Dynamic Interdomain Interactions Contribute to the Inhibition of Matrix Metalloproteinases by Tissue Inhibitors of Metalloproteinases*

Because of their important function, matrix metalloproteinases (MMPs) are promising drug targets in multiple diseases, including malignancies. The structure of MMPs includes a catalytic domain, a hinge, and a hemopexin domain (PEX), which are followed by a transmembrane and cytoplasmic tail domains or by a glycosylphosphatidylinositol linker in membrane-type MMPs (MT-MMPs). TIMPs-1, -2, -3, and -4 are potent natural regulators of the MMP activity. These are the inhibitory N-terminal and the non-inhibitory C-terminal structural domains in TIMPs. Based on our structural modeling, we hypothesized that steric clashes exist between the non-inhibitory C-terminal domain of TIMPs and the PEX of MMPs. Conversely, a certain mobility of the PEX relative to the catalytic domain is required to avoid these obstacles. Because of its exceedingly poor association constant and, in contrast with TIMP-2, TIMP-1 is inefficient against MT1-MMP. We specifically selected an MT1-MMP·TIMP-1 pair to test our hypothesis, because any improvement of the inhibitory potency would be readily recorded. We characterized the domain-swapped MT1-MMP chimeras in which the PEX of MMP-2 (that forms a complex with TIMP-2) and of MMP-9 (that forms a complex with TIMP-1) replaced the original PEX in the MT1-MMP structure. In contrast with the wild-type MT1-MMP, the diverse proteolytic activities of the swapped-PEX chimeras were then inhibited by both TIMP-1 and TIMP-2. Overall, our studies suggest that the structural parameters of both domains of TIMPs have to be taken into account for their re-engineering to harness the therapeutic in vivo potential of the novel TIMP-based MMP antagonists with constrained selectivity.

MMPs are synthesized as zymogens, which require the proteolytic processing of the N-terminal inhibitory PRO to generate the active enzymes (4). It is accepted that secretory tissue inhibitors of MMPs (TIMPs) play an important role in the regulation of the proteolytic activity of MMPs (6,7). Four TIMPs (TIMP-1, -2, -3, and -4) are present in humans (8). There is at least a 25% sequence identity among all TIMPs, including 12 conserved Cys residues that form 6 disulfide bridges resulting in 6 loop regions. There are two domains, N-terminal and C-terminal, in TIMPs. An N-terminal domain (NT-TIMP) binds the CAT, carries the MMP-inhibitory activity, and encompasses the three first loops. A non-inhibitory C-terminal domain (CT-TIMP) binds, albeit with distinct affinities, the PEX of MMPs. TIMP-1 forms a stoichiometric complex with the MMP-9 proenzyme via the binding of its CT-TIMP with the PEX (9). In turn, TIMP-2 forms a complex with the MMP-2 proenzyme (10,11).
Multiple studies have been performed to constrain the specificity of TIMPs and transform TIMPs into selective rather than wide-ranging inhibitors (14,(23)(24)(25)(26). The inhibitory NT-TIMP alone and the individual CAT of MMPs were predominantly used in these studies. Differences in contacts and chemistry of the interfaces of TIMP⅐MMP complexes provide a basis for re-* This work was supported, in whole or in part, by National Institutes of Health Grants CA77470 and CA83017 (to A. Y. S.). engineering of the inhibitory selectivity. Because of its exceedingly poor association constant, TIMP-1 is incapable of efficiently inhibiting the full-length MT1-MMP enzyme, albeit a certain level of inhibition was observed with the individual catalytic domain of MT1-MMP (21,23,24). A single point mutation of Thr 98 to Leu (T98L) at a distal site from the inhibitory loop, however, increased the association constant and transformed the NT-TIMP of TIMP-1 into a tight-binding inhibitor of MT1-MMP (14,27). There are, however, additional important structural elements, which are outside of the inhibitory and catalytic domains of the inhibitor and the proteinase, respectively. Thus, the contacts of the CT-TIMP of TIMP-1 and TIMP-3 are required with the C-terminal, non-catalytic domain to achieve an efficient inhibition of a disintegrin and metalloprotease-10 (ADAM10). The NT-TIMP alone is insufficient for the inhibition of ADAM10 (28). Conversely, the full-length TIMP-4 is a week inhibitor of ADAM17 (tumor necrosis factor-alpha converting enzyme; TACE), whereas the C-terminal truncation significantly increases the inhibitor's potency (29). Overall, both the CT-TIMP and its counterpart, the C-terminal noncatalytic PEX in MMPs, play a likely important, albeit unidentified, role in the mechanisms of inhibition of MMPs by TIMPs.
