Unveiling the Surface Epitopes That Render Tissue Inhibitor of Metalloproteinase-1 Inactive against Membrane Type 1-Matrix Metalloproteinase*

Membrane type 1-matrix metalloproteinase (MT1-MMP) is a zinc-dependent, membrane-associated endoproteinase of the metzincin family. The enzyme regulates extracellular matrix remodeling and is capable of cleaving a wide variety of transmembrane proteins. The enzymatic activity of MT1-MMP is regulated by endogenous inhibitors, the tissue inhibitor of metalloproteinases (TIMP). To date, four variants of mammalian TIMP have been identified. Whereas TIMP-2-4 are potent inhibitors against MT1-MMP, TIMP-1 displays negligible inhibitory activity against the enzyme. The rationale for such selectivity is hitherto unknown. Here we identify the surface epitopes that render TIMP-1 inactive against MT1-MMP. We show that TIMP-1 can be transformed into an active inhibitor against MT1-MMP by the mutation of a single residue, namely threonine 98 to leucine (T98L). The resultant mutant displayed inhibitory characteristics of a typical slow, tight binding inhibitor. The potency of the mutant could be further enhanced by the introduction of valine 4 to alanine (V4A) and proline 6 to valine (P6V) mutations. Indeed, the inhibitory profile of the triple mutant (V4A/P6V/T98L) is indistinguishable from those of other TIMPs. Our findings suggest that threonine 98 is critical in initiating MMP binding and complex stabilization. Our findings also provide a potential mechanistic explanation for MMP-TIMP selectivity.

Membrane type 1-matrix metalloproteinase (MT1-MMP, 1 MMP-14) is a member of the zinc-dependent endopeptidases of the metzincin family. The enzyme is involved in the degradation of extracellular matrix components and tissue remodeling (1)(2)(3). Among the six membrane-type MMPs identified, MT1-MMP is the most studied not only because of its role in activating other MMPs such as pro-MMP-2 (pro-gelatinase-A) and pro-MMP-13 (4,5), but also because of its compelling link with cell invasion and tumor malignancy (6). As with other members of the matrixin family, MT1-MMP is a multidomain enzyme with clear structural compartmentation. Preceding the cata-lytic domain is a propeptide that contains a cysteine switch that regulates the catalytic activity of the enzyme. The catalytic domain displays a tertiary fold typical of an enzyme of the metzincin family (7). Succeeding the catalytic domain are a short hinge, a hemopexin domain, a transmembrane domain, and a cytotail of 2.5 kDa. A crystal structure of the catalytic domain of MT1-MMP was first reported by Fernandez-Catalan et al. in 1998 (8). To date, this is the only known variant of membrane-associated MMP to be crystallized and published.
The activity of MMPs is regulated by their endogenous inhibitors, tissue inhibitor of metalloproteinases (TIMP). There are four mammalian TIMPs, namely TIMP-1-4. TIMP inhibit MT1-MMP by forming a tight, non-covalent 1:1 stoichiometric complex with the catalytic domain of the enzyme. The binding constants (K i app ) of wild type TIMP-2-4 with MT1-MMP are in the low picomolar range (9). Although TIMP-2-4 are superb inhibitors against MT1-MMP, TIMP-1, on the contrary, is not capable of forming a tight binding complex with MT1-MMP.
The molecular masses of TIMP vary from 21 to 28 kDa. Crystallography and protein NMR studies carried out on TIMP-1 and -2 revealed that the molecules are folded into two very discrete domains (10 -12). The N-terminal domain of the TIMPs (N-TIMP) is ϳ14 kDa and encompasses the first twothirds of the molecule. The architecture of N-TIMP is typical of a ␤-barrel scaffold, reminiscent of the oligosaccharide/oligonucleotide-binding (OB) motif. The C-terminal domain, on the other hand, consists mainly of ␤-sheets and is structurally less well defined. There are three disulfide bonds in each domain of TIMP. Mutagenesis and kinetic analysis on the full-length and N-terminal forms of TIMP-1-3 showed that the MMP-inhibitory function of TIMP resides exclusively at the N-terminal domain (13)(14)(15)(16).
Despite the fact that the co-crystal structure of MT1-MMP/ TIMP-2 has been available for some time, hitherto there is no explanation as to why TIMP-1 is not capable of inhibiting MT1-MMP. Here we report a novel strategy adopted by our laboratory in approaching this conundrum. We succeeded in identifying the surface epitopes that render TIMP-1 inactive against MT1-MMP. By exchanging these "obstructive" epitopes with those of other TIMP, we created a N-TIMP-1 mutant that exhibited inhibitory potency indistinguishable from those of other TIMPs.
