Human Tissue Inhibitor of Metalloproteinases 3 Interacts with Both the N- and C-terminal Domains of Gelatinases A and B

We compared the association constants of tissue inhibitor of metalloproteinases (TIMP)-3 with various matrix metalloproteinases with those for TIMP-1 and TIMP-2 using a continuous assay. TIMP-3 behaved more like TIMP-2 than TIMP-1, showing rapid association with gelatinases A and B. Experiments with the N-terminal domain of gelatinase A, the isolated C-terminal domain, or an inactive progelatinase A mutant showed that the hemopexin domain of gelatinase A makes an important contribution to the interaction with TIMP-3. The exchange of portions of the gelatinase A hemopexin domain with that of stromelysin revealed that residues 568–631 of gelatinase A were required for rapid association with TIMP-3. The N-terminal domain of gelatinase B alone also showed slower association with TIMP-3, again implying significant C-domain interactions. The isolation of complexes between TIMP-3 and progelatinases A and B on gelatin-agarose demonstrated that TIMP-3 binds to both proenzymes. We analyzed the effect of various polyanions on the inhibitory activity of TIMP-3 in our soluble assay. The association rate was increased by dextran sulfate, heparin, and heparan sulfate, but not by dermatan sulfate or hyaluronic acid. Because TIMP-3 is sequestered in the extracellular matrix, the presence of certain heparan sulfate proteoglycans could enhance its inhibitory capacity.

The tissue inhibitors of metalloproteinases (TIMPs) 1 are specific protein inhibitors of the matrix metalloproteinases (MMPs), a group of zinc-dependent enzymes that include collagenases, gelatinases, and stromelysins. Four forms of human TIMP have been cloned: TIMP-1 (1), TIMP-2 (2), TIMP-3 (3)(4)(5)(6), and, more recently, TIMP-4 (7). TIMP-1 and TIMP-2 are secreted by many cell types in culture and are found in body fluids and tissue extracts. TIMP-3 is unique in that it appears to be a component of the extracellular matrix (8 -10) and occurs in relatively small amounts, possibly being expressed during specific cellular events (11).
The TIMPs have comparable abilities to inhibit the active forms of the MMPs when assessed using macromolecular substrates (12,13) and have been shown to make tight binding noncovalent complexes with active MMPs with a 1:1 stoichiometry (14 -17). The inhibitors have related primary and secondary structures, consisting of an N-terminal subdomain of three disulfide bonded loops and a smaller C-terminal region also containing three loops (18 -20). The N-terminal domain of TIMP-1 and TIMP-2 can act as a functional inhibitor (19,21,22), interacting with the catalytic domain of the enzymes such that competition with low molecular weight substrate analogue inhibitors can be observed (23). 2 Using peptide substrate assays, it has been possible to demonstrate that TIMP-MMP complexes interact with K i values of 10 Ϫ9 to 10 Ϫ12 M (24). Comparative studies of the association rates of TIMP-1 and TIMP-2 with different members of the MMP family in our laboratory have shown exceptionally strong C-terminal domain interactions between TIMP-1 and gelatinase B and between TIMP-2 and gelatinase A, suggesting that complexes between the respective pro forms of these enzymes, the active sites of which are inaccessible, and inhibitors can also occur (20,25,26). This supports other biochemical studies of these complexes (27)(28)(29)(30).
In this study, we have assessed the ability of TIMP-3 to associate with active MMPs using a kinetic method, and we have compared this with TIMP-1 and TIMP-2. We have also investigated the contribution of the C-terminal domains of both gelatinase A and gelatinase B to the interaction with TIMP-3, because this has important implications for the regulation of proenzyme activation. We have tested the effect of heparin and other polyanions on TIMP-3 activity in our soluble kinetic assay to determine whether interaction with similar components of the extracellular matrix could affect the capacity of TIMP-3 to inhibit MMPs.
