Development of Selective Inhibitors and Substrate of Matrix Metalloproteinase-12*

Four phosphinic peptide libraries with compounds having the general formula p-Br-Ph-(PO2-CH2)-Xaa′-Yaa′-Zaa′-NH2 have been prepared and screened against 10 matrix metalloproteinases (MMPs). We identified two phosphinic peptides with Ki values of 0.19 and 4.4 nm toward MMP-12 (macrophage elastase) that are more than 2-3 orders of magnitude less potent toward the other MMPs tested. These highly selective MMP-12 inhibitors contain a Glu-Glu motif in their Yaa′-Zaa′ positions. Incorporation of this Glu-Glu motif into the sequence of a nonspecific fluorogenic peptide cleaved by MMPs provides a highly selective substrate for MMP-12. A model of one of these inhibitors interacting with MMP-12 suggests that the selectivity observed might be due, in part, to the presence of two unique polar residues in MMP-12, Thr239 and Lys177. These MMP-12-selective inhibitors may have important therapeutic applications to diseases in which MMP-12 has been suggested to play a key role, such as in emphysema, atherosclerosis, and aortic abdominal aneurysm.

The matrix metalloproteinases (MMPs) 2 form a group of structurally related extracellular zinc endoproteases able to degrade at least one protein component of the extracellular matrix (1). Based on this property, MMPs are considered to be critical mediators of both normal and pathological tissue remodeling processes (2,3). Their overexpression is observed in and associated with a variety of diseases, including cancer (4,5), arthritis (6), multiple sclerosis (7,8), and atherosclerosis (9,10). Therefore, there is substantial interest in developing MMP synthetic inhibitors for a variety of therapeutic indications (11)(12)(13)(14)(15). Results of the first clinical trials with broad spectrum MMP inhibitors in cancer therapy were disappointing, highlighting the need for better understanding of the exact role of each MMP during the different stages of tumor progression (16). Recent research in this field has focused on the development of inhibitors that fully differentiate one MMP from another (17). This is a particularly difficult task, since the topology and nature of the residues in the enzyme's active site are highly conserved among the different MMPs (18). Moreover, parts of the MMP catalytic domain, which play a critical role in enzyme specificity, seem to be highly flexible (19,20). This situation may explain why most previously reported MMP synthetic inhibitors preferentially inhibit some MMPs but are not exclusive inhibitors of a single MMP. A recent breakthrough in this field was achieved by identifying the first highly selective synthetic inhibitor of MMP-13 (21). Selective inhibitors for MMP-2 and MMP-9 have also been reported recently, but their degree of selectivity toward MMPs is less than that achieved for the MMP-13-selective inhibitor (22).
To identify highly selective MMP inhibitors, libraries of phosphinic peptides were prepared. Phosphinic peptides are good transition state mimics and have been shown to behave as highly potent inhibitors of different zinc metalloproteinases (23). To probe the SЈ 1 cavity of MMPs, a chemical strategy that makes it possible to prepare phosphinic peptides harboring various substituents in their PЈ 1 position was used (see Scheme 1). This strategy relies on the use of a common precursor, which can be modified in one step, to prepare phosphinic peptides displaying substituted isoxazole side chains in their PЈ 1 position (24). In such phosphinic peptides, the isoxazole ring is used as a rigid scaffold to project in the right orientation various chemical groups able to interact with the SЈ 1 subsite of MMPs, which corresponds to a deep cavity. Based on this strategy, four libraries of phosphinic peptides, containing four different isoxazole side chains in their PЈ 1 position, were prepared by introducing additional chemical diversity in the PЈ 2 and PЈ 3 positions of the inhibitors (see Scheme 1). Screening of these libraries against 10 different MMPs allowed us to identify highly selective inhibitors of MMP-12. Based on the unique structure of these inhibitors, a highly selective substrate of MMP-12 has been developed, and determinants of MMP-12 selectivity have been tentatively mapped by developing a model of interaction of one of these selective inhibitors with MMP-12.

