A residue in the S2 subsite controls substrate selectivity of matrix metalloproteinase-2 and matrix metalloproteinase-9.

Matrix metalloproteinase (MMP)-2 and MMP-9 are closely related metalloproteinases that are implicated in angiogenesis. The two proteins have a similar domain structure and highly homologous catalytic domains, making them an excellent comparative model for understanding the structural basis of substrate recognition by the MMP family. Although the two MMPs exhibit some overlap in substrate recognition, our recent work showed that MMP-2 can cleave a set of peptide substrates that are only poorly recognized by MMP-9 (Chen, E. I., Kridel, S. J., Howard, E. W., Li, W., Godzik, A., and Smith, J. W. (2002) J. Biol. Chem. 277, 4485-4491). Mutations at the P(2) position of these peptide substrates dramatically reduced their selectivity for MMP-2. Inspection of the corresponding S(2) pocket of the substrate-binding cleft of the protease reveals that MMP-9 contains an Asp, whereas MMP-2 contains Glu. Here, we test the hypothesis that this conservative substitution has a role in substrate selectivity. Mutation of Glu(412) in MMP-2 to Asp significantly reduced the hydrolysis of selective substrates, with only a minor effect on hydrolysis of non-selective substrates. The predominant effect of the mutation is at the level of k(cat), or turnover rate, with reductions reaching as high as 37-fold. The residues that occupy this position in other MMPs are highly variable, providing a potential structural basis for substrate recognition across the MMP family.

The matrix metalloproteinase (MMPs) 1 family consists of over 25 secreted and cell surface proteases (1). The MMPs are involved in a wide range of normal biologic processes and have been linked to a number of pathologic events (2,3). Because of the associations with disease, a number of synthetic MMP antagonists have been developed and tested in human clinical trials (2, 4 -7). Unfortunately, however, these trials did not succeed (8). One reason for the lack of success probably relates to the relatively broad spectrum of inhibition by the compounds that were tested. For example, the first compound to enter the clinic, Marimastat, has a similar affinity for at least five MMPs. The lack of selectivity of most of the small molecule MMP inhibitors can be traced to the fact that they were designed to exploit features common to all MMPs. These include a zinc ion and the catalytic glutamic acid (9), and especially, the deep S 1Ј pocket (10 -15). With these ideas in mind, one of the objectives of the present study was to identify regions of the MMP catalytic pockets that contribute to substrate distinction and that could be exploited for the future design of highly selective MMP antagonists.
Information on substrate selection by the MMPs could also help explain the unique biological roles of these proteases. MMPs are no longer looked upon as proteases whose action is limited to the destruction of collagen and gelatin in basement membranes. Rather, the MMPs are involved in the control of a number of events where proteolysis must be exerted in a precise manner. For example, MMPs can activate other MMP zymogens (proMMPs) by hydrolyzing propeptide bonds (16). They also influence that activity of the serine proteases by selectively degrading their macromolecular inhibitors such as serpin ␣1-proteinase inhibitor (17). MMPs can also influence that action of growth factors by activating and releasing them for their functions (18). Some MMPs potentiate the action of growth factors by selectively degrading growth factor-binding proteins (19). All of these functions are presumed to involve a relatively high degree of selectivity among the MMPs for their physiologic substrates. There are also circumstances in which closely related MMPs are present in the same locale but have different roles. For example, both MMP-2 and MMP-9 are expressed within tumors, but only one participates in the angiogenic switch (20). Similar distinctions in the role of these MMPs have also been observed in platelet function (21) and in cell migration (22). These findings strongly suggest that MMP-2 and MMP-9 operate by cleaving distinct substrates.
Despite the expanding awareness of selective hydrolysis by the MMPs, the structural basis for such selectivity is still not well understood. In a prior report, we used substrate phage display to gain an in-depth understanding of the substrate recognition properties of MMP-2. We identified a surprising number of highly selective substrates for MMP-2 (23). Interestingly, the selective substrates could be segregated into three subfamilies based on their sequences. However, within each family, the P 2 residue of substrate was key to selectivity for MMP-2 over MMP-9. The S 2 pocket, which interacts with the P 2 residue in substrate, is remarkably similar in both enzymes (14), save for the presence of Glu 412 in MMP-2, which is replaced by an Asp in MMP-9. In fact, our modeling work suggested that Glu 412 of MMP-2 forms a hydrogen bond with the backbone of the selective substrates, a bond that cannot be formed by the shorter Asp in MMP-9.