To highlight the effect of the PEX on the interactions of MT1-MMP with the full-length TIMPs, we designed and characterized the MT1-MMP mutants with the truncated domains and also the MT1-MMP chimeras. In these chimeras the original PEX was substituted in the MT1-MMP molecule by the PEX derived from MMP-2 and MMP-9. We specifically selected MMP-2 and MMP-9 because of their contrasting proenzyme complex formations with TIMP-2 and TIMP-1, respectively. In addition, we specifically selected TIMP-1 for our experiments because of its exceedingly low ability to interact with MT1-MMP and because any improvement of the affinity of TIMP-1 would be readily recorded. Based on our experiments, we now suggest that the interactions between the PEX of the full-length MT1-MMP and the CT-TIMP also contribute to the inhibitory efficacy of TIMPs. As it appears now, the global fold of the PEX, the interdomain dynamics, and the intrinsic protein flexibility of both MMPs and TIMPs play a role in defining the binding interface in the course of inhibition of the fulllength cellular MMPs by the full-length TIMPs. From the practical perspectives, it is likely that constraining the specificity of TIMPs toward the individual full-length MMPs also requires the modification in the non-inhibitory CT-TIMP and that the modifications in the inhibitory NT-TIMP alone are insufficient for the re-design of the inhibitor specificity.

MATERIALS AND METHODS
General Reagents and Antibodies-All reagents are purchased from Sigma-Aldrich unless indicated otherwise. A murine monoclonal antibody against the CAT of MT1-MMP (clone 3G4) and a hydroxamate inhibitor of MMPs (GM6001) were purchased from Chemicon. A SuperSignal West Dura Extended Duration Substrate kit, EZ-Link sulfo-NHS-SS-biotin (sulfosuccinimidyl-2-(biotinamido) ethyl-1,3-dithiopropionate), and EZ-Link sulfo-NHS-LC-biotin (sulfosuccinimidyl-6-(biotinamido)hexanoate) were from Pierce. The secondary species-specific antibodies conjugated with HRP were purchased from Jackson ImmunoResearch. The secondary speciesspecific antibodies conjugated with Alexa Fluor 594 (red) were obtained from Molecular Probes. A proteinase inhibitor mixture set (Set III) was from Calbiochem.
Recombinant Proteins-Recombinant human TIMP-1 was obtained from Invitrogen. Recombinant TIMP-2 was purified from the medium conditioned by the recombinant CHO cells (30). The secretory N-end-appended TIMP-2 mutant (A-TIMP-2) that exhibited the Glu-Ala-Glu-Ala-Tyr-Val-Glu-Phe sequence attached to the N-terminal Cys 1 was purified from the recombinant Pichia pastoris yeasts using FPLC on a Mono-Q column (31). The TIMP-2-free MMP-2 proenzyme was isolated from p2AHT2A72 cells derived from the fibrosarcoma HT1080 cell line sequentially transfected with the E1A and MMP-2 cDNAs (32). The individual CAT of MT1-MMP and MT6-MMP was expressed in Escherichia coli, purified from the inclusion bodies, and refolded to restore its catalytic activity (33,34). The concentrations of the purified, catalytically active MT1-MMP and MT6-MMP were measured by titration against a standard GM6001 solution of a known concentration and Mca-PLGL-Dpa-AR-NH 2 as a fluorescent peptide substrate (35).