EXPERIMENTAL PROCEDURES
Commercial reagents were used as received without additional purification. Solvents were of the reagent grade available from commercial sources and used without further purification. SynPhase lanterns, colored cogs, and spindles were from Mimotopes. Fmoc-amino acids were from Novabiochem. N,NЈ-diisopropylcarbodiimide (DIC), N-hydroxybenzotriazole (HOBt), trifluoroacetic acid, and triisopropylsilane were from Aldrich. N 2 -Fmoc-N 3 -2,4-dinitrophenyl-L-2,3-diaminopropionic acid and (7-methoxycoumarin-4-yl)acetyl (Mca) were from Bachem. Anhydrous N,N-dimethyl formamide (DMF) was from Fluka. Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 was from Novabiochem. Catalytic domains of mouse MMP-11, human MMP-7, and MMP-14 expressed in Escherichia coli BL21 DE3 pLysS cells were obtained and purified as described before (25). All of the other human MMPs were from R&D Systems. Analytical and preparative reverse phase HPLC separations were respectively performed on a Thermo separation product and Gilson apparatus using AIT C 18 Kromasil (250 ϫ 4.6 mm) and AIT C 18 Kromasil (250 ϫ 20 mm) columns with a flow rate of 1 and 3 ml⅐min Ϫ1 , respectively, with a detection at 230 nm. Each peptide and library was analyzed by electron spray mass spectrometry performed on a Micromass Platform II instrument (Atheris Laboratories, Geneva, Switzerland).
Qualitative and Quantitative Aspects-All sublibraries were analyzed by electrospray ionization mass spectrometry (ESI ϩ and ESI Ϫ ). In each library, 19 pseudotripeptides were clearly identified. For isobar compounds, tandem mass spectrometry experiments were performed, confirming the quality of the sublibrary. Concentration of each sublibrary was obtained by amino acid analysis.
Synthesis of Single Compounds-These compounds were synthesized according to the protocol described above. Pure diastereomeric forms of compounds 1 and 2 (see Table 1) were obtained by HPLC purification. The R and S configurations of the PЈ 1 position in these compounds were assigned as described previously (24). As expected, the diastereomeric form eluting first in HPLC (S configuration) turned out to be the most potent toward MMPs.
Synthesis of Fluorogenic Substrate-The substrate was prepared on SynPhase lanterns by using a standard Fmoc strategy. N 2 -Fmoc-N 3 -2,4-dinitrophenyl-L-2,3-diaminopropionic acid was first incorporated into the solid support using standard coupling conditions (10 eq, 2 ϫ 1.5 h). After sequential incorporation of each amino acid, (7-methoxycoumarin-4-yl)acetyl-proline was incorporated into the peptide sequence (5 eq, 2 ϫ 1.5 h). After cleavage from the lantern, the sample was concentrated and lyophilized from a CH 3 CN/H 2 O mixture. This substrate was purified by reversed-phase preparative HPLC. Analytical HPLC, UV measurements, and mass spectrometry were used to ascertain the purity and structural correctness of the substrate.

Enzyme Assays and Inhibition Studies
Enzyme inhibition assays were carried out in 50 mM Tris/HCl buffer, pH 6.8, 10 mM CaCl 2 , at 25°C as described before (31). With the exception of MMP-11, all of the assays were performed in black 96-well plates (nonbinding surface plates 3650; Corning Costar). All enzymatic studies were performed in order to remain below 10% of substrate hydrolysis (initial rate conditions). Progress curves were monitored by following the increase in fluorescence at 400 nm ( ex ϭ 340 nm), induced by the cleavage of the Mca-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 fluorogenic substrate by MMPs. Fluorescence signals were monitored using a photoncounter spectrophotometer (Fluoroskan Ascent; ThermolabSystems), equipped with a temperature device control and a plate shaker. The conditions of a typical experiment were 200 l of buffer, 0.2-0.5 nM MMP, and a 7.5 M concentration of the above Mca substrate. In the specific case of MMP-11, assays were performed in quartz cuvettes, with mixing by a magnetic stirrer, using a PerkinElmer Life Sciences 50b spectrophotometer. Mca-Pro-Leu-Ala-Cys(OMeBn)-Trp-Ala-Arg-Dpa-NH 2 was used as substrate, as described before, to monitor the activity of MMP-11 (32). Inhibitors or sublibraries at various concentrations were preincubated with the MMPs 30 min before the addition of substrate. The absolute concentration of inhibitor or sublibrary was determined by amino acid analysis. For each inhibitor or sublibrary, the percentage inhibition was determined in triplicate experiments at three concentrations. Inhibitor concentrations were selected in order to observe a 20 -80% range of inhibition. K i(app) values were determined using the method proposed by Horovitz and Levitski (33). This approach explicitly takes into account the effect of the substrate, enzyme, and inhibitor concentrations and applies to the situation of both standard and tight binding inhibition.