Here we test the importance of the Glu/Asp alteration in substrate recognition. Mutation of Glu 412 to Asp has a signifi-cant influence on the substrate recognition by MMP-2. The substitution causes a significant reduction in k cat /K m for peptide substrates that are selective for MMP-2 over MMP-9. The reduction in k cat /K m is predominantly driven by a reduction in k cat or the rate of turnover of these substrates. Interestingly however, this Glu/Asp substitution had far less of an effect on substrates that were recognized equally well by the two MMPs. In these cases, only a minor reduction in k cat (and correspondingly, k cat /K m ) was observed. The deleterious effect of the Glu/ Asp substitution on hydrolysis of selective substrates was also observed on EphB1, a protein substrate that we have found to be selectively cleaved by MMP-2 in vitro. The results of the Glu 412 /Asp mutation provide further proof that selective and non-selective substrates for MMP-2 dock via different contact points within the catalytic cleft. Altogether, the findings of this report substantiate the hypothesis that Glu 412 in the S 2 subsite of the MMP-2 is a key residue in conferring hydrolysis of selective substrates. The study also raises the interesting hypothesis that selectivity between MMP-2 and MMP-9 arises because MMP-9 is a "slower" protease than MMP-2.

EXPERIMENTAL PROCEDURES
Source of Commercial Proteins and Reagents-Ilomastat was purchased from AMS Scientific (Concord, CA). Restriction enzymes were from New England Biolabs. Oligonucleotides were synthesized by Integrated DNA technologies, Inc. (Coralville, IA). Tissue culture media and reagents were from Irvine Scientific (Irvine, CA). All other reagents, chemicals, and plastic ware were from Sigma or Fisher.
Docking of the MMP-2 Peptide Substrates into the Catalytic Domains of MMP-2 and MMP-9 -The complexes between MMP-2 and the peptide substrates were modeled using the Sybyl package from Tripos (www.tripos.com). The coordinates of MMP-2 are from the crystal structure (Protein Data Bank accession number 1qib). For docking, the initial structure of each peptide was built in an extended conformation within the catalytic cleft of the protease. Since the binding pocket of the enzyme is known, and some interactions between the protein and site around P 1 Ј of the substrate are obvious, these interactions were used as restrictions in the modeling procedure. The anchoring restrictions included a hydrogen bond between the NH group of P 1 Ј and the oxygen atom of Ala 165 , the hydrogen bond between the oxygen atom of P 1 Ј and the NH group of Ala 165 , the interaction between the side chain of P 1 Ј, and the hydrophobic patch of MMP-2 comprised of Val 198 and its surroundings. The substrate was first placed at the binding cleft at a position favoring the formation of bonds making up these restrictions. Under these constraints, the initial conformation of the substrate resembled an antiparallel ␤-strand lying against the neighboring ␤-strand in the catalytic cleft of MMP-2. Then a conformation search of the substrate was performed with only two restrictions: (i) P 1 Ј was maintained as the anchor point and (ii) the center (C␣ atom) of P 1 Ј remained fixed during the modeling. Since there are no known correlations between the conformation of substrate on opposite sides of the P 1 Ј position (the anchoring point), the conformational searches of the N terminus and C terminus were run separately. To reduce the complexity of computation, the searches were done in torsion angle space, where the bond length and bond angles of the whole substrate remained fixed. The model of substrates within the cleft of MMP-9 were generated in a similar manner, except that the structure of MMP-9 was first modeled on the crystal structure of MMP-2 as described previously (23).
Site-directed Mutagenesis of MMP-2-The mutations of Glu 412 to Asp in the MMP-2 catalytic domain was generated with the QuikChange site-directed mutagenesis kit from Stratagene (San Diego, CA). Oligo-nucleotides used for generating mutated MMP-2 were 5Ј-AAG CTT ATG GAG GCG CTA ATG GCC CG-3Ј and 5Ј-CGC CAT GGG GCT GGA TCA CTC CCA AGA CCC-3Ј. Each pair of oligonucleotides was annealed to the denatured template plasmid encoding the catalytic domain of MMP-2. Subsequently, the mutated MMP-2 was generated by PCR amplification of cDNA from these complexes. The mutations were verified by sequencing the newly generated catalytic cDNA. The mutated MMP-2 was subcloned into the pCDNA3 expression vector (Invitrogen).
