Specific, High Affinity Binding of Tissue Inhibitor of Metalloproteinases-4 (TIMP-4) to the COOH-terminal Hemopexin-like Domain of Human Gelatinase A

The binding properties of the newly described tissue inhibitor of metalloproteinases-4 (TIMP-4) to progelatinase A and to the COOH-terminal hemopexin-like domain (C domain) of the enzyme were examined. We present evidence for the first time of a specific, high affinity interaction between TIMP-4 and the C domain of human gelatinase A and show that TIMP-4 binds both progelatinase A and the C domain in a similar manner to that of TIMP-2. Saturable binding of recombinant C domain to TIMP-4 and to TIMP-2 but not to TIMP-1 was demonstrated using a microwell protein binding assay. The recombinant collagen binding domain of gelatinase A, comprised of the three fibronectin type II-like repeats, did not bind to TIMP-4, indicating that binding is mediated selectively by the C domain. Binding to TIMP-4 was of high affinity with an apparent K d of 1.7 × 10−7 m but slightly weaker than that to TIMP-2 (apparent K d of 0.66 × 10−7 m). Affinity chromatography confirmed the TIMP-4-C domain interaction and also showed that the complex could not be disrupted by 1 m NaCl or 10% dimethyl sulfoxide, thereby further demonstrating the tight binding. To verify the biological significance of this interaction, binding of full-length progelatinase A to TIMP-4 was investigated. TIMP-4 and TIMP-2 but not TIMP-1 bound specifically to purified TIMP-2-free human recombinant full-length progelatinase A and to full-length rat proenzyme from the conditioned culture medium of ROS 17/2.8 cells. Preincubation of the C domain with TIMP-2 was found to reduce subsequent binding to TIMP-4 in a concentration-dependent manner. Competition between TIMP-2 and TIMP-4 for a common or overlapping binding sites on the gelatinase A C domain may occur; alternatively TIMP-2 may prevent the binding of TIMP-4 by steric hindrance or induction of a conformational change in the C domain. We propose that the binding of progelatinase A to TIMP-4 represents a third TIMP-progelatinase interaction in addition to that of progelatinase A with TIMP-2 and progelatinase B with TIMP-1 described previously. This new phenomenon may be of important physiological significance in modulating the cell surface activation of progelatinase A.

Gelatinase A (72-kDa gelatinase, MMP-2, EC 3.4.24.24) belongs to the family of matrix metalloproteinases (MMPs), 1 a group of zinc-dependent enzymes, which together can degrade all components of the extracellular matrix (1,2). Gelatinase A cleaves several connective tissue matrix proteins including denatured collagens (gelatins), native type IV, V, VII, X, and XI collagens, aggrecan, elastin, and fibronectin (3)(4)(5)(6)(7)(8). The MMPs share common structural features (3,9), and gelatinase A is typical of the family in that it is secreted as a proenzyme that is activated by proteolytic processing to remove an 80-amino acid propeptide (10,11). The remainder of the NH 2 -terminal domain (N domain) of gelatinase A contains a catalytic zincbinding site that is common to all MMPs and also a fibronectin type II-like module triple repeat (12,13) found only in the gelatinases that forms a collagen binding domain (CBD) (14 -16). The hemopexin/vitronectin-like COOH-terminal domain (C domain) present in all MMPs except matrilysin does not appear to be required for catalysis by gelatinase A (17,18), although it may have a role in binding substrates such as fibronectin and other extracellular matrix components such as heparin (19) and fetuin. 