Site-directed Mutagenesis of N-TIMP-1-TIMP-1 mutants in this work were created by PCR using Vent DNA polymerase. Mutations were incorporated into full-length and N-terminal domain forms of human TIMP-1 cDNAs (in pRSET-c Escherichia coli expression vector, Invitrogen) by either forward or reverse primers, depending on the locations. All constructs have been sequenced to confirm that no unwanted mutation had been introduced during the mutagenesis process.
Protein Refolding and Assessment of Activity by Titration-N-TIMP-1 mutants were expressed in E. coli and refolded as described previously (19). Full-length wild type and V4A/P6V/T98L mutant were refolded essentially as with N-TIMP-1 with the exception that 0.5 M arginine was included in the refolding buffer. The concentration and activity of active TIMP in each preparation were assessed and determined by titration against MMP-2 as reported in our previous paper (19) on N-TIMP-1 engineering. With the exception of T98F, all N-TIMP-1 mutants retained full inhibitory activity against MMP-2, irrespective of their loci. Refolded full-length TIMP-1 and its V4A/P6V/ T98L mutant were highly active against MMP-2, their association rates with MMP-2 (k on ) being over 10 Ϫ8 M Ϫ1 s Ϫ1 . The amount of T98F was determined by Bio-Rad protein assay kit.
Inhibition Constant Measurement (K i app )-MT1-MMP (0.15 nM) was pre-incubated with increasing concentrations of N-TIMP-1 mutants, ranging from 0 to 500 nM, depending on the potency of the inhibitors. Incubation was allowed at room temperature for 3 h before steady state (V s ) measurement. Quenched fluorescent peptide QF-24 was added to a final concentration of 1 M to initiate the assays. Measurements of enzyme activities were performed at 27°C throughout this work. All data were fitted into competitive tight binding equations with the computer program Grafit to obtain an estimation of K i app values with Equation 1, where V 0 is the rate in the absence of inhibitor; E t is the total enzyme concentration; and I t is the total inhibitor concentration. Association Rate Constant Measurement (k on )-k on measurements were performed by adding N-TIMP mutants (3-50 nM) to MT1-MMP enzyme (0.05 nM). The rate of inhibition was followed using a continuous fluorometric assay at 27°C until steady state was reached. The progress curve was analyzed using Equation 2, where p is the product concentration; V 0 is the initial velocity; V s ϭ steady state velocity; and k is the apparent first order rate constant of equilibrium between enzyme and TIMP complex. k on values were calculated by linear regression of k on TIMP concentrations.

RESULTS
Strategy for TIMP-1 Mutagenesis-Although members of the TIMP family are well conserved in both primary and tertiary conformation, the inability of TIMP-1 to inhibit MT1-MMP has been a long standing conundrum for protein engineers interested in the mechanism of TIMP-MMPs selectivity (20,21). First and foremost, there is no obvious contrast between the amino acid distribution of TIMP-1 and those of TIMP-2-4 that could account for its inability to inhibit MT1-MMP (Fig. 1). Hypothetical docking of MT1-MMP and TIMP-1 models in our laboratory also failed to reveal significant interfacial clashes that could satisfactorily explain the lack of ability of TIMP-1 to form tight binary complex with MT1-MMP. The most challenging aspect, in our view, is that although the structures of TIMP-1 and -2 and MT1-MMP have been delineated by protein NMR and crystallography for a considerable length of time, all the hypotheses regarding the inactivity of TIMP-1 against MT1-MMP remained hitherto speculative.
Taken together, we propose that the lack of TIMP-1 activity against MT1-MMP is not due to simple incompatibility between the surface topology of the two molecules. Instead, we believe that TIMP-1 is not capable of inducing conformational changes in MT1-MMP, an isomerization process that needs to be overcome for tight enzyme-inhibitor (EI) complex formation (22). The reason for this, we hypothesize, is the presence of some "obstructive epitope(s)" that prevent TIMP-1 from initiating conformational changes with the enzyme. These obstructive epitope(s), in our view, must be rather subtle and are not readily identifiable by primary sequence alignment and structural examination.