Expression and Purification of the Catalytic Domain of Mutated MMP-2-Briefly, the cDNA encoding the mutated catalytic domain of MMP-2 was used to transfect HEK 293 cells. Individual antibioticresistant clones were isolated with cloning discs, expanded, and then screened by reverse transcription-PCR and zymography. The mutated catalytic domains of MMP-2 were purified from conditioned medium by the method described previously (23). The purity of the mutant MMP-2 was greater than 90% judging by silver-stained acrylamide gels. The purified mutant MMP-2 was stored at Ϫ70°C at concentrations of ϳ0.3 mg/ml. For kinetic studies, the mutated MMP-2 was activated and active site-titrated as described previously (23).
Quantifying the Kinetic Parameters of Peptide Hydrolysis-The kinetic parameters of substrate hydrolysis were measured using a fluorescamine incorporation assay (24). The method for determining the K m and k cat of peptide hydrolysis was described previously (23).
Assessing the Extent of Hydrolysis of EphB1 by Mutated MMP-2-Recombinant fusion proteins between EphB1, EphB2, and the Fc domain of IgG were purchased from R&D systems Inc. The fusion proteins (1.8 M) were incubated for 4 h at 37°C with 280 nM of wild type MMP-2, wild type MMP-9, or MMP-2 E412D . Following incubation, samples were resolved by 10% SDS-PAGE, and samples were visualized by Coomassie staining.

Mutagenesis of Glu 412 to Asp in the Catalytic Domain of MMP-2-Previously
, we observed that the residue at the P 2 position within peptide substrates has a major role in determining the selectivity of the peptide for MMP-2 versus MMP-9 (23). A key distinction between MMP-2 and MMP-9 is the substitution of a Glu for an Asp within the S 2 subsite, and we suggested that this substitution could account for differences in substrate recognition. To test this hypothesis, we mutated Glu 412 to Asp in a construct encoding the catalytic domain of MMP-2. The mutated MMP-2 was expressed in HEK 293 cells as described under "Experimental Procedures." The mutated MMP-2 was expressed at levels similar to wild type MMP-2 and could be purified by affinity chromatography on gelatin-Sepharose. To ensure that kinetic parameters were accurately calculated, the amount of properly folded MMP was quantified by active site titration with the small molecule inhibitor, Ilomastat, as described under "Experimental Procuedures." Both proteins had similar levels of catalytic activity per mg of purified protein.
Assessing the Ability of Mutated MMP-2 to Cleave the Nonselective and Selective Peptide Substrates-We measured the ability of wild type and MMP-2 E412D to hydrolyze non-selective peptides substrates that contain the PXXL motif (Table I). Only minor changes were observed in the k cat /K m values for peptides such as C15 and m1A11, which display no selectivity between MMP-2 and MMP-9 (Table I). In contrast, however, the E412D mutation had significant and consistent effects on the The hydrolysis of synthetic peptides by each MMP was quantified using procedures outlined under "Experimental Procedures." For each protease, hydrolysis was measured across a concentration range of peptide. Values for k cat and K m were derived from Lineweaver-Burk plots. Each measurement was repeated at least three times and the relative error of k cat /K m in these measurements was less than 10%. hydrolysis of a panel of peptides that were found previously to be selective for MMP-2 (Table II). Each of the selective peptides has a higher k cat /K m ratio for MMP-2 than for MMP-9. In all cases, the mutation of Glu to Asp caused a decrease in the k cat /K m such that the ratio for the mutant falls between the value for MMP2 and MMP-9. This observation is consistent with the fact that the E412D mutation alters the structure of the catalytic pocket to more closely resemble that of MMP-9. Interestingly, the primary effect of the mutation seemed to be at the level of k cat , a parameter that was decreased for all of the selective peptides. Decreases in k cat ranged from 3 to 36fold, in most cases approximating the values we have measured for MMP-9. In contrast, the effects of the E412D mutation on K m were less consistent. We observed this parameter to increase and decrease depending on the peptide being tested. Consequently, the E412D mutation in the S 2 pocket appears to primarily affect the rate of substrate turnover.