Cloning of the MT1-MMP Chimeras-To design both chimeras, we used the cDNA coding for the full-length human wildtype MT1-MMP (WT), MMP-2, and MMP-9 cDNAs. The sequences coding for the PEX of MMP-2-(470 -659) and MMP-9-(517-703) were each inserted between Cys 319 and Cys 508 of the MT1-MMP sequence to generate the PEX/ MMP-2 and PEX/MMP-9 chimeras, respectively. The authenticity of the recombinant constructs was confirmed by DNA sequencing.
In addition, to quantitatively assess the inhibitory potency of TIMP-1 and TIMP-2 against the WT and PEX/MMP-9 constructs, we employed MCF7-␤3 cells transiently transfected with the pcDNA3.1-zeo plasmid encoding for WT and PEX/ MMP-9. The cells were transfected with the recombinant plasmids using Lipofectamine LTX and the Plus reagent (Invitrogen) according to the manufacturer's protocol. The cells were used in our inhibitory experiment in 48 h post-transfection.
For the uptake experiments, cells were surface-biotinylated using the cleavable membrane-impermeable EZ-Link sulfo-NHS-SS-biotin. Immediately following the biotinylation procedure, cells were incubated for 25 min at 37°C in serum-free DMEM supplemented with 1% insulin-transferrin-selenium to allow the internalization of biotin-labeled MT1-MMP (41,42). To remove the residual cell surface biotin, cells were incubated for 25 min on ice in Sorensen phosphate buffer containing membrane-impermeable MESNA (150 mM). Cells were next extensively washed using Sorensen phosphate buffer and lysed, and the lysates were precipitated using streptavidin beads and analyzed as above.
Enzymatic Assay-MMP activity was measured in triplicate in wells of a 96-well plate in 0.2 ml of 50 mM HEPES, pH 7.5, containing 10 mM CaCl 2 and 50 M ZnCl 2 . Mca-PLGL-Dpa-AR-NH 2 (10 M) was used as a fluorescent substrate. The concentration of MT1-MMP and MT6-MMP in the reactions was 5 nM. The steady-state rate of substrate hydrolysis was monitored continuously ( ex ϭ 320 nm and em ϭ 400 nm) at 37°C for 3-25 min using a fluorescent spectrophotometer. Where indicated, TIMP-1 (25-125 nM) and TIMP-2 (25-125 nM) were co-incubated for 30 min at 20°C with the MMP samples prior to adding the substrate.
Immunostaining of Cells-Cells grown on 15-mm glass coverslips were fixed for 20 min with 4% formaldehyde. Where indicated, cells were permeabilized for 4 min using 0.1% Triton X-100 or left untreated. Cells were then blocked for 1 h at ambient temperature using 10% BSA in PBS and then stained overnight at 4°C with the MT1-MMP 3G4 antibody (dilution 1:1000) or the polyclonal rabbit MT1-MMP AB815 antibody (dilution 1:200) followed by a 1-h incubation with the secondary species-specific antibody (dilution 1:200) conjugated with Alexa Fluor 594. The slides were mounted in the Vectashield medium containing DAPI for the nuclear staining. The slides were analyzed using an Olympus BX51 fluorescence microscope equipped with a MagnaFire digital camera.
In Situ Gelatin Zymography Using FITC-gelatin-FITC-gelatin was prepared as described earlier (33). Cells (1 ϫ 10 4 ) were seeded onto the gelatin-coated coverslips and incubated for 16 h at 37°C in serum-free DMEM supplemented with TIMP-1 (100 nM), TIMP-2 (100 nM), or GM6001 (50 M). The cells were then fixed with 4% formaldehyde for 16 min, permeabilized for 4 min using 0.1% Triton X-100, and stained for MT1-MMP as described above. The dark regions of degraded FITC-gelatin can be readily detected using a fluorescent microscope.
