Protein engineering of the tissue inhibitor of metalloproteinase 1 (TIMP-1) inhibitory domain. In search of selective matrix metalloproteinase inhibitors.

Studies of the structural basis of the interactions of tissue inhibitors of metalloproteinases (TIMPs) and matrix metalloproteinases (MMPs) may provide clues for designing MMP-specific inhibitors. In this paper we report combinations of mutations in the major MMP-binding region that enhance the specificity of N-TIMP-1. Mutants with substitutions for residues 4 and 68 were characterized and combined with previously studied Thr(2) mutations to generate mutants with improved selectivity or binding affinity to specific MMPs. Some combinations of mutations had non-additive effects on DeltaG of binding to MMPs, suggesting interactions between subsites in the reactive site. The T2L/V4S mutation generates an inhibitor that binds to MMP-2 20-fold more tightly than to MMP-3(DeltaC) and over 400-fold more tightly than to MMP-1. The T2S/V4A/S68Y mutant is the strongest inhibitor for stromelysin-1 among all mutants characterized to date, with an apparent K(i) for MMP-3(DeltaC) in the picomolar range. A third mutant, T2R/V4I, has no detectable inhibitory activity for MMP-1 but is an effective inhibitor of MMP-2 and -3. These selective TIMP variants may provide useful tools for investigation of biological roles of specific MMPs and for possible therapy of MMP-related diseases.


Studies of the structural basis of the interactions of tissue inhibitors of metalloproteinases (TIMPs) and matrix metalloproteinases (MMPs) may provide clues for
designing MMP-specific inhibitors. In this paper we report combinations of mutations in the major MMP-binding region that enhance the specificity of N-TIMP-1. Mutants with substitutions for residues 4 and 68 were characterized and combined with previously studied Thr 2 mutations to generate mutants with improved selectivity or binding affinity to specific MMPs. Some combinations of mutations had non-additive effects on ⌬G of binding to MMPs, suggesting interactions between subsites in the reactive site. The T2L/V4S mutation generates an inhibitor that binds to MMP-2 20-fold more tightly than to MMP-3(⌬C) and over 400-fold more tightly than to MMP-1. The T2S/V4A/S68Y mutant is the strongest inhibitor for stromelysin-1 among all mutants characterized to date, with an apparent K i for MMP-3(⌬C) in the picomolar range. A third mutant, T2R/V4I, has no detectable inhibitory activity for MMP-1 but is an effective inhibitor of MMP-2 and -3. These selective TIMP variants may provide useful tools for investigation of biological roles of specific MMPs and for possible therapy of MMP-related diseases.
Degradation of the extracellular matrix is essential for normal biological processes including embryonic development and morphogenesis (1,2), reproduction (3) and wound healing (4), and enhanced turnover is associated with diseases including arthritis (5,6), tumor angiogenesis and metastasis (7), multiple sclerosis (8), and cardiovascular diseases (9). The matrix metalloproteinases (MMPs) 1 are a family of more than 20 zinc-dependent proteases that catalyze extracellular matrix turnover (10). Activity and zymogen activation in MMPs are regulated by a group of endogenous proteins named the tissue inhibitors of metalloproteinases (TIMPs) (11).
TIMPs are distributed in both invertebrates and vertebrates (11)(12)(13). The mammalian TIMPs are a family of four members (TIMP-1-4) that have about 40% sequence identity and fold into two domains, each containing three disulfide bonds (11). The isolated N-terminal domains (N-TIMPs) are able to form the correct native structure that carries the inhibitory activity against the MMPs (14). Although there are four TIMPs, their inhibitory activities toward different MMPs are not particularly specific. A notable exception is that TIMP-1 is a weak inhibitor of MT-MMPs, whereas TIMP-2 and TIMP-3 are much more effective (15)(16)(17).
Reported structures of TIMPs include crystal structures of TIMP-1 in a complex with the MMP-3 catalytic domain (18), TIMP-2 in a complex with the catalytic domain of membrane type MT1-MMP (19) and in a free form (20), and the solution NMR structures of N-TIMP-1 (21) and N-TIMP-2 (22). These structures show that the N-terminal inhibitory domain consists of a 5-stranded ␤-barrel with three associated ␣-helices resembling the folds of members of the oligonucleotide/oligosaccharide binding (OB) protein family (23). The structure of the TIMP-1/MMP-3 complex reveals that about 75% of all intermolecular contacts are made by residues adjacent to the disulfide bond between Cys 1 and Cys 70 , especially residues 1-5 and 66 -70. These two sections of chain insert into the active site cleft of the MMP, thus blocking its accessibility to substrates (18). The N-terminal Cys 1 coordinates the catalytic Zn 2ϩ through the ␣-amino group and the peptide carbonyl group and is crucial for the inhibitory activity of TIMPs for MMPs, as shown by the complete loss of inhibitory activity for MMPs in TIMP-2 on carbamylation of the ␣-amino group of the NH 2terminal Cys 1 (24) or mutation to append an alanine extension to the amino terminus (25).