One peptide was identified that has an increased k cat /K m ratio for the E412D mutant of MMP-2. This peptide, A13R, was originally synthesized and characterized in our prior report on substrates selective for MMP-2 (23). The parent peptide, A13, contains the SX 2 L motif that is selective for MMP-2, but A13R contains the RX 2 L sequence and is cleaved better by MMP-9. Here, we compared the rate of hydrolysis of the A13R peptide with all three of the MMPs (Fig. 1). In three independent replications of this experiment, the rate of peptide hydrolysis by MMP-9 and by MMP-2 E412D was higher than that of wild type MMP-2. This finding provides further support for the idea that mutation of Glu 412 to Asp shifts the substrate recognition of MMP-2 closer to that of MMP-9. Because the hydrolysis of this peptide by MMP-2 E412D is better than that of wild type MMP-2, the observation also excludes the possibility that the mutation within MMP-2 simply causes general diminution in proteolytic activity.
Comparing the Hydrolysis of EphB1 by Wild Type and Mutated MMP-2-The EphB1 receptor tyrosine kinase contains the SXL motif and is selectively cleaved by MMP-2 (23). In fact, we found that this protein is generally resistant to hydrolysis by MMP-9. Here we tested the ability of MMP-2 E412D to hydrolyze EphB1 (Fig. 2). The EphB1 fusion protein was incubated with equimolar amounts (280 nM) of wild type MMP-2, mutated MMP-2, and MMP-9 for 4 h. The quantity of each protease was measured by active site titration prior to initiating the experiment. The extent of hydrolysis of EphB1 was gauged by SDS-PAGE (Fig. 2). The EphB1-Fc fusion protein was almost quantitatively cleaved by MMP-2 (Fig. 2, lane 2). As expected, MMP-9 had no effect on the migration of the protein, indicating a lack of hydrolysis (Fig. 2, lane 4). Interestingly, the hydrolysis of the EphB1-Fc fusion protein by MMP-2 E412D fell between that of wild type MMP-2 and MMP-9 (Fig. 2, lane 3). This result is consistent with the observations made with the panel of peptide substrates and corroborates the conclusion that the E412D mutation shifts the phenotype of the protease away from that of MMP-2 and toward that of MMP-9. DISCUSSION MMP-2 and MMP-9 are considered to be close homologs because of their unique domain structure and their high sequence similarity and because they both have gelatinase activ-FIG. 1. Effect of the E412D mutation on hydrolysis of peptide substrates. A13R peptide was cleaved by MMP-2 (f), MMP-2 E412D (q), and MMP-9 (OE), and the extent of hydrolysis for individual protease was measured by relative fluorescent units (RFUs) at 2-min intervals up to 10 min. Hydrolysis of A13R showed that the activity of MMP-2 E412D resembles MMP-9 rather than the wild type MMP-2. Four independent replicates were performed for each time point, and standard deviation was less than 10% as indicated by the error bars.  The hydrolysis of a panel synthetic peptides previously shown to be selective for MMP-2 (23) was quantified using procedures outlined in "Experimental Procedures." For MMP-2 and mutant MMP-2, hydrolysis was measured over a concentration range of peptide such that values for k cat and K m were derived from Lineweaver-Burk plots. For MMP-9, the k cat /K m ratio was measured as described in the previous paper (23). Each measurement was repeated at least three times and the relative error for k cat /K m and for k cat independently less than 10%. ity. Nevertheless, the two proteases clearly have distinct biological functions. In many instances, the two proteases have different effects even when they are analyzed in the same biological system (21,22). Consequently, it follows that MMP-2 and MMP-9 are likely to cleave different substrates.
We have recently begun to examine differences in the way that MMPs recognize peptide substrates. This analysis places focus on structural and functional distinctions at the catalytic cleft, a region that has generally been overlooked as a feature that can distinguish one MMP from another. In fact, early comparisons of peptide substrates for MMP-2 and MMP-9 showed that they each cleaved the same set of collagen-like peptide substrates (25), a finding that indicates structural and functional similarity at the catalytic pocket. Nevertheless, we recently made the observation that MMP-2 and MMP-9 recognize distinct sets of peptide substrates (23). We found that both MMP-2 and MMP-9 hydrolyze peptides with the canonical recognition motif PXXX Hy , as originally described by Netzel-Arnett et al. (25), but we also found MMP-2 to uniquely recognize three additional sets of peptide substrates. Like the canonical motif, the three other families of substrates all contain a hydrophobic residue at the P 1 Ј position, but they lack the characteristic proline at P 3 . In an extension of this study, we found that the substrates selective for MMP-2 could be converted to substrates for MMP-9 by inserting Arg into the P 2 position (23). This finding underscored the significance of the P 2 position within substrate and strongly suggested that the corresponding S 2 subsite within the catalytic clefts of MMP-2 and MMP-9 must interact differently with substrate. Interestingly, the S 2 subsite contains one of the few differences between the catalytic clefts of the two MMPs; Glu 412 of MMP-2 is replaced by Asp 410 in MMP-9. The objective of the present study was to test the hypothesis that this conservative substitution accounts for the distinct substrate recognition profiles of the two enzymes.