Hence, to approach this problem in a precise manner, we decided to confine our investigation to the N-terminal domain forms of TIMP (N-TIMP) and the catalytic domain form of MT1-MMP. Structural delineation of stromelysin-1⅐TIMP-1 and MT1-MMP⅐TIMP-2 complexes by x-ray crystallography showed that the TIMPs inhibit MMPs by inserting an "MMPbinding ridge" into the catalytic site grooves of the enzymes (8, 11). This MMP-binding ridge, by and large, is composed of the very N terminus, the AB-loop, the CD-loop and the EF-loop of the molecule (Fig. 2). In our opinion, the obstructive epitope(s) are more likely to be located at these loci rather than the OB-core of the molecule.
Based on this hypotheses, we divided the MMP-binding ridge of TIMP-1 into five "divisions," namely the "N terminus" division, the "Pro 6 " division, the "AB-loop" division, the "CD-loop" division, and finally the "EF-loop" division ( Fig. 2). In turn, each of these divisions is sub-divided into independent entities termed "epitopes." These epitopes could either be a single or multiple amino acids, depending on the locus. Systematically, residues constituting these epitopes were swapped with the corresponding amino acids of TIMP-2-4, and the kinetic profiles of the resultant mutants were monitored throughout the mutagenesis process.
Kinetic Analysis of Wild type N-TIMP-1-4 against the Catalytic Domain of MT1-MMP-Before any mutagenesis work was carried out, we started by examining the binding affinities of wild type N-TIMP-1-4 with the catalytic domain of MT1-MMP. The binding constant (K i app ) and association rate (k on ) are shown in Table I. Clearly, N-TIMP-2-4 have far superior affinities than N-TIMP-1, their K i app values being in the range of 0.3 to 1.5 nM. In comparison, N-TIMP-1 is at least 2 orders of magnitude higher in value (K i app 178 nM). In term of association rate, N-TIMP-2 is slightly superior to N-TIMP-3 and N-TIMP-4 (k on N-TIMP-2, 10 ϫ 10 Ϫ5 M Ϫ1 s Ϫ1 , and N-TIMP-3 and -4, 3-6 ϫ 10 Ϫ5 M Ϫ1 s Ϫ1 ) (Table I). N-TIMP-1, on the other hand, is not capable of establishing tight binding complexes with MT1-MMP.
N-terminal Mutants-The side chains of the second and the fourth residues of TIMP (also termed P1Ј and P3Ј subunits) are critically important in the determination of the selectivity profile of the inhibitor. The residues dock directly into the S1Ј and S3Ј catalytic pockets of MMPs. The second residue of TIMP-1 and -3 is threonine and that of TIMP-2 and -4 is serine (Fig. 1). Hence, we created only one mutant at the position of Thr 2 (T2S). The fourth residue of TIMP-1 is valine. The corresponding residues in TIMP-2-4 are either serine or alanine (Fig. 1). Therefore, two mutants were created at this locus, namely V4S and V4A. The results of the mutagenesis are shown in Table II. Among the three mutants of the N terminus division (T2S, V4A, and V4S), V4A exhibited substantial improvement in binding affinity with MT1-MMP (K i app 66 nM versus wild type N-TIMP-1 of 178 nM), closely followed by V4S (K i app 81 nM). Nonetheless, none of the mutants was capable of potentiating tight complex formation with MT1-MMP.
Pro 6 Mutants-Pro 6 is located at the junction between the N terminus and the first ␣-helix loop of TIMP-1 (Fig. 2). TIMP-1 is unique as it is the only TIMP to have a proline at this locus. In TIMP-2-4, the corresponding residues are valine, serine, and alanine (Fig. 1). Substitution of Pro 6 by a smaller residue might arguably release the constraints and bring about more relaxed local dynamics, although none of the crystal or NMR papers on TIMP published so far have emphasized the significance of this residue. Here we mutated Pro 6 to valine (P6V), serine (P6S), and alanine (P6A) to mimic TIMP-2-4, respectively. Despite the 2-fold reduction in the   2. Anatomy of N-TIMP-1. TIMPs inhibit metalloproteinases by inserting the MMP-binding ridge into the active site of the enzyme. We propose that the inactivity of TIMP-1 against MT1-MMP is most likely to be due to the presence of obstructive epitopes at the MMP-binding ridge of the molecule. Hence, we divide the ridge into five "divisions," each composed of one or more independent epitopes. The amino acid residues chosen for mutagenesis in this work are highlighted in the ball-and-stick format. that this cavity functions as a receptacle for Tyr-36, a conspicuous residue at the tip of the AB-loop of TIMP-2. Although Williamson et al. (23) did confirm that the Tyr-36 residue is crucial for TIMP-2 to establish tight binding complex with MT1-MMP, the findings did not explain why TIMP-3 and TIMP-4 are equally good inhibitors against MT1-MMP despite the complete lack of a tyrosine (or an amino acid of similar nature) at the tip of the AB-loop. Hence, the rationale of grafting the entire AB-loops of TIMP-2-4 onto TIMP-1 was not to find out if any of the residues on the AB-loop of TIMP-3 and -4 could interact with the MT1-loop "receptacle" of MT1-MMP. Rather, we were interested in finding out whether the TIMP-1 AB-loop "obstructs" the molecule from initiating conformational changes with MT1-MMP. Our data indicated that TIMP-2 and TIMP-3 AB-loops were slightly beneficial in enhancing the affinity of TIMP-1 mutants against MT1-MMP (K i app of 77-106 nM) (Table II and Fig. 3). Disappointingly, none of the mutants could be considered tight binding inhibitor in MT1-MMP inhibition.