Structural Modeling-The structural coordinates of the porcine full-length MMP-1 enzyme complexed with a specific inhibitor N-  (12), the individual CAT of MT1-MMP complexed with the C-terminally truncated, V4A, P6V, and T98L triple mutant TIMP-1 (PDB 3MA2) (14,23), the individual PEX of human MMP-2 (PDB 1RTG) (46), and the individual PEX of MT1-MMP (PDB 3C7X) were used in our study. The structures were analyzed and superimposed using PyMOL. The images were prepared also in PyMOL.

TIMP⅐MMP Inhibitory
Complex-Like all of the other members of the MMP family, the active MT1-MMP proteinase is regulated by TIMPs. The association constant of TIMP-1 with

Inhibition of MT1-MMP by TIMP-1 and TIMP-2
the MT1-MMP enzyme, however, is exceedingly poor. This association constant is significantly less efficient compared with those of TIMP-2 and TIMP-3 (23,24). As a result, TIMP-1 is not capable of inhibiting the cellular MT1-MMP activity, especially under a physiological range of inhibitor concentrations and especially if compared with TIMP-2 (8,21). Our work and multiple publications by others suggested that the PEX is involved in the MMP homodimerization and the association of the individual MMPs with TIMPs as well as in the interactions of MMPs with their cleavage substrates (9,10,31,(47)(48)(49)(50)(51)(52)(53). Based on this general assumption, we hypothesized that the structure of the PEX also contributes to the inhibitory interactions of active MMPs with TIMPs.
To support our hypothesis, we performed a thorough superimposition analysis of the available MMP and TIMP structures, including PDB 1FBL, 1SU3, 1GXD, 1CK7, 1UEA, 1BUV, 3MA2, 1RTG, and 3C7X. Thus, the overall fold of the PEX in the MMP-1 proenzyme (PDB 1SU3) is highly similar with that of the individual PEX of MMP-2 (PDB 1RTG) and MT1-MMP (PDB 3C7X). The PEX position, however, is shifted relative to the CAT in the MMP-1 enzyme (PDB 1FBL) compared with the MMP-1 proenzyme (PDB 1SU3). Similarly, there is a difference in the relative positions of the PEX and the CAT in the MMP-2 proenzyme alone (PDB 1GXD) and the MMP-2 proenzyme⅐TIMP-2 complex (PDB 1CK7). Our estimate suggests that the mobility range of the PEX relative to the CAT exceeds 10°as measured from the extreme conformational states in the MMP-1 (PDB 1FBL and 1SU3) and MMP-2 (PDB 1GXD and 1CK7) structures we analyzed (supplemental Fig.  S1). These structural differences are distinct and additional to those local differences that were observed in the interface between the PRO and the PEX (44).
In a similar fashion, the overall fold of TIMP-1 and TIMP-2 is highly similar in their respective complexes with MMP-3 (PDB 1UEA) and MT1-MMP (PDB 1BUV and 3MA2). Notably, a significant difference is in the AB loop of NT-TIMP because of the protruding six-amino acid insert in TIMP-2 relative to TIMP-1 (54). There is, however, an ϳ15°difference in the position of TIMP-1 and TIMP-2 bound to the CAT of MMP-3 (PDB 1UEA), and MT1-MMP (PDB 1BUV and 3MA2), respectively (supplemental Fig. S2).
Taken together, our analysis suggests that a certain level of motion of the PEX is required to permit an inhibitory TIMP⅐CAT complex. In the absence of this motion, the CT-TIMP collides with the PEX moiety. As a result, the NT-TIMP appears incapable of the productive inhibitory interactions with the CAT active site ( Fig. 1; supplemental Fig. S3). Naturally, molecular packing interactions in the crystal may affect the relative orientation of domains in multi-domain proteins. The mobility of the PEX relative to the CAT domains we recorded, however, significantly exceeds these parameters. To determine experimentally if the PEX affects the inhibitory characteristics of TIMPs, we performed an extensive mutagenesis of the MT1-MMP sequence followed by a characterization of the mutants, including the analysis of their sensitivity to the inhibition by TIMP-1 and TIMP-2.
