensitive homodimer (48). X-ray crystallography revealed that the reduction sensitivity of the dimer was not due to an intermolecular disulfide bridge but an intramolecular bridge. The disulfide bond between the conserved Cys-516 and Cys-704 connects blades I and IV and is critical for the structural integrity of PEX9. There is a remaining free cysteine residue within the PEX9 domain, Cys-674, which is buried and hence not involved in the dimerization of PEX9. The crystal structure showed that the dimerization was due to non-covalent and mainly hydrophobic interactions of the two PEX9 domains where most of the dimer contacts involved blade IV. In addition, one salt bridge between the C terminus of PEX9(A) and the side chain of Arg-677 of PEX9(B) also contributed to the PEX9 dimer contact.
Formation of different MMP-9 complexes results in altered biochemical properties of the enzyme. In cells that produce both pro-MMP-9 and TIMP-1, these two molecules are bound together through their C-terminal domains, and the presence of TIMP-1 affects the activity of the enzyme (28). When pro-MMP-9 forms a dimer with collagenase, the binding to TIMP-1 is prevented (21). There are conflicting data concerning whether the pro-MMP-9 homodimer is able to form a complex with TIMP-1 (29,32,48). There are however characteristics that are somewhat different between the pro-MMP-9 monomer and homodimer, because the monomer is more rapidly activated by MMP-3 than the homodimer (32). In its heterodimer form with neutrophil gelatinase-associated lipocalin, pro-MMP-9 can bind TIMP-1 and form a ternary complex (49). Activation of the enzyme with plasma kallikrein and HgCl 2 is enhanced in the pro-MMP-9⅐neutrophil gelatinase-associated lipocalin complex (50), and the enzyme is protected from degradation (51). The interaction between the C-terminal domain of pro-MMP-9 and a PG core protein has also been shown to alter the ability of pro-MMP-9 to be converted into an active enzyme (41). The pro-MMP-9 monomer and homodimer are known to be activated by the organomercurial compound p-aminophenylmercury acetate, whereas the pro-MMP-9⅐PG complex was not activated by this compound (41). In contrast to this, Ca 2ϩ , which is known to stabilize MMP-9 and other MMPs but not induce activation of the proenzymes, induced an autoactivation of the pro-MMP-9 in the complex. The presence of Ca 2ϩ also resulted in activated enzyme forms released from the complex, due to cleavage of both a part of the PG core protein as well as the C-terminal hemopexin domain of the enzyme (41). Both MMP-9 and MMP-2 interact with gelatin as well as collagen through the three FnIIlike modules in their catalytic domain (18,21,22). This interaction is also important for the ability of these enzymes to degrade these physiological substrates, but has no effect on their degradation of other physiological substrates or chromogenic peptide substrates (18 -20, 24, 25, 27). In the present work it is shown that, when pro-MMP-9 is bound to PG core proteins, the enzyme cannot bind gelatin. Hence, the interaction with PG core proteins results in hiding of the gelatin binding sites in the FnII-like modules of the enzyme. Thus, the interaction between pro-MMP-9 and PG core proteins has resulted in changes of several biochemical properties of the enzyme. Based on this it is tempting to assume that an activation of pro-MMP-9⅐PG, which results in an active MMP-9 that is still attached to the PG core protein, will have altered biochemical properties compared with unbound active MMP-9. Such properties may include substrate specificity, catalytic efficiency, ability to interact with inhibitor molecules, and hence an altered regulation of the enzyme activity.
Putative Physiological Functions of the MMP-9⅐PG Complex-Because the pro-MMP-9⅐PG complexes are produced and secreted from cells, and is not an artifact produced during the isolation procedure (34), the complex is likely to play a physiological role. Despite the small amount of pro-MMP-9⅐PG heteromer formed, the complex may have fundamental physiological importance, because only catalytic amounts of the enzyme are required to digest physiological targets. Also, the heteromers have biochemical properties different from the monomeric enzyme. The PG part of the heteromers may mediate binding to proteins that are not recognized by monomeric MMP-9 and other MMP-9 dimers and hence expose the enzyme to new targets.
PGs are known to interact with structural ECM molecules as well as cell surface receptors through either their core proteins or the GAG chains. Among the various structural ECM and cell surface receptor molecules to which PGs can bind are collagens, laminins, fibronectin, fibrin, elastin, fibrillins, hyaluronic acid, CD44, epidermal growth factor receptor, integrins, and selectins (46). Several types of molecules such as proteinases, proteinase inhibitors, growth factors, cytokines, and chemokines bind to the GAG chains of PGs (1,7). This binding appears to be of importance for signaling events where the bound molecules have an altered activity. In other cases the bound molecules are protected from degradation. MMPs such as MMP-2 and MMP-7 have been found associated with GAG chains, the former through its C-terminal hemopexin-like domain (17,52,53). In contrast to MMP-2, MMP-9 does not bind to GAG chains (54,55), but instead can be bound strongly to one or more PG core proteins (34). The nature of the PG in the pro-MMP-9⅐PG heteromer is still not known. We have previously shown that THP-1 cells synthesize serglycin (6), and the PG in the pro-MMP-9⅐PG heteromer may therefore be serglycin. It is interesting to note that serglycin is a potent inhibitor of bone mineralization in vitro (9). Because both the inorganic (calcium and phosphate) and the organic (collagen I) components of the bones have a high turnover rate, it is tempting to assume that, under conditions where PGs like serglycin accumulate, the presence of a pro-MMP-9⅐PG heteromer would have resulted in an calcium-induced activation and release of the bound gelatinase, which may participate in the remodeling process. Tumor cells migrate to various tissues, organs, and body cavities. The synthesized PGs and proteolytic enzymes like MMPs seem to be of importance for the migration of cancer cells. The invasive capability of the monocytic leukemia cell line THP-1 increases after PMA stimulation, and this appears to be correlated to an increase in CD147 (EMMPRIN), MMP-2, and MMP-9 (56). It may be assumed also that the MMP-9⅐PG heteromers can be of importance for cellular migration and invasion.