k cat /K m values were determined from first-order full-time reaction curves (S Ͻ Ͻ K m ; S ϭ 0.5 M), fitted with the integrated Michaelis-Menten Equation 1, by nonlinear regression, where k ϭ (k cat /K m )⅐E, P represents product concentration, and S 0 is substrate concentration at t ϭ 0, and E is enzyme concentration. MMP concentrations were determined by titration experiments using a phosphinic peptide inhibitor displaying K i values lower than 0.1 nM toward MMPs. Cleavage site identification for the hydrolysis of Mca-Pro-Leu-Gly-Leu-Glu-Glu-Ala-Dpa-NH 2 by MMP-12, -9, and -13 was performed by HPLC analysis (Atlantis dC 18 5; 4.6 ϫ 150 mm; Waters) coupled to an ionic trap (Esquire HCT, Bruker Daltonics). These experiments demonstrated that these MMPs cleaved only the peptide bond located between the Gly-Leu sequence of this substrate.

Molecular Modeling
The interaction of compound 1 with the catalytic domain of MMP12 was modeled using a protocol based on molecular dynamics with the program CHARMM (version 27) (34). CHARMM force field version 22 was used (35). The starting three-dimensional structure of MMP-12 was that of the catalytic domain of this enzyme in complex with an acetohydroxamic acid inhibitor obtained by x-ray diffraction, Protein Data Bank code 1Y93 (20). The initial position of compound 1 was defined by superimposing the catalytic domain of MMP-11 complexed with the phosphinic inhibitor RXP03, Protein Data Bank code 1HV5 (36) on that of the catalytic domain of MMP-12. The structure of compound 1 was then obtained by modification of the structure of RXP03 using the BUILDER facility of InsightII (Accelrys Inc.). Geometrical and nonbonded parameters for the phosphinic inhibitor were derived from ab initio quantum calculations with the program GAUSSIAN98 (37). SCHEME 1 These calculations were done at the MP2 level of theory using a 6 -31 ϩ G(d,p) basis set. The force constants for the phosphinic inhibitor were approximated to 5 kcal mol Ϫ1 rad Ϫ2 , 50 kcal mol Ϫ1 rad Ϫ2 , and 500 kcal mol Ϫ1 Å Ϫ2 for the dihedral, bond angle, and bond distance terms, respectively. In order to preserve the structure of the protein during the relaxation of the complex, harmonic restraints were applied to the atomic positions of several sets of atoms. The harmonic constants were set to 30, 10, and 5 kcal mol Ϫ1 Å Ϫ2 for the cations (zinc and calcium) and the residues chelating these cations, the nonhydrogen backbone atoms, and the nonhydrogen side chain atoms, respectively. No harmonic restraints were applied to the residues whose atoms were located at a distance greater than 5 Å from the inhibitor's atoms. In addition, a set of distance restraints between the phosphinic inhibitor and the catalytic domain MMP-12 was used in order to preserve the geometry of the hydrogen bond between the main chain of the inhibitor and the protein.
During the calculations, the nonbonded interactions were modeled using a Lennard-Jones function and a coulombic electrostatic term with a nonbonded cut-off of 15 Å. The dielectric constant was set to 4. The first step of the relaxation protocol consisted of an initial 2000 cycles of adopted basis Newton-Raphson energy minimization. Then 100,000 steps of molecular dynamics using the Verlet algorithm were calculated. The integration step was set to 0.0005 ps. The temperature was gradually increased by 25 K every 100 steps to reach 450 K. These molecular dynamics calculations were followed by 5000 cycles of energy minimization. The resulting structure was then analyzed with the program PyMOL (Delano).