Expression and Analysis of the MT1-MMP Constructs-To identify the role of the individual structural domains in the functionality of MT1-MMP, we constructed mutants in which the CAT, the PEX, and the CYTO were truncated in the MT1-MMP sequence (⌬CAT, ⌬PEX, and ⌬CYTO, respectively). We also constructed the MT1-MMP chimeras in which the PEX of MMP-2 and MMP-9 replaced the original PEX in the MT1-MMP sequence (PEX/MMP-2 and PEX/MMP-9, respectively). As controls, we used the wild-type MT1-MMP construct and the E240A mutant in which Ala substituted for the catalytically essential Glu 240 residue of the proteinase active site (Fig. 2A).
The constructs were then stably co-expressed with the ␤ 3 integrin subunit in human breast carcinoma MCF-7 cells. We
To assess the expression level and the catalytic activity of MT1-MMP, the obtained stably transfected cells were then examined by Western blotting and gelatin zymography (Fig.  2B). The 3G4 and AB815 antibodies (against the CAT and the hinge, respectively) were used in Western blotting. MT1-MMP immunoreactivity was not detected in the mock cell control transfected with the original plasmid lacking the MT1-MMP insert, whereas other cell types expressed the comparable level of MT1-MMP. Naturally, the 3G4 antibody did not detect the ⌬CAT construct and the degraded, ϳ40-kDa, MT1-MMP species, which were lacking the CAT. In turn, the degraded species were absent in the inert E240A and ⌬CAT mutants. As expected, the size of the degraded ⌬PEX construct was ϳ20 kDa lower compared with the WT construct.
The functional activity of cellular MT1-MMP was measured using the ability of cells to activate the latent MMP-2 zymogen, a direct downstream target of MT1-MMP (4,47,55). Because MCF7 cells do not synthesize MMP-2 naturally, the purified MMP-2 proenzyme was added to the cells. As expected, mock, E240A, and ⌬CAT cells did not activate MMP-2, whereas all other cell types, including ⌬PEX, PEX/MMP-2, and PEX/ MMP-9 cells, readily activated MMP-2 and transformed the latent 68-kDa zymogen into the 64-kDa intermediate and, predominantly, the 62-kDa mature enzyme of MMP-2 (Fig. 2B). These results directly suggest that the original PEX of MT1-MMP is not crucial for the MMP-2-processing function of MT1-MMP, and they agree well with the results of others (41, 56 -58).
Mutations Do Not Affect the Internalization Rate of MT1-MMP-To determine whether the mutations affected the cell surface presentation and internalization rate of MT1-MMP, we used cell immunostaining. To inactivate cellular MT1-MMP and block its self-degradation, prior to immunostaining procedures the cells were cultured in the presence of GM6001 (42). The cells were then fixed and permeabilized or left untreated. The cells were next stained with the 3G4 antibody that recognizes the CAT and, as a result, reacts with the full-length proenzyme-enzyme species but not with the degraded forms of MT1-MMP. Because the CAT was missing in the ⌬CAT construct, ⌬CAT cells were stained with the AB815 antibody against the hinge region of MT1-MMP.
In agreement with the immunoblotting (Fig. 2B), there was no MT1-MMP immunoreactivity in the mock MCF-7 cell control. Immunostaining demonstrated the presence of the MT1-MMP immunoreactivity in all MT1-MMP constructs (Fig. 3A). Cell surface-associated MT1-MMP expression was especially evident in the non-permeabilized cells, whereas the predominantly vesicular MT1-MMP immunoreactivity was observed in the permeabilized cells. In agreement with our previous observations (39) and in contrast to other constructs, the immunoreactivity pattern of the permeabilized and non-permeabilized ⌬CYTO cells was similar. Predominant association with the caveolin-enriched lipid rafts and the resulting slow internalization rate via the caveolae pathway explain this immunostaining pattern of the ⌬CYTO construct (41,59). Based on these results, we conclude that all of the MT1-MMP constructs we designed are efficiently trafficked through the cell compartment and presented on the plasma membrane of MCF-7 cells.