2 Matrix metalloproteinase activity in the extracellular matrix is regulated by a family of specific inhibitors known as TIMPs (tissue inhibitor of metalloproteinases). Until recently, three members of this family had been characterized: TIMP-1 (20,21), TIMP-2 (22,23), and TIMP-3 (24,25). The N domains of these proteins share structural similarities (26 -28) and inhibit the activity of the MMPs by forming a 1:1 molar stoichiometric complex with the active enzyme, which is essentially irreversible (20,29,30). A new fourth member of the TIMP family (TIMP-4) has very recently been identified (31,32); the predicted sequence of this protein shares a 37% homology with TIMP-1 but a 51% identity with TIMP-2 and TIMP-3 (31). Gelatinase A is unique among the MMPs in that the latent enzyme binds TIMP-2 to form a tightly bound 1:1 molar stoichiometric complex (22,(33)(34)(35), whereas only the active forms of the other MMPs can bind this inhibitor. Binding of TIMP-2 to progelatinase A occurs via the C domains of the enzyme (17,18,34,36) and inhibitor (37,38); and with the other MMPs, an additional binding site for the N domain of TIMP-2 is also present in the catalytic domain of active gelatinase A (18,22,33,34). A further binding site for the N domain of TIMP-2 is also found on the C domain of gelatinase A (39), although the biological significance of this interaction is not yet clear. The binding of TIMP-2 to the C domain of progelatinase A is involved in the activation mechanism of the enzyme. Unlike the other MMPs, progelatinase A fails to activate after treatment with proteinases such as plasmin, plasma kallikrein, neutrophil elastase, or cathepsin G (5,8). Instead, a cell membranemediated activation process takes place (36, 40 -42) in which the activator has been identified as a membrane type (MT) MMP (43,44). Activated MT-MMP can act as a cell surface receptor for TIMP-2, which in turn acts as a receptor for progelatinase A, the latter binding via its C domain (43).
The interaction of progelatinase A with TIMP-2 appears to be specific in so far as a similar interaction does not take place between the latent enzyme and TIMP-1 as it does for gelatinase B (13,45). However, the binding properties of TIMP-3 and TIMP-4 to progelatinase A have not yet been ascertained. We have therefore investigated the binding of gelatinase A to the newly identified TIMP-4 protein and identified the site of interaction utilizing recombinant domains of the human enzyme. These investigations have demonstrated specific binding of TIMP-4 to both full-length progelatinase A and to the C domain, which is closely similar to that of the TIMP-2-progelatinase A interaction.

EXPERIMENTAL PROCEDURES
Expression and Purification of Recombinant Gelatinase A C Domain and CBD-Expression and purification of recombinant C domain (rC domain) and recombinant CBD (rCBD) was carried out essentially as described previously (14,19). Briefly, the recombinant protein encoding the C domain and linker (Gly 417 -Cys 631 , exons 9 -13) and the CBD (Val 191 -Gln 364 , exons 5-7) of human gelatinase A were both expressed with a short NH 2 -terminal fusion tag that included an initiation methionine and a (His) 6 tag, using the expression vector pGYMX (14). Inclusion bodies from LE392 Escherichia coli expressing the recombinant proteins were solubilized in 6 M guanidine HCl, and recombinant protein was refolded in 0.1 M Na 2 B 4 O 7 , pH 10, followed by dialysis into either 20 mM or 100 mM Na 2 HPO 4 ⅐7H 2 0, NaH 2 PO 4 , pH 8.0, 0.5 M NaCl. rC domain and rCBD were purified using Zn 2ϩ -charged chelating Sepharose 6B (Pharmacia Biotech Inc.) affinity chromatography essentially as described previously (14,19). rCBD eluted from the Zn 2ϩchelate column was additionally applied to gelatin Sepharose 4B (Pharmacia) to select functionally folded CBD from the Zn 2ϩ -chelating Sepharose elute (14). Protein yields were determined by the BCA assay (Pierce) and by spectroscopy after extinction coefficient determination (see below).
Expression and Purification of Recombinant TIMP-4 -Expression and purification of recombinant TIMP-4 protein was carried out according to Liu et al. 3 Briefly, the DNA sequence encoding TIMP-4 was ligated into the pA2-GP vector derived from pVL94 (46), and the vector was transfected into HB101 cells to produce recombinant baculovirus, which was then used to infect Sf9 insect cells. TIMP-4 was purified from the clarified cell supernatant by cation exchange, hydrophobic, and size exclusion chromatography columns. A soluble gelatin degradation assay (47) was used to assay for anti-MMP activity of the recombinant TIMP-4 during purification. 200 g of purified TIMP-4 was obtained for analysis.