RESULTS
Four libraries, with the general formula p-Br-Ph-(PO 2 -CH 2 )-XaaЈ-YaaЈ-ZaaЈ-NH 2 , were synthesized by combinatorial chemistry. They were designed by selecting four different isoxazole side chains in the XaaЈ position of the inhibitor to probe the SЈ 1 cavity of MMPs (Scheme 1).
The size of the side chain in the XaaЈ position increased from library 1 to 3. In libraries 3 and 4, the side chain in the XaaЈ position is of similar size, but, due to para or ortho substitution on the phenyl group attached to the isoxazole ring, the second phenyl group (chlorophenyl or phenoxy) adopts a different orientation, thus probing a different part of the SЈ 1 cavity. Each library was prepared as a set of 19 sublibraries, each containing a single variable amino acid in the YaaЈ position, whereas 19 different amino acids in an equimolecular mixture were present in the ZaaЈ position. The determination of the relative potency displayed by each sublibrary enables determination of the preference of MMPs for the YaaЈ position (corresponding to the PЈ 2 position of the inhibitor) in each library.
Role of the PЈ 1 Position in Inhibitor Selectivity-Sublibraries tested at 1 M displayed weak inhibition toward MMP-1 and MMP-7; thus, data for these two MMPs are not shown in Figs. 1 and 2. This low potency is probably related to the presence of a long side chain in the inhibitor PЈ 1 position. In contrast to most MMPs, these two MMPs are known to possess a shallow SЈ 1 cavity (1). Data reported in Figs. 1 and 2 indicate that the presence of a long side chain in the XaaЈ position results in the preparation of inhibitor sublibraries exhibiting high potency toward most of the MMPs tested. Sublibrary potency clearly depends on the size of the R side chain and the MMP tested ( Fig. 1; for clarity, only the results for the sublibrary YaaЈϭPhe are reported in this figure). For MMP-12, inhibitors harboring the R 3 side chain display the highest potency, whereas the lowest potency was observed for inhibitors containing the R 4 group. R 1 and R 2 side chains are well accommodated by MMP-12, with the R 2 substituent providing more selective inhibitors for MMP-12. The preference of MMP-12 for the R 3 group was also observed for MMP-2 and -11, but for these two MMPs, the corresponding inhibitors displayed a reduced potency, as compared with MMP-12 (Fig. 1). MMP-8, -9, -13, and -14 form a group in which the R 1 side chain provides the most potent inhibitors. In this group, inhibitors with R 4 exhibited low potency toward MMP-8 and -9, with MMP-14 being an extreme case. MMP-3 displays a unique inhibition profile, with a preference for R 4 over R 1 and R 2 (Fig. 1). Based on this preference, many sublibraries of inhibitors harboring this R 4 group display good selectivity toward MMP-3 (Fig.  2, fourth panel). But given the weak affinity observed for the corresponding sublibraries, this library was not characterized further. Overall, Fig. 1 illustrates that the presence of the R 2 and R 3 side chains in the inhibitor PЈ 1 position yields compounds that exhibit some selectivity for MMP-12. The same conclusion can be reached by analyzing data reported in Fig. 2.

Role of the PЈ 2 Position in Inhibitor
Selectivity-For most MMPs, the presence of an aromatic residue in the PЈ 2 position of the inhibitor enhanced potency, yielding sublibraries that displayed low selectivity (Fig. 2). Proline in the PЈ 2 position yielded sublibraries exhibiting low potency, regardless of the nature of the R group in the PЈ 1 position. As mentioned above, due to the presence of the R 2 side chain, library 2 contains many sublibraries exhibiting selectivity for MMP-12 (Fig. 2,  second panel). However, inspection of the library 2 diagrams (Fig. 2, second panel) clearly shows that the nature of the residue in the PЈ 2 position influences the selectivity of sublibraries toward MMP-12. The highest selectivity for MMP-12 was observed when the PЈ 2 position was occupied by a glutamate. Even sublibraries containing a closely related residue such as aspartate and asparagine in the PЈ 2 position appeared less selective. In fact, the presence of a glutamine in the PЈ 2 position gave a sublibrary exhibiting higher potency toward most MMPs, thus decreasing its selectivity toward MMP-12. The sublibrary containing an aspartate was less potent toward MMP-12, so its selectivity was reduced, as compared with the sublibrary with a glutamate.