To corroborate this conclusion, we compared the internalization rate of cellular MT1-MMP (Fig. 3B). The cells were surface-biotinylated with membrane-impermeable, cleavable, EZ-Link NHS-SS-biotin. Biotinylation was followed by incubation of the cells at 37°C to initiate protein uptake. Cells were next transferred on ice to arrest protein trafficking and then treated with MESNA to release the biotin moiety from the residual cell surface-associated MT1-MMP molecules. The biotin-labeled internalized MT1-MMP was protected from MESNA. The biotin-labeled MT1-MMP pool was then captured on streptavidin beads and then analyzed by Western blotting using the 3G4 and AB815 antibodies. These tests demonstrated that a major portion of cell surface-associated MT1- The cells were grown in the presence of GM6001 on glass coverslips for 24 h. The cells were then fixed, permeabilized (ϩTriton X-100), or left untreated (ϪTriton X-100), and stained using the MT1-MMP 3G4 antibody raised against the CAT (red). The nuclei are DAPI-stained (blue). Bar, 10 m. The star indicates the ⌬CAT construct, which was stained using the MT1-MMP AB815 antibody raised against the hinge region. B, the uptake of MT1-MMP by the cells. Cells were incubated for 24 h in DMEM-10% FBS with GM6001 and then surface-biotinylated with cleavable EZ-Link NHS-SS biotin. The cells were next incubated for 25 min at 37°C to allow cell surface-associated MT1-MMP to be internalized. Biotin-labeled MT1-MMP was captured on streptavidin beads and examined by Western blotting with the MT1-MMP 3G4 and AB815 antibodies (top and bottom panels, respectively). Where indicated, MESNA was used to release a biotin moiety from the cell surface-associated proteins. Control, WT cells were treated as above but at 0°C to demonstrate the quantitative removal of the biotin label by MESNA.

Inhibition of MT1-MMP by TIMP-1 and TIMP-2
MMP (except the ⌬CYTO construct) was already internalized following a 25-min incubation. In contrast, only a small fraction of the ⌬CYTO construct was protected from MESNA, thus consistently suggesting that the ⌬CYTO MT1-MMP was inefficiently internalized (39,41,59).
Inhibition of MT1-MMP by TIMPs-We next evaluated the inhibitory effect of TIMP-1 and TIMP-2 on the MT1-MMPmediated activation of MMP-2. The similarly high inhibitory activity of the TIMP-1 and TIMP-2 samples we used was confirmed using the purified individual CAT of MT6-MMP and the Mca-PLGL-Dpa-AR-NH 2 peptide as a substrate (Fig. 4A). According to our earlier data, TIMP-1 and TIMP-2 were similarly potent in the inhibition of the MT6-MMP proteolytic activity (33). In agreement, a 5 molar excess of TIMP-1 or TIMP-2 over MT6-MMP was sufficient in our current tests to achieve a near complete inhibition of the proteolytic activity, thus suggesting the equal inhibitory potency of our inhibitor samples. In sharp contrast, the purified individual CAT of MT1-MMP retained its full proteolytic activity in the presence of the high, 125 nM, TIMP-1 concentrations (at a 1:25 molar ratio of MT1-MMP/TIMP-1), whereas no proteolytic activity was recorded in the presence of a 1:5 molar ratio MT1-MMP/TIMP-2.