Mass Spectrometry and Amino Acid Analysis-The mass of the rC domain was measured by electrospray mass spectrometry using a PESCIEX API 300 after sample injection on a C18 high pressure liquid chromatography column at 50 l/min. 10 l of rC domain in phosphatebuffered saline (PBS) (0.14 M NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 ⅐7H 2 0, 1.5 mM KH 2 PO 4 , pH 7.4) was subjected to amino acid analysis (amino acid analyzer from Applied Biosystems) using norleucine as an internal standard. The extinction coefficient was then determined from measurement of the pmol content of the protein sample.
Microwell Protein Binding Assay-Binding of the rC domain to TIMP-4, TIMP-2, and TIMP-1 was determined using an enzyme-linked immunosorbent type assay (14) with bovine serum albumin as a negative control. Purified recombinant TIMP-2 and purified natural TIMP-1 were kindly supplied by Dr. Y. DeClerck (Children's Hospital, Los Angeles, CA) and Dr. I. Clark (Addenbrooke's Hospital, Cambridge, UK), respectively. Each protein at 0.5 g/well in 15 mM Na 2 CO 3 , 35 mM NaHCO 3 , 0.02% (w/v) NaN 3 , pH 9.6, was coated on 96-well microtiter plates for 18 h at 4°C. Wells were blocked with 2.5% (w/v) bovine serum albumin, 0.1% (w/v) NaN 3 in PBS for 1 h at 21°C. Serially diluted rC domain (7.1 ϫ 10 Ϫ10 M to 1.15 ϫ 10 Ϫ5 M) was added in 100 l of PBS for 1 h at 21°C followed by washing with PBS, 0.05% (v/v) Tween 20 to remove unbound protein. Bound rC domain was quantitated using affinity purified anti-peptide ␣His 6 antibody (19) followed by incubation with goat anti-rabbit alkaline phosphatase conjugated secondary antibody (H & L chains, Bio-Rad Laboratories) and p-nitrophenyl phosphate disodium (Sigma) as substrate. Absorbance measurements (405 nm) were made in a microplate reader (Thermomax, Molecular Devices). Specificity was confirmed by comparing binding of the rC domain with that of the rCBD because both proteins contain the same fusion tag and were expressed in the same E. coli strain.
Human recombinant full-length gelatinase A uncomplexed with TIMP-2 and predominantly in the latent form (M r ϭ 72,000) (kindly supplied by Dr. R. Fridman, Department of Pathology, Wayne State University, Detroit, MI) was applied at approximately 2 g in 100 l of PBS to microwell plates coated with TIMP-4, TIMP-2, or TIMP-1 with gelatin as a positive control and myoglobin as a negative control (0.5 g/well). Gelatin was prepared from purified acid-soluble rat tail tendon type I collagen by heat denaturation at 56°C for 30 min. Rat progelatinase A complexed with TIMP-2 in 100 l of conditioned culture medium from ROS 17/2.8 cells (see below) was also applied to identically coated plates. Microwell plates were incubated at 21°C for 2 h followed by collection of unbound material for enzymogram analysis. Any remaining unbound and nonspecifically bound enzyme was removed by washing plates with PBS, 0.05% (v/v) Tween 20. Bound gelatinase A was solubilized in 50 l of 2 M urea, 2% (w/v) SDS, 0.125 M Tris-HCl, pH 6.8, and analyzed by gelatin enzymography. Unbound enzyme was also prepared and analyzed using the same buffer.