Kinetic Studies-Active enzymes were active site titrated against a standard preparation of TIMP-1 (20). TIMP-2 and TIMP-3 were active site titrated with stromelysin-1 that had been titrated against the standard TIMP-1. Assays were performed at 25°C for gelatinase A and gelatinase B or at 37°C for stromelysin-1 and matrilysin in a buffer containing 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10 mM CaCl 2 , and 0.05% Brij 35 (fluorometry buffer). Hydrolysis of 1 M substrate Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 for the gelatinases and matrilysin or Mca-Pro-Leu-Ala-Nva-Dpa-Ala-Arg-NH 2 for stromelysin-1 was followed using a Perkin Elmer LS 50B fluorescence spectrometer (20,25,38). Inhibition of the matrix metalloproteinases by TIMPs was analyzed under pseudo-first-order conditions using suitable ratios of enzymes: inhibitors as described previously (20,25). Association rate constants (k on ) were estimated from the progress curves using published equations (20,25) and the Enzfitter (Biosoft) or Grafit (Erithacus Software) program. The effect of ionic strength was analyzed by increasing the concentration of NaCl in the standard buffer from 0.1 M to 0.25 M and 0.5 M. For competition assays, various concentrations of (⌬1-414)gelatinase A or proE375A-gelatinase A were added to the cuvette with the gelatinase A before the addition of TIMP-2 or TIMP-3. Because the K i values for the TIMP:gelatinase A interaction are unknown, the K i and the K d for the TIMP:competitor interaction are expressed as relative values using an arbitrary value of 1 for the K i . The relationship between the two dissociation constants is given in Equation 1: in which EI and E f are the TIMP:gelatinase A complex and free gelatinase A, respectively, whereas FI and F f are the TIMP:competitor complex and free competitor, respectively. Equation 1 can be rewritten as: in which F t , E t , and I t are total reagent concentrations. In our assays, F t Ͼ ϾFI, and I f is negligible, so Equation 2 can be simplified to Equation 3, from which the relative K d can be readily calculated.
The effect of various polyanions on the rate of association was carried out using a constant amount of enzyme and inhibitor (concentrations similar to those used to calculate the k on values listed in Table I) with increasing concentrations of each test polyanion in the fluorometry buffer.
Binding of TIMP-3 to Progelatinases-TIMP-3 was incubated in the presence or absence of progelatinases in TCABN for 1-2 h at 25°C. Complexes with progelatinases were isolated on gelatin-Sepharose that had been blocked with 0.2 mg/ml bovine serum albumin in TCABN. The column was washed with TCABN, and bound material was eluted with TCABN containing 15% dimethyl sulfoxide. Eluates were analyzed by rabbit collagenase diffuse collagen fibril assays (39) and reverse zymography (40).
Binding of TIMPs to Heparin-Agarose-Approximately 1 g of each TIMP was applied to heparin-agarose (blocked with 0.2 mg/ml bovine serum albumin) in TCABN buffer. Columns were washed with TCABN, and proteins were eluted stepwise with the same buffer containing 0.5 M NaCl and then 2 M NaCl. Bound and unbound fractions were analyzed for TIMP content by SDS-polyacrylamide gel electrophoresis and silver staining and by rabbit collagenase diffuse collagen fibril assay (39).
Deglycosylation of TIMPs-5 g of TIMP-3 or TIMP-1 were incubated for 4 h at 37°C in the presence or absence of 1250 units of PNGase F (New England Biolabs). TIMPs were diluted in fluorometry buffer and used in assays as above.

RESULTS
We analyzed the inhibition of active gelatinase A, gelatinase B, stromelysin-1, and matrilysin by TIMP-3 using continuous fluorometric assays with the appropriate fluorescent peptide substrate (see "Experimental Procedures"). As discussed previously for TIMP-1 and TIMP-2 (20, 25), we were unable to obtain accurate values of K i (Ͻ200 pM). Our measurements were therefore limited to the association rate constants (k on ) at low reagent concentrations, over a range where the observed rate was linear with TIMP concentration. In Table I, the data are compared with k on values for TIMP-2 that were re-assayed at the same time and k on values for TIMP-1 derived from our previous work (25,26). All three TIMPs bound relatively slowly to stromelysin-1 and matrilysin. In general, we found that TIMP-3 was more like TIMP-2 than TIMP-1, showing rapid binding to gelatinase A and slower association with gelatinase B. The contribution of the C-terminal domains of gelatinase A and gelatinase B to TIMP-3 binding was assessed by measuring the association rate of the isolated catalytic domains, (⌬418 -631)gelatinase A and (⌬426 -688)gelatinase B. Whereas TIMP-2 binding was only affected by the loss of the gelatinase A C-terminal domain, TIMP-3 association was slower in the absence of the C-terminal domains of both gelatinase A and gelatinase B (1400-fold and 12.5-fold, respectively).
The effect of ionic strength on the rate of association of gelatinase A and TIMP-3 was analyzed at increasing NaCl concentrations. Similar to TIMP-2 (20), there was a marked decrease in k on from 16.0 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 in 0.1 M NaCl to 9.6 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 (0.25 M NaCl) and 7.3 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 (0.5 M NaCl), suggesting that ionic interactions are involved in the association of gelatinase A and TIMP-3.