CD-loop Mutants-The fourth group of mutants consisted of the CD-loop mutants. Only two residues were considered important at this site, namely Met-66 and Val-69. We know from the crystal structures of free and TIMP-1-bound forms of stromelysin-1 (Protein Data Bank codes 1QIA and 1UEA) that Met-66 was capable of inducing some degree of conformational change in stromelysin-1 (11). Hence, a series of mutants were created to replace Met-66 at this site. The mutations included amino acids of different characteristics, ranging from acidic to basic as well as those hydrophobic in nature (M66K, M66D, M66L, M66I, M66V, M66A, and M66G). On the other hand, given that leucine is the only variant at the Val-69 locus (Fig.  1), only one mutant was created to replace the residue (V69L). Subsequent kinetic analysis showed that the majority of the mutations only impaired the affinity against MT1-MMP (K i app from 146 nM to an excess of 500 nM) (Table II).

EF-loop Mutants-
The last mutant in the series was the Thr 98 to leucine mutant (T98L) from the EF-loop division. Thr 98 is situated right before the second disulfide bond (Cys 3 -Cys 99 ) of TIMP-1. Interestingly, TIMP-1 is the only member of the TIMP family that has a threonine at this locus (Fig. 1). The corresponding residue in TIMP-2-4 is leucine. Indeed, replacement of threonine by leucine vastly enhanced the affinity of the resultant mutant against MT1-MMP (K i app 11 nM) (Table II and Fig. 4). Indeed, the K i app value is 16-fold lower than that of the wild type N-TIMP-1. The most striking effect is that the T98L mutant clearly manifested inhibitory profiles reminiscent of a slow, tight binding inhibitor (Fig. 4). Throughout the series, T98L is the first mutant exhibiting inhibitory profiles akin to those of N-TIMP-2-4.
Thr 98 Point Mutants-So far, we have identified Thr 98 to be the key obstructive residue that renders TIMP-1 inactive against MT1-MMP. It would be interesting to find out the effects of other amino acids on MT1-MMP inhibition. With the exception of cysteine, we mutated Thr 98 to all the available amino acids, and the kinetic profiles of the mutants are shown in Table III. Not surprisingly, isoleucine produced the same potentiation effect as leucine, the K i app value of T98I mutant (12 nM) being indistinguishable from that of T98L. The next two most potent amino acids are valine and methionine (K i app T98V and T98M ϳ30 nM). Glutamine and tyrosine were slightly poorer, the affinity of the T98Q and T98Y mutants (90 -120 nM) being marginally better than the wild type protein. Replacement of Thr 98 by serine, on the other hand, produced no apparent effect on the activity of N-TIMP-1. The remaining amino acids severely impaired the affinity of N-TIMP-1 against MT1-MMP (Table III).