Affinity Chromatography-Binding and elution properties of the rC domain to TIMP-4 were also determined by affinity chromatography. A Zn 2ϩ -charged chelating Sepharose 6B minicolumn (V t ϭ 75 l) was equilibrated in PBS, and 50 g of rC domain was applied. Saturation of the binding sites on the affinity matrix occurred resulting in a small amount of excess protein being recovered in the unbound material. 20 g of purified TIMP-4 was applied, and elution was attempted with 1 M NaCl in 20 mM Na 2 HPO 4 ⅐7H 2 0, NaH 2 PO 4 , pH 7.4, 0.02% (w/v) NaN 3 followed by 10% (v/v) dimethyl sulfoxide in PBS and finally with 50 mM EDTA, pH 8.0. Chromatography fractions were analyzed by SDS-PAGE.
Cell Culture-ROS 17/2.8 cells were cultured in 75-cm 2 flasks (Becton Dickinson Labware) in ␣-minimum essential medium (Life Technologies, Inc.) supplemented with 10% (v/v) newborn calf serum (Life Technologies, Inc.). Confluent cells were washed twice in PBS, pH 7.4, and then cultured in serum-free medium with and without 20 g/ml concanavalin A (Sigma). Conditioned culture medium was collected after an 18-h incubation to obtain active (41) and latent secreted gelatinase A complexed with TIMP-2.

RESULTS
Purification of Gelatinase A rC Domain-In agreement with our previous studies (19), purified rC domain electrophoresed as a single band on 15% SDS-PAGE gels with an apparent M r of 26,500 under reducing condtions (Fig. 1B). Western blot analysis using two anti-peptide antibodies (19) verified the identity of the purified protein (not shown). The presence of an intact disulfide bond (Cys 440 and Cys 631 ) was indicated by a downshift in apparent M r of 0.8 under nonreducing conditions (not shown). The absence of intermolecular disulfide-linked multimeric forms of the protein was shown by SDS-PAGE and Western blotting under nonreducing conditions. Electrospray mass spectrometry measured the precise mass of the protein to be 25,924 (within 1 Da of the predicted M r of 25,924.9 for the NH 2 -terminal methionine processed form), thereby confirming the fidelity of correct expression. The molar extinction coefficient determined from amino acid analysis of the purified protein was 5.10 ϫ 10 4 M Ϫ1 cm Ϫ1 .
TIMP-4 Binding Properties of the rC Domain-Saturable binding of the rC domain to TIMP-4 in PBS was found with an apparent K d of 1.7 ϫ 10 Ϫ7 M using the microwell protein binding assay; bovine serum albumin was used as a negative control (Fig. 1A). To confirm the TIMP-4-rC domain interaction, the binding of rC domain to TIMP-4 was investigated further by affinity chromatography. Fig. 1B shows that TIMP-4 was not detected in either the unbound fraction or in fractions eluted with 1 M NaCl or 10% dimethyl sulfoxide after application to a Zn 2ϩ -chelating Sepharose minicolumn saturated with rC domain. TIMP-4 was recovered from the column by elution with 50 mM EDTA, pH 8.0, together with the bound rC domain (Fig. 1B).
Binding Specificity of rC Domain to TIMP-4 -The binding of rC domain to TIMP-4 was compared with that to TIMP-1 and TIMP-2 using the microwell protein binding assay. Saturable binding to both TIMP-4 and TIMP-2 was observed ( Fig. 2A) with the apparent K d for the interaction with TIMP-2 being 0.66 ϫ 10 Ϫ7 M (c.f. 1.7 ϫ 10 Ϫ7 M for TIMP-4). These data indicate that both TIMP-4 and TIMP-2 bind the rC domain with similar high affinities but that the interaction with TIMP-2 was slightly stronger. No significant binding to TIMP-1 occurred ( Fig. 2A), demonstrating that interaction of the rC domain with TIMP-4 and TIMP-2 was specific. Moreover, because the rCBD from gelatinase A did not bind TIMP-4 ( Fig. 2B), this demonstrated that binding to TIMP-4 is mediated selectively by the C domain of gelatinase A.