The contribution of the C-terminal domain of gelatinase A to TIMP-3 binding was assessed by measuring the effect of adding increasing amounts of (⌬1-414)gelatinase A (the isolated Cterminal domain) or proE375A-gelatinase A (an inactive form of progelatinase A) to the inhibition assay and observing the effect on the association rate for active full-length gelatinase A. The effect on inhibition by TIMP-2 was also measured for matrix metalloproteinases Association rate constants (k on ) were estimated from the inhibition progress curves using equations described previously (20,25). The data for TIMP-1 were taken from our previous work [25,26]. comparison. The increase in the final steady-state velocity and the decreased rate of inhibition observed with increasing concentrations of (⌬1-414)gelatinase A and proE375A-gelatinase A were deduced to be due to an effective decrease in TIMP-3 concentration by binding to the C-terminal domain, as was seen for TIMP-2 (20). The data were analyzed as described under "Experimental Procedures" to obtain an estimate for K d , the dissociation constant, relative to the K i for the appropriate TIMP:gelatinase A interaction (Table II). The interaction of TIMP-3 with (1-414)gelatinase A was significant but was around 16-fold weaker than the interaction of TIMP-2. The interaction between TIMP-3 and proE375A-gelatinase A was about five times weaker than that for TIMP-2. In both cases, the interaction of the TIMPs with proE375A-gelatinase A was stronger than that with the isolated C-terminal domain, which suggests that additional sites of interaction exist in the proenzyme-TIMP complex.
To further characterize the region of gelatinase A responsible for the C-terminal domain interaction, we used two C-terminal domain mutants: regions of the C-terminal domain of gelatinase A were exchanged for the corresponding regions of the C-terminal domain of stromelysin-1, which does not interact significantly with the TIMPs (25). As was the case for TIMP-2 (36), replacement of residues 418 -474 in N-G.C-SGG did not affect the rate of association with TIMP-3 (k on ϭ 17.0 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 , compared with 16.5 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 for gelatinase A). However, the additional substitution of residues 568 -631 in N-G.C-SGS reduced the rate of association of TIMP-3 with gelatinase A by a factor of 100 to 0.1 ϫ 10 6 M Ϫ1 ⅐s Ϫ1 , suggesting that residues 568 -631 of gelatinase A are crucial for the interaction with TIMP-3.
Because the kinetic data suggested that TIMP-3 has significant interactions with the hemopexin domains of gelatinase A and gelatinase B, we assessed the ability of TIMP-3 to bind to various pro form constructs of gelatinases A and B, in which normal catalytic domain interactions are precluded due to the presence of the propeptide domain (Table III). A small amount of TIMP-3 alone bound to the gelatin-Sepharose matrix. Enhanced retention of TIMP-3 was observed after preincubation with progelatinase A or progelatinase B, suggesting that TIMP-3 shows significant binding to both proenzymes. TIMP-3 was recovered in the unbound fraction after incubation with pro(⌬418 -631)gelatinase A or pro(⌬426 -688)gelatinase B. TIMP-3 bound to gelatin-Sepharose after preincubation with proN-G.C-SGG but did not bind if proN-G.C-SGS or proN-GL.C-SL were used. TIMP-2 was retained on the gelatin-Sepharose after incubation with progelatinase A but not after incubation with progelatinase B.
The Effect of Heparin on the Rate of Association of TIMPs with Gelatinase A-Increasing concentrations of heparin in the fluorometry buffer reproducibly resulted in a bell-shaped distribution for the association rate of TIMP-3 with gelatinase A (Fig. 1a). As the heparin concentration was increased to 100 g/ml, the association rate increased 3.7-fold compared with the k on measured in the absence of heparin. Further increases in the amount of heparin resulted in a decrease in the rate of association to levels approaching that observed in the absence of heparin. The addition of heparin to TIMP-2 and gelatinase A had a negligible effect on the association rate. The association rate of TIMP-1 and gelatinase A was increased by 4.6-fold with 800 g/ml heparin, but the distribution was not bell-shaped, as it was for TIMP-3. Although TIMP-1 appears to be more dramatically affected than TIMP-3 due to the manner in which the data is presented, the k on for TIMP-3 increased to 10 8 M Ϫ1 ⅐s Ϫ1 and exceeds the maximum rate accurately measurable using this system, whereas values for TIMP-1 plateaued at around 2 ϫ 10 7 M Ϫ1 ⅐s Ϫ1 . Preincubation of either TIMP-3 or gelatinase A with heparin or the addition of heparin to the buffer did not affect the association rate obtained. Increasing amounts of heparin did not affect the association rate of (⌬418 -631)gelatinase A and TIMP-3 (data not shown). SDS-polyacrylamide gel electrophoresis and silver staining (data not shown) and a rabbit collagenase diffuse collagen fibril assay revealed that TIMP-1 and TIMP-2 did not bind at all to heparin-agarose in 0.15 M NaCl, whereas TIMP-3 did bind, and 95% was eluted by 0.5 M NaCl, and 5% was eluted by 2 M NaCl.