Our previous mutagenesis studies of N-TIMP-1 (26,27), together with work with N-TIMP-2 by others (28), suggest that the affinity and specificity of TIMP for MMPs can be modified by site-directed mutagenesis. A major determinant of the affinity of N-TIMP-1 for different MMPs is the residue at position 2 in the sequence (threonine 2 in the wild-type protein) which interacts with the S1Ј pocket of MMPs, a key to MMP substrate specificity (27). Based on this, it is reasonable to hypothesize that other residues that make contact with MMPs also contribute to the binding affinity and specificity of N-TIMP-1 and that selective variants can be generated by combining suitable mutations at these sites. In the present study, we show that N-TIMP-1 mutants with substitutions at positions 4 and 68 showed changes in affinity and specificity for MMPs. Combinations of mutations in these positions with those in position 2 led to the discovery of N-TIMP-1 variants with higher binding affinity and specificity for individual MMPs.
PCR products were cloned into pET-3a vector using the NdeI and BamHI restriction sites. All constructs were confirmed by DNA sequencing using T7 promoter primer. The N-Timp-1 mutants were expressed in Escherichia. coli BL21(DE3) cells. Protein was purified from inclusion bodies and folded in vitro as described previously (29).
Further Purification of N-TIMP-1 and Mutants-N-TIMP-1 and mutant proteins purified by cation exchange chromatography with CM-52 were dialyzed overnight against 15 volumes of 20 mM bis-Tris-HCl, pH 5.5 (pH 6.0 for T2R/V4I) and applied to a cation exchange Mono S HR 5/5 column previously equilibrated with the same buffer and connected to a Biologic DuoFlow medium pressure chromatography system. The protein was eluted with a linear salt gradient of 0 -0.5 M NaCl over 60 min at a flow rate of 1 ml/min. The activity of different fractions was estimated by the fluorescence assay method using MMP-3(⌬C) and NFF-3 substrate as described (26).
Two similarly sized peaks, which slightly overlap, were obtained in this separation, and fractions corresponding to the second peak, which contained the MMP-3 inhibitory activity, were pooled and titrated with MMP-3(⌬C). Various concentrations of the inhibitors were incubated with MMP-3(⌬C) (300 nM) for 4 h at 37°C, diluted 300-fold with TNC buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM CaCl 2 and 0.02% Brij 35) and immediately assayed with 1.5 M NFF-3 substrate as described (26). These results showed that this protein is Ͼ85% active for MMP inhibition.
Inhibition Kinetic Studies-K i (app) of N-TIMP-1 and mutants against MMP-1, -2, and -3(⌬C) were determined using fluorescence assay as described previously (26)  , the following formulas were used (30) as seen in Equations 1 and 2, where I t is the total inhibitor concentration, E t is the total enzyme concentration, A t is the total substrate concentration, and K m is the Michaelis constant. In our assays the value of E t /K i (app) does not exceed 100 so that the inhibitor is distributed in both free and bound forms, and K i (app) can be calculated by fitting inhibition data to Equation 1 (30). Because the inactive portion of the N-TIMP-1 does not interfere with the binding of the active inhibitor with MMPs, as shown by isothermal titration by isothermal titration, 2 the true K i values can be determined by multiplying the K i , calculated as described above, by the fraction of active TIMP determined by titration. Because the substrate concentration is very low relative to the estimated K m , the apparent K i values are essentially identical to the true K i values.  (27), these mutations produce more moderate changes (Table I). Mutations of Val 4 into isoleucine, lysine, and serine cause significant increases in the K i (app) for MMP-1 while having only a small effect on the affinity for MMP-2.