In library 3, the presence of a glutamate in the PЈ 2 position of the inhibitor also resulted in a sublibrary exhibiting selectivity toward MMP-12, but in this case the sublibrary also exhibited good potency toward MMP-13 (Fig. 2, third panel). As shown in Fig. 1, MMP-13 accommodates the R 3 side chain very well in the PЈ 1 position of the inhibitor. Given the good selectivity toward MMP-12 displayed by the mixture of compounds contained in library 2, in particular when the PЈ 2 position is occupied by a glutamate, phosphinic peptides of the general formula p-Br-Ph-(PO 2 -CH 2 )-XaaЈ-Glu-ZaaЈ-NH 2 were prepared, with the XaaЈ position occupied by the isoxazole side chain R 2 and the ZaaЈ position occupied by 19 different amino acids. Similar peptides with the XaaЈ position occupied by the R 3 side chain were also synthesized to identify selective MMP-12 inhibitors.
Role of the PЈ 3 Position in Inhibitor Selectivity-When the R 2 chain was present in the XaaЈ position, the results reported in Fig. 3a show that the presence of a glutamate in the PЈ 2 position gives several inhibitors displaying high selectivity toward MMP-12, in particular when the ZaaЈ position is occupied by Asn, Gln, Asp, and Glu, the last inhibitor showing the highest selectivity. Interestingly, selectivity toward MMP-12 is decreased when the ZaaЈ position is occupied by basic residues. Similarly, with the R 3 side chain in the XaaЈ position, the presence of Asn, Asp, Gln, and Glu in the ZaaЈ position increases the selectivity of the inhibitors toward MMP-12 (Fig. 3b), leading to a clear differentiation between MMP-12 and MMP-13 (Figs. 2c and 3b).
Compounds 1 and 2 Display High Selectivity toward MMP-12-Due to the presence of an asymmetric center at the PЈ 1 position of the inhibitor, all of the phosphinic peptides reported in this study were prepared as a mixture of two diastereoisomers. Thus, to report potency of the inhibitor in optically pure form, two inhibitors were selected (compounds 1 and 2, Table 1) and purified by HPLC to resolve these diastereoisomers. Only data for the diastereoisomer exhibiting the shorter retention time in HPLC were reported in Table 1, since this diastereoisomer exhibits the highest selectivity for MMP-12, both for compounds 1 and 2. A previous study has showed that for this family of isoxazole derivatives, the diastereoisomer eluting first in HPLC possesses the S configuration in the PЈ 1 position (24). Compound 1 with the R 3 group behaves as a highly potent inhibitor of MMP-12 and displays selectivity factors from 2 to 3 orders of magnitude when tested with other MMPs. Compound 2 is less potent toward MMP-12 but displays a similar range of selectivity. Each compound has its own selectivity profile, so, depending on the envisaged application, one compound might be more useful than the other. These results demonstrate that in addition to the size of the isoxazole side chain in the PЈ 1 position, also the presence of a Glu-Glu motif induces selectivity for MMP-12. Moreover, compounds 1 and 2 behave as very weak inhibitors of TACE, NEP, and ACE.