The TIMPs are differentially glycosylated by our NS0 cell expression system: TIMP-2 is nonglycosylated, TIMP-1 is glycosylated, and TIMP-3 is produced in glycosylated and nonglycosylated forms. The potential role of glycosylation in binding to the polyanions was investigated by comparing the effect of heparin on the inhibition of gelatinase A by TIMP-1 and TIMP-3 in their glycosylated and deglycosylated forms. After treatment of TIMP-1 and TIMP-3 with PNGase F, which cleaves off the carbohydrate at its link with asparagine, there was a decrease in the apparent molecular weight of both TIMP-3 and TIMP-1, giving a distinct band on a silver-stained 12% polyacrylamide gel, but no decrease in apparent molecular weight where the inhibitors were incubated under the same conditions without PNGase F (data not shown). This suggests that the carbohydrate had been removed. Using the collagenase fibril assay, we found that 98% of both glycosylated and deglycosylated TIMP-3 bound heparin-agarose in 0.15 M NaCl and both were eluted by 0.5 M NaCl, whereas neither form of TIMP-1 bound significantly. In the fluorimetric assay, PNGase F had no activity against Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 , and neither gelatinase A activity nor the rate of inhibition of gelatinase A by TIMP-1 was affected by the addition of PNGase F (data not shown). Deglycosylation of TIMP-1 and

)gelatinase A or proE375A-gelatinase A to TIMP-3 and TIMP-2
Inhibition assays were carried out at 25°C with increasing amounts of either (⌬1-414)gelatinase A or proE375A-gelatinase A. The dissociation constant (K d ) was estimated as described under "Experimental Procedures" and is given as a relative value compared to that of the appropriate TIMP:gelatinase A interaction. n is the number of assays carried out.

Analysis of complex formation between TIMP-3 and progelatinases A
and B using gelatin-Sepharose Each proenzyme was incubated for 1-2 h at 25°C in an approximate 1:1 molar ratio with 1 g of TIMP-2 or TIMP-3. Complexes of TIMP bound to proenzyme were isolated on gelatin-Sepharose that had been blocked with bovine serum albumin. TIMP activity in the bound and unbound fractions was measured in a rabbit collagenase diffuse fibril assay (39), and values are expressed as the percentage of total activity recovered for each incubation. TIMP-3 had no effect on the rate of inhibition of gelatinase A in either the presence or absence of heparin (data not shown).
Hence, it appears that the carbohydrate component of TIMP-1 and TIMP-3 is not responsible for the effect seen with heparin.
To confirm that the effect of heparin is mediated by ionic interactions, the ionic strength of the fluorimetry buffer was increased, and the association rate of TIMP-3 and gelatinase A was measured. Increasing the NaCl concentration from 0.1 M to 0.25 M or 0.5 M in the presence of 10 g/ml heparin abolished the effect of heparin on the association rate: in 0.1 M NaCl, heparin increased the k on 1.4-fold, whereas in 0.25 M NaCl and 0.5 M NaCl, the k on was identical in the presence and absence of heparin. TIMP-3 was also eluted from heparin-agarose by 0.5 M NaCl. Raising the ionic strength had an identical effect on deglycosylated TIMP-3 in the presence or absence heparin (data not shown). This indicates that the effect of heparin is mediated by ionic interactions, probably between its negatively charged sulfate groups and the positively charged residues in TIMP-3 and gelatinase A.
The Effect of Other Polyanions on the Inhibition of Gelatinase A by TIMP-3-The effect of various polyanions on the rate of association of TIMP-3 with gelatinase A was tested using the standard fluorometric assay. Like heparin, dextran sulfate resulted in a bell-shaped distribution for the association rate over the concentration range studied, with an increase in k on of 4.4-fold at 50 g/ml dextran sulfate (Fig. 1b). Heparan sulfate resulted in a slight increase in the association rate over the relatively small concentration range studied (Fig. 1b). Hyaluronic acid and dermatan sulfate had no effect, although the former did result in an increase in the steady-state rate, probably due to increasing viscosity (data not shown). There was no effect on the rate of association of TIMP-3 and gelatinase A when de-N-sulfated heparin was used (Fig. 1c).