Substitutions for
Mutants V4K and V4S are more selective for MMP-2 than the wild-type N-TIMP-1 as a result of minor changes in affinity for MMP-2 and larger reductions in binding to MMP-1 and MMP-3. The V4S mutant has an unchanged K i for MMP-2 but 5-fold and 8-fold increased K i values for MMP-1 and -3, respectively. The other two mutants also have modified inhibition activities; V4A has an increased affinity for MMP-3(⌬C), whereas V4I results in a 15-fold reduced affinity for MMP-1 but unchanged activity with MMP-2 and MMP-3(⌬C).
Substitutions for Ser 68 -Residues 66 -70 of the C-D loop also form part of the core of the TIMP/MMP contact site ( Fig. 1 and Ref. 18), and previous studies have shown that substitutions for Met 66 or Val 69 affect TIMP activity (26). Here we mutated the Ser 68 to Ala, Glu, Arg, and Tyr. These mutations have large effects on MMP binding (Table I). Three of the four mutants have improved selectivity for MMP-2 relative to the other two MMPs, whereas the fourth mutant, S68Y, inhibits MMP-3 much more strongly than MMP-1 and -2.
Combined Mutations of Thr 2 , Val 4 , and Ser 68 -Based on the mutagenesis studies of Thr 2 , Val 4 , and Ser 68 , we constructed double and triple mutants containing combinations of the more selective single-site mutations. Characterization of these mutants shows that effects of the individual mutations are generally additive but with some exceptions (Table II), suggesting that these residues do not always contribute independently to the stability of the TIMP/MMP complex.
The T2L mutation was combined with V4S to generate a N-TIMP-1 variant that is more selective for MMP-2. As predicted, with MMP-2 the resulting double mutant inhibits 20fold more strongly than with MMP-3(⌬C) and about 470-fold more strongly than with MMP-1. Introduction of a third mutation, S68A, produced a mutant binding very weakly to MMP-1 while retaining a good activity for MMP-2. However, this inhibitor is much more effective with MMP-3 than is predicted based on additive effects on the free energy of binding (Table  II). This triple mutant, T2L/V4S/S68A, is less selective against MMP-3(⌬C) than the double mutant T2L/V4S.
The double mutant, T2S/V4A, has an increased inhibitory activity for MMP-3. Binding with MMP-2 is unchanged, whereas that with MMP-1 is weaker than the wild-type protein. Interestingly, the triple mutant T2S/V4A/S68Y has the highest inhibitory activity for MMP-3(⌬C) among all mutants characterized so far, with a K i of 50 pM. Unexpectedly, it also has a greatly improved affinity for MMP-2, whereas binding to MMP-1 was reduced over 36-fold.
The T2R/V4I mutant was purified as a ϳ95% active form using a Mono S column at pH 6.0 instead of pH 5.5, which is used for the wild-type protein. We could not detect any inhibition of MMP-1 by this mutant even at a concentration of 10 M, yet it retains good activity as an inhibitor of MMP-2 and MMP-3(⌬C).

DISCUSSION
The TIMPs are important regulators of extracellular matrix metabolism (11), principally through their primary activities as endogenous inhibitors of MMPs. Imbalances between the levels of TIMPs and active MMPs are linked to disease processes such as tumor metastasis, arthritis, and atherosclerosis (31,32). Low molecular weight MMP inhibitors have been developed and extensively studied, and a few have been used in clinical trials but with little success (33). Nonspecific inhibition of housekeeping MMP functions has been proposed as the major drawback of these inhibitors (34). We proposed previously that N-TIMP-1 can be engineered to produce more selective MMP inhibitors that could be applied in gene therapy of MMP-related diseases (35,36), and residue 2 of TIMP-1 was found to be a major determinant of affinity and specificity for MMPs (27). In this paper we report the effects of mutagenesis of two other residues that are important in TIMP/MMP binding; combinations of mutations in these positions with Thr 2 mutations generate N-TIMP-1 variants with increased selectivity and/or affinity for specific MMPs.