TIMP-4 and TIMP-2 Bind at Shared or Overlapping Sites on the rC Domain-To investigate whether TIMP-4 and TIMP-2 bind at a common site on the gelatinase A C domain, 10 pmol of rC domain was incubated for 1 h at 21°C in the presence of TIMP-2 (100, 40, 20, 10, 4, and 0 pmol). The reaction mix was then applied to microwell plates coated with TIMP-4 (0.5 g, 20 pmol/well) for 1 h. Binding to TIMP-4 was reduced in a concentration-dependent manner by preincubation of the rC domain with increasing amounts of TIMP-2, and maximal inhibition was reached at a 2:1 molar ratio of TIMP-2 to rC domain (Fig. 3).
Binding of Full-length Progelatinase A to TIMP-4 -Recombinant full-length progelatinase A uncomplexed with TIMP-2 was observed to bind TIMP-4 using the microwell protein binding assay (Fig. 4A), thereby confirming the biological relevance of the interaction found with the isolated rC domain. Binding to TIMP-2 and gelatin was also demonstrated (positive controls) but not to TIMP-1 and myoglobin (negative controls) (Fig. 4A). Of note, the small amount of active enzyme present (M r ϭ 59,000) was bound by gelatin but not by TIMP-4 or TIMP-2, indicating that the interaction observed for the proenzyme was not through the catalytic site. Proenzyme from concanavalin A treated and untreated ROS 17/2.8 cells also showed binding to TIMP-4, TIMP-2, and gelatin but no binding to TIMP-1 or myoglobin (Fig. 4B). However, the quantity of recombinant progelatinase A activity binding to microwell plates was much greater than the enzyme activity bound from ROS 17/2.8 cell conditioned medium. Although not quantitative, this was indicated by the marked difference in enzymogram digestion times required to obtain visible lysis bands of bound enzyme: 18 h for conditioned medium and 1 h for recombinant enzyme. Indeed, the majority of progelatinase A from conditioned culture medium did not bind to the microwells as shown by enzymogram analysis of the enzyme before application (digestion time of 3.3 h) (Fig. 4B). This indicates that most progelatinase A produced by ROS 17/2.8 cells is unavailable for binding to TIMPs, probably due to prior formation of a complex with TIMP-2.

DISCUSSION
All members of the TIMP family are characterized by their ability to inhibit MMP activity by forming essentially irreversible 1:1 molar stoichiometric complexes with the active enzymes (20,30). However, the association of TIMPs with the proforms of the MMPs is more specific, and so far only two such interactions have been described: that of TIMP-2 with the proform of gelatinase A (22,(33)(34)(35) and that of TIMP-1 with the proform of gelatinase B (13,45). In this report, we present evidence for the first time of a third TIMP-progelatinase interaction, that is, of the newly cloned TIMP-4 with progelatinase A.
Binding of TIMP-2 to progelatinase A has been shown to occur via the C domain of the enzyme (17,18,34,36); indeed, the C domain alone can bind TIMP-2 (18,34,36,37,39). Similarly, we have shown that both progelatinase A and the rC domain alone can bind TIMP-4, indicating that the binding site in the full-length proenzyme resides mainly or entirely within the C domain. The lack of binding of the rCBD to TIMP-4 demonstrates that this domain is unlikely to be involved in complex formation. However, the contribution of N domain binding sites elsewhere in the catalytic domain cannot be precluded, because for TIMP-2, the existence of such a site or sites on the proenzyme has been shown (50). Our data suggest that the binding mechanism of TIMP-4 to the C domain of gelatinase A may be closely similar to that of TIMP-2. First, the K d equilibrium values for the binding of rC domain to immobilized TIMP-2 and TIMP-4 on microwell plates were comparable, although binding to TIMP-2 was slightly stronger. Secondly, the elution profile of TIMP-4 following application to rC domain immobilized on an affinity column was identical to that of TIMP-2 under the same experimental conditions (39). In addition, preincubation of the rC domain with TIMP-2 (which would be expected to result in the formation of an rC domain-TIMP-2 complex) prevented subsequent binding to TIMP-4. This demonstrates that both inhibitors cannot bind simultaneously, suggesting the presence of a common or overlapping binding sites on the rC domain. Latent gelatinase A from ROS 17/2.8 cell conditioned culture medium is present mainly as an enzyme-TIMP-2 complex (22,33). The reduced binding of this enzyme to exogenous TIMP-4 and TIMP-2 therefore further indicates competition for a common or overlapping binding sites. However, the existence of separate sites in which binding of TIMP-2 blocks TIMP-4 binding, either by steric hinderance or by inducing a conformational change in the rC domain, cannot be precluded by our data.