The Effect of Heparin on the Rate of Association of TIMP-3 and Other MMPs-The association rate of TIMP-3 and stromelysin-1 was not affected by heparin over the concentration range of 0 -800 g/ml (data not shown), probably because stromelysin does not bind to heparin (34). The rate of interaction of TIMP-3 with matrilysin was increased 2-fold by heparin, but the rate did not decrease with high heparin concentrations, as it did for gelatinase A, and de-N-sulfated heparin also increased the association rate slightly (Fig. 2a). There was also a slight increase in the rate of association with heparan sulfate, hyaluronic acid (as well as an increase in the steady-state rate as for gelatinase A), and dermatan sulfate (data not shown). Dextran sulfate increased the association rate 15-fold, and the distribution was bell-shaped, as it was for gelatinase A (Fig.  2b). However, the pattern of these results differed from those of TIMP-3 and gelatinase A, suggesting a different mode of action for the effect on TIMP-3 and matrilysin.

DISCUSSION
Our data show that TIMP-3 is able to associate with the MMPs stromelysin-1 and matrilysin with rates similar to TIMP-1 and TIMP-2: the association rate is relatively slow (10 5 M Ϫ1 ⅐s Ϫ1 ), presumably because stromelysin-1 has negligible Cterminal domain interactions with TIMPs (17,25), and matrilysin comprises solely a catalytic domain. The interaction of TIMP-3 with active gelatinase A (10 7 M Ϫ1 ⅐s Ϫ1 ) and gelatinase B (10 5 M Ϫ1 ⅐s Ϫ1 ) is more similar to that of TIMP-2. The rate of association of TIMP-3 with these gelatinases is enhanced by the hemopexin domains of the enzymes. The apparent K d data suggest that significant interactions occur between the C-terminal domain of gelatinase A and TIMP-3. The interaction between TIMP-3 and the C-terminal domain of gelatinase A is slightly weaker than that of TIMP-2, probably due to the absence of the highly negatively charged C-terminal tail in TIMP-3 that is present in TIMP-2 (these last 8 residues of TIMP-2 have been shown to be highly significant in the interaction with the C-terminal domain of gelatinase A (20)), but serves to increase the rate of association of TIMP-3 and gelatinase A 1000-fold. As for TIMP-2 (36), C-terminal domain interactions with residues 568 -631 are particularly important for rapid association of TIMP-3 and gelatinase A. The decrease in the rate of inhibition of gelatinase A by TIMP-3 with increasing ionic strength suggests the involvement of charged residues in the interaction, as seen for TIMP-2 (20). We also reported previously that TIMP-3 inhibition of the catalytic domains of MT1 MMP and MT2 MMP was similar to that of TIMP-2 (41,42). It is known that TIMP-2 and TIMP-4 bind to progelatinase A via C-terminal domain interactions (20,43). Here we demonstrate that TIMP-3 is also able bind to progelatinase A. Complex formation between TIMP-3 and progelatinase A involves C-terminal domain interactions: the binding of progelatinase A to TIMP-3 was reduced by removal of the hemopexin domain or by replacement with the C-terminal domain of stromelysin-1. As with the active enzyme, residues 568 -631 but not residues 418 -474 of the hemopexin domain play an important role in the association of progelatinase A and TIMP-3. These residues constitute part of blade 3 and the whole of blade 4 in the four-bladed propeller structure determined by x-ray crystallography (44,45) and border a surface patch of lysine residues (residues 566, 567, and 568) that may be important for the electrostatic interaction. This region was also important for the association of gelatinase A with TIMP-2 (36) and suggests that TIMP-2 and TIMP-3 share common features of the binding site for progelatinase A. Although TIMP-3 is able to bind progelatinase A and MT1 MMP (41) like TIMP-2, we have been unable to convincingly demonstrate its involvement in progelatinase A activation as we did for TIMP-2 (36). It is unclear whether this is due to technical difficulties caused by the adherent nature of TIMP-3 or an alternative pericellular activation pathway involving TIMP-3 bound to the matrix or whether the binding of TIMP-3 to progelatinase A is not strong enough to support the formation of a membrane receptor (36).