The side chain of Val 4 occupies a site similar to the substrate P3Ј-subsite in currently known TIMP/MMP complexes (18,19). In the complex of TIMP-1 with MMP-3, the side chain of Val 4 sits in a shallow groove at the margin of the interaction site close to the side chains of Gly 161 , Asn 162 , Leu 164 , and Tyr 223 of the protease (15). These residues are Ͼ4 Å from Val 4 , and the insignificant effect of truncating the side chain through the Ala substitution on the affinity for all three MMPs suggests that interactions between the side chain of Val 4 and the protease contributes little to the free energy of binding. The Val 4 side chain projects away from the protease surface in the TIMP-1/ MMP-3 complex (18), and the more extended Ile side chain can be accommodated in MMP-2 and -3 without perturbing the protein-protein interaction, but in MMP-1 it produces a 14-fold   loss in affinity. The basis for this is uncertain, because the structures of complexes of TIMP-1 with MMP-1 and -2 have not been determined, but this result suggests that the S3Ј site of MMP-1 does not readily accommodate a larger side chain; it is also affected more than the other proteases by the Lys substitution (Table I). Differences between the MMPs in the residues that form the P3Ј site do not readily account for these effects; residues Leu 164 and Tyr 223 of MMP-3 are conserved in MMP-1 and -2, whereas residues corresponding to Gly 161 -Asn 162 are Gly-Gly in MMP-1 and Asp-Gly in MMP-2, so changes in side chain size are not responsible for the reduced steric tolerance in this subsite in MMP-1. The binding of TIMP-1 to MMP-2 is least affected by substitutions for Val 4 , suggesting that there is greater separation between MMP-2 and N-TIMP-1 in this part of the interaction site. The side chain of Ser 68 of TIMP-1 interacts with the S2 subsite of MMP-3 in the crystal structure of their complex and is in contact with Ala 167 of the metalloproteinase (18). It is also near (Ͻ4.5 Å distance) to His 166 , Tyr 168 , Ala 169 , and His 205 . Mutation of Ser 68 to Ala, Glu, and Arg reduces the affinity of N-TIMP-1 for all three MMPs, but the effect is less for MMP-2 than for MMP-1 and MMP-3. The Tyr mutant has a major effect on inhibitor specificity, producing a 150-fold loss in affinity for MMP-1, a 7-fold reduction in affinity for MMP-2, but essentially no change in binding to MMP-3. The side chain of Ser 68 of TIMP-1 projects toward Ala 169 of MMP-3 in their complex (18), and it appears that the Tyr side chain can be accommodated without major perturbation of the interaction interface. Ala is conserved at this site in MMP-2, but in MMP-1 it is replaced by Gln. A steric conflict between the Gln side chain and Tyr 68 in the N-TIMP-1 mutant is a possible explanation of the effects of this mutation on inhibitor selectivity.
Substitutions for residues 2, 4, and 68 with similar selectivity were combined in an attempt to engineer N-TIMP-1 variants with higher selectivity and/or affinity for specific MMPs. In some cases, the effects of these mutations are essentially additive (Table II), indicating that interactions with the S1Ј, S3Ј, and S2 subsites of MMPs contribute independently to the stability of the TIMP/MMP complex. However, some mutants containing combinations of mutations have much lower K i values for particular MMPs than predicted based on the assumption that individual substitutions have additive effects on the ⌬G of binding. For example, the T2L/V4S/S68A mutant has a Ͼ100-fold higher affinity for MMP-3(⌬C) than expected, yet its K i values for MMP-1 and MMP-2 are in good agreement with predictions. Similarly, the T2S/V4A/S68Y is a 60-fold better inhibitor of MMP-1 and also a 68-fold better inhibitor of MMP-2 than predicted based on additivity. These discrepancies greatly exceed the compounded errors of the single-site mutations and suggest interactive effects between the sites in these triple mutants. Because of the complexity of protein-protein interactions, there are many possible explanations for this, such as changes in relative orientation of TIMP and protease, structural changes introduced by the mutations, and changes in dynamics (11). Structural studies are in progress to address this question.
Several combined mutants have interesting and potentially useful properties. T2L/V4S is selective for MMP-2, providing a possible tool for gene therapy of gelatinase-related diseases. Another mutant, T2S/V4A/S69Y, is the best inhibitor for stromelysin-1 among all mutants discovered so far, although it also exhibits excellent inhibitory activity for MMP-2. The apparent K i of this inhibitor for MMP-3(⌬C) is 50 pM, providing the most potent inhibitor of stromelysin-1 among all N-TIMP-1 mutants. The third mutant, T2R/V4I, has no detectable activity for MMP-1 while retaining good inhibition for MMP-2 and -3. This mutant is of special interest, because the failure of many general MMP inhibitors in clinical trials was the result of muscular-skeletal disorders that are thought to be caused by nonspecific inhibition of MMP-1 (38).
Our mutational studies have demonstrated the feasibility of generating selective MMP inhibitors by engineering TIMP. Using high throughput screening methods, it should be possible to identify TIMP variants that are specific inhibitors of individual MMPs for application in future clinical trials.