Combination of Good Epitopes-Even though T98L was much more active than wild type N-TIMP-1, the mutant was still not as potent as N-TIMP-2, -3, or -4. Hence, in the second phase of this work, we combined four of the major positive epitopes in an attempt to study their effects on MT1-MMP inhibition. Two multiple mutants were made: 1) V4A/P6V/ T98L triple mutant, and 2) V4A/P6V/TIMP-2 AB-loop/T98L quadruple mutant. The results of the combination are shown in Table IV. First of all, incorporation of V4A and P6V significantly improved the affinity of the T98L mutant. The affinity of the V4A/P6V/T98L triple mutant (K i app 1.66 nM) was practically equal to those of N-TIMP-2 (K i app 1.30 nM) and N-TIMP-3 (K i app 1.38 nM). The association rate of the mutant (k on 1.48 ϫ 10 Ϫ5 M Ϫ1 s Ϫ1 ) exceeded 10 Ϫ5 M Ϫ1 s Ϫ1 , closely resembling N-TIMP-3 and -4 (k on 3-6 ϫ 10 Ϫ5 M Ϫ1 s Ϫ1 ). Unexpectedly, the incorporation of the TIMP-2 AB-loop significantly impaired the affinity of V4A/P6V/T98L (K i app of V4A/P6V/TIMP-2 AB-loop/T98L being 5.2 nM). The association rate, however, did not seem to be affected (Table IV).
Effects of the C-terminal Domain on V4A/P6V/T98L Activity-So far, we have succeeded in identifying the obstructive epitopes on the N-terminal domain of TIMP-1 that hinder the inhibitor from establishing a tight binding complex with MT1-MMP. What are the effects of the C-terminal domain on MT1-MMP association? To address this issue, we introduced V4A/

of MT1-MMP with the first generation N-TIMP-1 mutants
The first generation N-TIMP-1 mutants were created based on the sequence of TIMP-2-4 as described in text. Met 66 was mutated to various amino acids to find out the role of the CD-loop in the MT1-MMP interaction. Among the mutants created, only T98L displayed a clear sign of a slow, tight binding inhibitor with MT1-MMP. The association rate (k on ) of the mutant is 1.35 Ϯ 0.08 ϫ 10 Ϫ5 M Ϫ1 s Ϫ1 , almost parallel with those of wild-type N-TIMP-3 and -4 (see Table I). In the AB-loop mutants section, T2, T3, and T4 indicate for TIMP-2-4 respectively.  P6V/T98L mutations into full-length TIMP-1, and the results are summarized in Table V. At first sight, it appeared that the C-terminal domain might improve the affinity of TIMP-1 against MT1-MMP, because full-length wild type TIMP-1 displayed significantly better affinity than its N-terminal counterpart (K i app , full-length wild type TIMP-1 91 nM versus N- TIMP-1 of 178 nM). Subsequent comparison of the V4A/P6V/ T98L mutant, however, revealed that the C-terminal domain has no beneficial effects on MT1-MMP inhibition (K i app fulllength V4A/P6V/T98L 1.86 nM). Furthermore, the association rate of the full-length mutant (0.9 ϫ 10 Ϫ5 M Ϫ1 s Ϫ1 ) was slightly lower than the N-terminal version (Table V and Fig. 5). DISCUSSION Not only are TIMP the endogenous inhibitors of MMPs, they also modulate the enzymatic activities of the ADAM (a disintegrin and metalloproteinase) and ADAM-TS (ADAM with thrombospondin-like repeats) proteinases (reviewed in Ref. 20). Membrane-type MMPs, in general, are poorly inhibited by TIMP-1. What is the physiological significance of such inactivity and can protein engineers ever fully understand the ultimate molecular mechanism of TIMP-metalloproteinases selectivity?
Contradicting previous perceptions, our study shows that Thr 98 is the pivotal obstructive epitope that renders TIMP-1 inactive against MT1-MMP. Thr 98 is TIMP-1-specific, and other TIMPs have leucine at the equivalent position. How does leucine potentiate the binding of TIMP-1 to MT1-MMP? Reexamination of the available TIMP/MMP structures failed to provide a satisfactory answer to the question. In the stromelysin-1 (MMP-3)-TIMP-1 complex (Protein Data Bank code 1UEA), Thr 98 is situated right before His 211 (HEXX-HXXGXXH 211 ) of the enzyme, the last of the three conserved histidines that forms the catalytic zinc-binding ligands. Distance-wise, the two MMP-3 residues closest to Thr 98 are His 211 and Pro 221 , the amino acids being ϳ4 Å from the side chain of Thr 98 (Fig. 6). This aside, Thr 98 does not seem to be in close contact with any particular hydrophobic residue on the surface of stromelysin-1. The TIMP-2 equivalent of Thr 98 is Leu 100 (Fig. 1). The crystal structure of the MT1-MMP⅐TIMP-2 complex (Protein Data Bank code 1BUV) again shows that Leu 100 is located before the third conserved histidine (His 249 ) of the zinc-binding motif (HEXXHXXGXXH 249 ) (Fig. 6). The two MT1-MMP residues closest to Leu 100 are His 249 and Pro 259 , almost a spitting image of the setting found in TIMP-1⅐MMP-3 mentioned above (Fig. 6). Hence, could this "His 249 /Pro 259 pair" in MT1-MMP be a deciding factor in its rejection of TIMP-1? As part of our modeling simulation, we replaced the TIMP-2 molecule in MT1-MMP⅐TIMP-2 (Protein Data Bank code 1BUV) complex with TIMP-1 bearing a Thr 98 to leucine mutation. Our study suggests that substitution of Thr 98 by leucine does not enhance the interfacial contact between TIMP-1 and MT1-MMP enzyme (not shown).