Selective Substrate of MMP-12-To assess further whether a Glu-Glu sequence could be a motif leading to a preferential recognition by MMP-12, this motif was introduced into the sequence of a fluorogenic

TABLE 1 K i values of compounds 1 and 2 towards MMPs (R is defined in Scheme 1)
peptide. The design of this substrate was based on the structure of a fluorogenic peptide classically used to monitor MMP activity (38) (Scheme 2). MMP-12 was observed to cleave the Glu-Glu substrate with very high efficiency (Fig. 4) and exclusively between the Gly-Leu sequence of this substrate (see "Experimental Procedures"). Thus, when this substrate is cleaved by MMP-12, the Glu-Glu motif interacts with the SЈ 2 and SЈ 3 subsites of MMP-12.
Mca-Pro-Leu-Gly2Leu-Dpa-Ala-Arg-NH 2 Mca-Pro-Leu-Gly2Leu-Glu-Glu-Ala-Dpa-NH 2 SCHEME 2 In contrast, with the exception of MMP-9 and MMP-13, the Glu-Glu-containing peptide turned out to be a poor substrate of other MMPs, a result illustrating that the Glu-Glu motif may also govern the selectivity of synthetic substrates toward MMP-12 (Fig. 4). Based on the values of the catalytic efficiency parameters (k cat /K m ) reported in Fig. 4, the Glu-Glu substrate is cleaved 3.5 and 5 times more rapidly by MMP-12, as compared with MMP-13 and MMP-9, respectively.
Model of Compound 1 Interacting with MMP-12-A model of compound 1 interacting with MMP-12 was developed to map the residues of MMP-12 that are in close proximity to the Glu-Glu side chains of the inhibitor (Fig. 5). This model reveals that the Glu in the PЈ 2 position is close to Thr 239 . Interestingly, with the exception of MMP-7 and MMP-1, MMPs possess bulky hydrophobic residues in place of Thr 239 in MMP-12. As mentioned above, tight interaction of compound 1 with MMP-1 and MMP-7 is prevented due to the presence of a long side chain in the PЈ 1 position of this inhibitor; thus, the presence of a small polar side chain in these two MMPs, in a position equivalent to Thr 239 in MMP-12, cannot counterbalance the effect of the PЈ 1 residue in this inhibitor.
This model also indicates that the Glu in the PЈ 3 position of compound 1 is in close proximity with Lys 177 . With the exception of MMP-2 and MMP-9, most other MMPs do not have a lysine at the MMP-12 equivalent position but a proline. However, in MMP-2 and MMP-9, the lysine residue is followed by an aspartate, whereas in MMP-12 a glycine was observed after the lysine. Inspection of the x-ray structures of MMP-2 and -9 reveals that the aspartate residue in MMP-2 and MMP-9 points toward the PЈ 3 glutamate of the selective inhibitor. Thus, based on this observation, it is tempting to speculate that whereas a favorable interaction between the glutamate of compound 1 and Lys 177 of MMP-12 seems possible, similar interactions would be prevented by the presence of the aspartate residue in MMP-2 and MMP-9, explaining the low potency of compound 1 toward MMP-2 and MMP-9. Inspection of the SЈ 1 loop reveals that the PЈ 1 side chain of compound 1 is in close contact with several residues of this loop. However, two positions occupied by Lys 241 and Val 243 may deserve particular interest, since a great variability in amino acid composition between the different MMPs is observed for these two positions ( Table 2).

DISCUSSION
This study demonstrates that the screening of a limited number of phosphinic peptide sublibraries allows the identification of highly potent and selective inhibitors of MMP-12. In contrast to some hydroxamate derivatives developed for MMP inhibition, the inhibitors reported in this study behave as weak inhibitors of TACE, NEP, and ACE (39). The size of the PЈ 1 residue contributes significantly both to the selectivity and potency of this family of compounds toward MMPs. However, in the particular case of MMP-12, this study demonstrates that the inhibitor selectivity, in addition to the PЈ 1 residue, also depends on the nature of the side chain present in the PЈ 2 and PЈ 3 positions.