These results are consistent with the effect of the inhibitors on the self-proteolysis of cellular MT1-MMP and on the level of the degraded forms of the cellular proteinase (Fig. 4B). As expected, GM6001 almost quantitatively repressed MT1-MMP self-proteolysis, and, as a result, only insignificant amounts of the degraded forms were generated. TIMP-2 (100 nM) also repressed, albeit less efficiently, the self-degradation of the WT, ⌬CYTO, ⌬PEX, PEX/MMP-2, and PEX/MMP-9 constructs, whereas 100 nM A-TIMP-2 was without a significant effect. The inhibitory effect of 100 nM TIMP-1 was observed only with both the PEX/MMP-2 and PEX/MMP-9 chimeric constructs. Overall, the ability of TIMP-1 to inactivate the PEX/MMP-2 and PEX/MMP-9 chimeras is contrasting relative to the resistance of the individual CAT of MT1-MMP to the similar concentrations of the inhibitor.
Quantitative Assessment of the Inhibitory Potency of TIMPs-To quantitatively assess the inhibition of MT1-MMP by TIMPs, we used MCF7-␤3 cells transiently transfected with the WT and PEX/MMP-9 constructs. Cells were co-incubated with both the MMP-2 proenzyme and the increasing concentrations of TIMP-1 or TIMP-2. Following gelatin zymography of the concentrated medium aliquots, the residual levels of the proenzyme and the generated levels of the MMP-2 intermediate were measured using the digitized gel images (Fig. 5). In agreement with multiple earlier reports (reviewed in Refs. 8,62,63), the cellular WT construct was readily inhibited by TIMP-2. In contrast, the cellular WT activity was not significantly inhibited by TIMP-1. Thus, only a 20% inhibition of MT1-MMP was observed at TIMP-1 concentrations as high as 1000 nM.
In agreement with our other results (Fig. 4), the PEX/MMP-9 construct was significantly inhibited by both TIMP-1 and TIMP-2 (Fig. 5). Thus, a near quantitative inhibition of PEX/ MMP-9 was observed at the concentrations of TIMP-1 as low as 100 nM. In turn, TIMP-2 was similarly effective against WT and PEX/MMP-9. Because TIMP-1 does not bind the PEX domain of MMP-2, our inhibitory results, especially if combined together, cannot be explained only by the non-inhibitory TIMP-1 binding with the PEX domain in the PEX/MMP-9 and PEX/MMP-2 chimeras.
TIMP-1 Inhibits Gelatin Degradation by MT1-MMP Chimeras-We next measured the ability of the cellular MT1-MMP constructs, including WT, PEX/MMP-2, and PEX/MMP-9 to degrade FITC-gelatin. Mock cells were used as a control. Where indicated, TIMP-1, TIMP-2, or GM6001 were added to the samples. In 16 h the cells were fixed, stained with the MT1-MMP antibody, and observed using a fluorescence microscope. Gelatinolytic activity was detected by the presence of the dark zones of digested gelatin on the fluorescent background of the intact FITC-gelatin (Fig. 6).
WT, PEX/MMP-2, and PEX/MMP-9 cells readily degraded FITC-gelatin, whereas mock cells were clearly negative. GM6001 and TIMP-2 each blocked gelatin degradation by WT, PEX/MMP-2, and PEX/MMP-9 cells. In turn, TIMP-1 inhibited only the PEX/MMP-2 and PEX/MMP-9 chimeras and was without any effect on the WT cells. Thus, these results suggest that not only the ability of activating MMP-2 and of self-proteolysis but also of the gelatinolytic activity of the MT1-MMP chimeras is repressed by both TIMPs.

DISCUSSION
MT1-MMP, the first characterized, archetype member of the MT-MMP family, was discovered as an MMP-2 cellular activator (47,64). Pro-tumorigenic MT1-MMP is a key proteinase in cell migration, and its inhibitors are urgently required to combat multiple diseases, including malignancies. MT1-MMP is known to be readily inhibited by TIMPs, excluding TIMP-1, whereas other MMPs are similarly sensitive to the inhibition by TIMPs, including TIMP-1, -2, -3, and -4 (8,14,21). From practical perspectives, this unique relation between TIMP-1 and MT1-MMP facilitates the discrimination of the latter from other individual MMPs.