Removal of the C-terminal domain of gelatinase B significantly reduced the rate of inhibition of this enzyme by TIMP-3. Hence, as for TIMP-1 (26), the hemopexin domain of gelatinase B is important for association with TIMP-3, although it contributes little to the association with TIMP-2. Like TIMP-1 (26), TIMP-3 can also bind to progelatinase B, but not to pro(⌬426 -688)gelatinase B, indicating that these C-terminal domain interactions are sufficient and necessary to yield a stable proenzyme-inhibitor complex. The precise biological role of this property of the TIMPs is not yet known, although a role in progelatinase B activation is possible.
The assays of TIMP-3 inhibitory activity described above were carried out in solution. Although these studies are valuable from a comparative point of view, it must also be borne in mind that TIMP-3 is apparently largely extracellular matrixbound in vivo, although the components to which it binds remain to be determined. Proteoglycans consist of core proteins with numerous attached glycosaminoglycan chains: the latter are negatively charged polysaccharides composed of repeating disaccharides (for reviews, see Refs. 46 and 47). TIMP-3 possesses 9 positively charged residues (8 lysines and 1 arginine) that are not present in TIMP-1 or TIMP-2 (see alignment, Fig.  3 in Ref. 3) and that are generally conserved in TIMP-3 from different species. It is likely that these charged residues may be involved in the interaction of TIMP-3 with cell surface or extracellular matrix glycosaminoglycans. We therefore tested the effect of some commercially available polyanions on the inhibition of various MMPs by TIMP-3.
The rate of inhibition of gelatinase A but not (⌬418 -631)gelatinase A by TIMP-3 was increased by heparin. Both gelatinase A and TIMP-3 bind to heparin, but there is no heparin binding site in (⌬418 -631)gelatinase A (48), suggesting that a heparin binding site is required in both interacting proteins. The bell-shaped distribution of the association rate over the heparin concentration range studied is reminiscent of similar curves for the effect of heparin on progelatinase A autoactivation (48) or for the inhibition of thrombin by antithrombin III (49). The curve suggests a biomolecular mode of binding to heparin that increases the local concentration of reactants, thereby increasing their rate of interaction, rather than a conformational effect.
The effects of the polyanions tested on the rate of inhibition of gelatinase A by TIMP-3 appear to correlate with negative charge density. The sulfated compounds, dextran sulfate, heparin, and heparan sulfate (4 -5 O-linked, 2-3 O-and N-linked and 1 O-or N-linked sulfate per disaccharide, respectively (50)), enhanced the rate of interaction, whereas dermatan sulfate (1 O-linked sulfate per disaccharide), de-N-sulfated heparin, and hyaluronic acid (unsulfated) had no effect, suggesting that the interaction of enzyme and inhibitor with these polyanions is based on charge density as well as structure. A specific recognition domain in heparin has been described for basic fibroblast growth factor and antithrombin (51). The existence of such a domain for TIMP-3 would be compatible with a surface concentration mechanism like that of antithrombin and thrombin (51). TIMP-3 does not contain any of the reported linear heparin binding motifs, but a motif defined by the threedimensional structure could exist (52). It is likely that TIMP-3 interacts with cell surface and extracellular matrix glycosaminoglycans via the large number of positively charged residues in TIMP-3, and that this is the basis for its location in the extracellular matrix both in vivo and in cell culture. Hence, colocalization of TIMP-3 with proenzymes in the pericellular environment may be a mechanism for increasing the rate of inhibition of MMPs and regulating extracellular matrix breakdown during morphogenetic processes.
Heparan sulfate proteoglycans such as perlecan (53) and syndecans (54) are also implicated in binding growth factors that promote angiogenesis. A recent study demonstrated that TIMP-3 can inhibit endothelial cell migration and angiogenesis in response to the angiogenic factors basic fibroblast growth factor and vascular endothelial growth factor (55). TIMP-2 had similar effects upon endothelial cell migration in vitro, but TIMP-1 was ineffective (55,56). This implicates MT1 MMP in the angiogenic process, an enzyme that can degrade matrix components and initiate the autoactivation of gelatinase A (36, 40, 57). The study presented here suggests that the effects of TIMP-3 and TIMP-2 might be due to the specific ability of these inhibitors to bind to progelatinase A as well as to inhibit MT1 MMP. Colocalization of TIMP-3 in the pericellular environment via binding to the extracellular matrix, including heparan sulfate proteoglycans, would place this inhibitor in a key position to inhibit MMPs produced by endothelial cells, thus regulating degradation of the extracellular matrix and release of the angiogenic factors required for migration and angiogenesis.