Thr 98 is not the only residue we considered that is unique to TIMP-1. As mentioned earlier, Pro 6 was also featured prominently on our list of mutagenesis study. Substitution of the residue by valine or serine improved the binding affinities against MT1-MMP significantly. Yet again, examination of the stromelysin-1/TIMP-1 structure suggests that the residue is not directly involved in MMP association.
Why should leucine and isoleucine be the best residues? Table III demonstrates that the activity of N-TIMP-1 is critically dependent on the biophysical characteristics of the amino acids occupying the Thr 98 position. The best amino acids are

TABLE IV Apparent inhibition constant (K i app ) and association rate (k on ) of MT1-MMP with N-TIMP-1 mutants of combined epitopes
Epitopes that contribute positively towards MT1-MMP inhibition, i.e. V4A, P6V, T98L, and TIMP-2 AB-loop, were combined into two multiple mutants. The K i app and k on values of the V4A/P6V/T98L mutant are indistinguishable from that of the wild-type N-TIMP-3 and -4 (see Table  I). T2-AB-loop, TIMP-2 AB-loop.  Table I). However, the C-terminal domain does not seem to exert a significant effect on the affinity and association rate of the V4A/P6V/T98L mutant. NA, not able to determine. those similar in nature to leucine, namely isoleucine, valine, and methionine. Too bulky and hydrophobic (tryptophan and phenylalanine) or minute (glycine and alanine) a side chain resulted in complete abrogation of activity. Strangely, even though phenylalanine is poorly tolerated, tyrosine is beneficial, notwithstanding its similar size with phenylalanine. In general, acidic (aspartate and glutamate) or basic residues (lysine and arginine) are poorly tolerated.
Throughout this work, we were intrigued by the fact that some of the individual positive epitopes identified are not always mutually complementary as hoped. Incorporation of V4A and P6V mutations into T98L, for example, enhanced the affinity of N-TIMP-1 against MT1-MMP to a level essentially equal to those of N-TIMP-2-4. The effect is additive. The same could not be said for the TIMP-2 AB-loop; the epitope is clearly incompatible with T98L mutation. The problem is again highlighted by our findings on the full-length forms of the V4A/P6V/ T98L mutant. In comparison with wild type N-TIMP-1, fulllength TIMP-1 is marginally superior in MT1-MMP binding. Subsequent study on the V4A/P6V/T98L mutant, however, suggests that the C-terminal domain was either inert or slightly detrimental to the affinity and association rate against MT1-MMP.
Taken together, it is highly unlikely that the findings in this work could be explained satisfactorily by steric reasons alone. The true answer, we believe, lies in the molecular dynamics of TIMP-1 and MT1-MMP that governs the course of their interaction. Crystallographic studies of MMP⅐TIMP complexes only portray static representations of the enzymes and their inhibitors. Undeniably, the parts of TIMP with the closest intimacy with MMPs are the N terminus, the AB-loop, and the CD-loop. Here, we show that the key to TIMP/MMPs selectivity might in fact, lie elsewhere. What is most striking is that the pivotal amino acid, Thr 98 , is a rather "insignificant" residue that has never been emphasized in any of the literature on TIMP analysis and engineering so far (21, 24 -28).
As mentioned under "Experimental Procedures," with the exception of T98F, all Thr 98 mutants were highly active against MMP-2 regardless of the nature of the amino acids occupying the position. What about other MMPs, ADAM and ADAM-TS proteinases? We are currently evaluating the activity profiles of our TIMP mutants with these proteinases, and the findings will be presented in the near future.
This paper is the first in which a variant of inactive TIMP has been successfully transformed into a fully active one against a specific MMP backdrop. We hope the findings in this work will broaden our views on the mechanism of MMP/TIMP selectivity.