The apparent dissociation constant for binding of the rC domain to TIMP-2 (0.66 ϫ 10 Ϫ7 M) is somewhat higher than that previously reported for the full-length enzyme or for the binding of the C domain to cells via TIMP-2 (34,43). This may be because additional interactions of TIMP-2 with other sites on the N domain outside of the active site and the CBD occur to increase the affinity of binding (50). Alternatively, in our experiments, TIMP-2 and TIMP-4 are bound to microwell plates, possibly via sites that are involved in binding to the rC domain, resulting in a weaker interaction. The location of the rC domain binding site on the TIMP-4 molecule requires further investigation; in the case of TIMP-2, a highly charged sequence QEFLDIEDP located at the COOH terminus of the inhibitor is proposed to occur as an exposed "tail" and to mediate binding to progelatinase A (37). The homologous sequence in TIMP-4 (KEFVDIVQP) (31) shares four conserved residues (bold type) and an additional two conservative substitutions (underlined), none of which are found in TIMP-1 or TIMP-3, suggesting that this sequence may be important in forming the binding site. The remaining nonconserved residues reduce the net negative charge of the peptide from Ϫ4 in TIMP-2 to Ϫ1 in TIMP-4, which may cause the binding of TIMP-4 to be weaker than that of TIMP-2. The location of binding sites for TIMP-2 and TIMP-4 on the rC domain are currently under investigation in our laboratory and appear to involve several positively charged clusters unique to gelatinase A on hemopexin-like modules III and IV.
The findings of this study raise the important question of the physiological role of TIMP-4 binding to progelatinase A. TIMP-4 appears to mimic TIMP-2 in this capacity, thereby implying functional redundancy between the two inhibitors. However, the expression of TIMP-4 has been shown to be highly tissue-specific (31,32) in contrast to TIMP-2, which is widely and constitutively expressed. This implies that TIMP-4 has a unique role that is distinct to that of TIMP-2. Binding of the C domain of progelatinase to TIMP-2 is required to mediate activation of the enzyme; this is believed to occur through the formation of a trimolecular complex of MT1-MMP, TIMP-2, and proenzyme on the cell membrane (43). We speculate that binding of progelatinase A to TIMP-4 could be involved in the formation of an alternative trimolecular complex, perhaps preferentially with one of the other MT-MMPs, giving rise to an alternative activation pathway which is functional only in tissues which express TIMP-4. Formation of the MT-MMP-TIMP-2-progelatinase A trimolecular complex appears to be sequential, such that preformed TIMP-2-progelatinase A complex will not bind (43) and is therefore resistant to activation (36,42,43). Hence, an alternative possibility is that complex formation between progelatinase A and TIMP-4 prevents the enzyme from association with membrane bound TIMP-2 and therefore blocks activation. Interestingly, the highest level of TIMP-4 expression is in the heart (31), in which cancer metastasis rarely occurs, a process that is believed to involve the expression and activation of gelatinase A. Perhaps this is because the presence of excess TIMP-4 prevents gelatinase A activation via complex formation with the proenzyme.
In summary, we have demonstrated specific, high affinity binding of the newly described inhibitor TIMP-4 to progelatinase A. Binding appears to occur mainly via the C domain of the enzyme and to resemble that of TIMP-2. Future work will address the sites of interaction of TIMP-4 on the C-domain of gelatinase A and the physiological significance of this phenomenon.