Our structural studies provided a basis for the hypothesis that the global fold and the relative positions of the PEX and the CAT manipulate the way TIMPs, including TIMP-1 and TIMP-2, interact with MMPs, including MT1-MMP. Overall, our computational analysis suggests that a noticeable motion of the PEX relative to the CAT is required to allow an inhibitory complex with the TIMP moiety. In the absence of this motion, the loop 6 of the CT-TIMP clashes with the C-terminal region of the first N-terminal propeller blade of the PEX, including MMP-1 and MMP-2 (PDB 1FBL and 1CK7), and, most probably, also MT1-MMP. As a result, the inhibitory NT-TIMP cannot interact in a productive way with the active site of the CAT ( Fig. 1 and supplemental Fig. S3). The interdomain dynamics and the intrinsic protein flexibility of both MMPs and TIMPs seem to control their binding interface and play a role in the mechanisms involved in MMP inhibition by TIMPs (14). This suggestion agrees well with the studies by others who suggested that conformational adaptations are required to avoid obstacles for interaction between the full-length TIMP-1 and the CAT of MT1-MMP and MT3-MMP (65).
These previously underexploited structural data allowed us to hypothesize that the nature and the fold of the MMP's PEX contribute to the selectivity of MMP inhibition by TIMPs. These parameters are distinct and additional to the direct interaction of the NT-TIMP with the MMP's CAT. Conversely, we suggested that, if the natural PEX is modified in MT1-MMP, the proteolytic activity of the resulting mutant may become sensitive to TIMP-1. To test our hypothesis, we constructed the ⌬PEX, ⌬CAT, and ⌬CYTO MT1-MMP truncations and the PEX-swapped constructs. In the latter, the PEX of MMP-2 and MMP-9 replaced the original PEX in the MT1-MMP structure. The wild-type and the catalytically inert E240A constructs of MT1-MMP were used as controls.
We then tested the functionality of the MT1-MMP constructs. These tests included the level of the MT1-MMP presentation on the cell surface and the rate of internaliza- tion inside the cells. In addition, we measured the ability of the constructs we designed to activate MMP-2, to self-degrade, and to hydrolyze gelatin in situ and also their response to TIMP-1, TIMP-2, and appended A-TIMP-2 with the inactivated inhibitory NT-TIMP (60). A broad-range hydroxamate inhibitor, GM6001, was used as a control in our inhibitory tests.
We determined that, in contrast to other MT1-MMP constructs, the PEX/MMP-2 and PEX/MMP-9 chimeras became sensitive to the inhibition by both TIMP-1 and TIMP-2. These results provide evidence that the presence of these PEX moieties in the MT1-MMP structure, but not the original PEX, allows both TIMP-1 and TIMP-2 to interact with the active site in the CAT of MT1-MMP.
It is possible that in the chimeras the unnatural PEX stabilizes the interactions of the otherwise weak TIMP-1 inhibitor with the CAT of MT1-MMP. The effects of the PEX are not as prominent for TIMP-2 because of its intrinsic high affinity to the MT1-MMP's CAT. Overall, it becomes increasingly clear that there is an interplay between the CT-TIMP and the PEX in the inhibitory mechanism of MMPs by TIMPs.
The structural parameters that are involved in these dynamic interactions are not precisely clear as yet. Nevertheless, our biochemical studies suggest that constraining the TIMP specificity appears even more challenging than before and that the structural parameters of the PEX of MMPs should be taken into account for TIMP re-engineering to harness the therapeutic potential of new MMP antagonists with constrained selectivity. The use of the model systems involving the inhibitory NT-TIMP alone and the CAT of the individual MMPs may not satisfy the criteria that are required for the efficient and selective inhibition of the full-length MMPs in vivo.