Elongation of the R side chain from R 1 to R 2 reduces the inhibitory potency toward MMPs (Fig. 1). This result suggests that some steric hindrance within the MMP SЈ 1 cavity may interfere with the binding of the R 2 group. However, since further elongation of the R group from R 2 to R 3 increases inhibitory potency toward MMPs (Figs. 1 and 2), the reduced potency displayed by the inhibitors in library 2 should involve factors other than steric hindrance. Inspection of the SЈ 1 cavity for most MMPs screened in this study does not reveal the presence of particular residues occluding the cavity, MMP-8 being the sole exception, with an arginine pointing toward the center of the SЈ 1 cavity (Arg 222 , MMP-8 numbering). This property may explain the preference of MMP-8 for inhibitors harboring the shortest R group in this series. According to the R side chain preference, MMPs can be assembled in two different groups, one exhibiting a preference for the R 3 side chain and the other for R 1 (Fig. 1). The model of compound 1 in complex with MMP-12 reveals that among the residues of the SЈ 1 loop that are in close proximity to the PЈ 1 residue of compound 1, two may play a particular role in the preferential recognition of the PЈ 1 residue. These two residues, colored in purple in Fig. 5, are of interest because they are not conserved between  MMPs ( Table 2). The great diversity of residues in these two positions of the SЈ 1 loop may influence the binding to MMPs of inhibitors harboring long PЈ 1 group. However, no simple correlation could be found between these SЈ 1 residues and the data reported in Figs. 1 and 2. As mentioned in the Introduction, the great flexibility occurring at the level of the SЈ 1 loop in most MMPs makes it difficult to predict all of the interactions that may take place between PЈ 1 group and the different residues of the SЈ 1 loop that point toward the SЈ 1 cavity. This flexibility, confirmed by recent NMR studies (19,20), severely limits the use of MMP crystal structures to fully explain the factors contributing to inhibitor potency and selectivity. Mutagenesis experiments will be helpful in assessing the role of these residues in inhibitor recognition. Strikingly, few studies dealing with the role of active site residues of MMPs have been reported (40,41), and to the best of our knowledge there are none concerning the SЈ 1 loop residues. Results from such studies may facilitate the selection of the PЈ 1 side chains that should be introduced into the inhibitor structure to optimize selectivity. The selective MMP-12 inhibitors reported in this study contain a Glu-Glu motif in their sequence. The side chains of these two Glu are in close proximity to the side chains of Thr 239 and Lys 177 , according to our model of compound 1 in interaction with MMP-12 (Fig. 5). The model suggests that the selectivity of compound 1 might be explained either by a direct interaction between the Glu and Thr side chains or alternatively that a hydrophobic residue in position 239, as observed in many MMPs, constitutes an unfavorable environment for the Glu side chain. The fact that Asp and Gln in PЈ 2 also give rather selective sublibraries argues in favor of the second hypothesis. The same arguments can be used to explain the preference of MMP-12 toward inhibitors possessing a Glu in the PЈ 3 position. Determination of the MMP-12 structure in complex with compound 1 or 2, together with mutagenesis experiments, will be necessary to definitively establish the role of Thr 239 and Lys 177 in MMP-12 selectivity. The catalytic efficiency displayed by MMP-12 in cleaving the fluorogenic substrate incorporating the Glu-Glu motif is another observation that illustrates the pivotal role played by this motif in the preferential recognition of a ligand by MMP-12. Further optimization of the sequence of this peptide should lead to substrates exclusively cleaved by MMP-12.
To the best of our knowledge, compounds 1 and 2 are the first examples of highly potent and selective MMP-12 inhibitors (42)(43)(44). Several studies have assessed the functional role of MMP-12 in different diseases, using MMP-12-deficient mice. Based on these studies, MMP-12 was proposed to play a key role in the development of emphysema (45)(46)(47), atherosclerosis (48 -50), and abdominal aortic aneurysm (51,52). The role of MMP-12 in cancer progression is more elusive (53)(54)(55). Given the implication of MMP-12 in different pathologies, it will therefore be of great interest to see whether the development of the corresponding diseases in animal models can be prevented with compounds 1 and 2.
Screening of a few phosphinic libraries has allowed the identification of highly selective MMP-12 inhibitors. As compared with the hydroxamates, these compounds do not block other proteases such as TACE, ACE, and NEP. These results suggest that selective MMP inhibitors can be selected by screening new phosphinic peptide libraries.