We swapped this residue by mutating each protease. Although we were able to express the mutated MMP-2, the mutant MMP-9 could not be expressed (see below). Consequently, most of our discussion focuses on results obtained with the mutant of MMP-2. The mutated MMP-2 was expressed to high levels in HEK 293 cells, could be purified on gelatin-agarose affinity columns, and exhibited gelatinase activity in zymography gels. In addition, the E412D mutant hydrolyzed substrates with the canonical PXXX Hy recognition motif to essentially the same degree as wild type MMP-2. This is to be expected since MMP-9, which displays the Asp rather than Glu at this position, also hydrolyzes this subfamily of substrates with high efficiency. In conjunction with the fact that the mutation is conservative and represents a sequence found in the closest homolog, these observations allow us to exclude the possibility that substitution of Glu for Asp at position 412 causes a general diminution of catalysis by forcing improper folding.
The E412D mutation had substantial effects on the hydrolysis of peptide substrates that are selective for MMP-2. The mutation significantly reduced the k cat /K m ratio for all of the MMP-2-selective substrates. Reductions to the k cat /K m ratio were primarily driven by reductions to the turnover rate, k cat , which was reduced in every case. Changes in both directions were observed for K m , so the E412D mutation has no uniform influence on binding affinity. The E412D mutation also extended to larger protein substrates. We recently demonstrated that the extracellular domain of the EphB1 receptor tyrosine kinase is cleaved by MMP-2, but not by MMP-9. The ability of the E412D mutant of MMP-2 to cleave EphB1 fell in between that of wild type MMP-2 and MMP-9, indicating the mutation shifted the function of the protease closer to that of MMP-9. This is taken as another indication that the Glu 412 /Asp 410 alteration has a role in substrate distinction.
Altogether the results of this study suggest a structural basis for the role of the S 2 subsite in substrate recognition and catalysis. In large part, our results support the idea that the S 2 subsite can engage in two types of interactions that depend on the composition of the substrate (Figs. 3 and 4). When Arg is present at P 2, substrates show selectivity for MMP-2 and MMP-9 over other MMPs (26). In this case, the Arg in the substrate extends far enough into the S 2 subsite to interact favorably with the side chains of both Asp in MMP-9 or the Glu in MMP-2 (Fig. 3). Because Arg at this position decreases K m (26), we suggest that the salt bridge that Arg forms with the side chains of Glu or Asp increases binding affinity for the protease (Fig. 3). A second type of interaction at the S 2 subsite is likely with the peptides that lack Arg and are selectively cleaved by MMP-2. In the absence of Arg at P 2 , only the Glu of MMP-2 extends far enough into the S 2 pocket to form a hydrogen bond with the backbone of bound substrate (Fig. 4A) concept that is generally supported by our mutational studies. The corresponding Asp in MMP-9 fails to extend far enough into the pocket to make a similar contact (Fig. 4B). Interestingly, our findings also suggest that a switch to this binding mode in MMP-2 changes the overall effect of the S 2 subsite on substrate recognition. Rather than playing a role in binding affinity, the hydrogen bond between Glu 412 and the backbone of bound substrate influences k cat , indicating that it helps to properly position the substrate for optimal catalysis.
Although our findings strongly support the role of the Glu 412 / Asp 410 substitution in determining substrate recognition by the two MMPs, this mutation alone was not sufficient to completely convert the recognition profile of MMP-2 to that of MMP-9. Consequently other factors must be involved. These are likely to include cooperativity among other subsites where subtle differences exist or a global difference in the shape of the catalytic pocket that is governed by sequence differences outside of the catalytic cleft. A resolution to this particular issue will require a three-dimensional structure for MMP-9.
The importance of the residue corresponding to Glu 412 in the S 2 subsite may extend across the MMP family. Most of the residues surrounding the S 2 subsite are strictly conserved among all MMPs (Fig. 5). The two histidines that flank the subsite and the glycine in the center of the subsite are conserved across all family members. In contrast, the position taken by Glu 412 in MMP-2 is highly variable. It is occupied by acidic residues, large hydrophobic residues, and in some cases, glycine (Fig. 5). The distinctions among these residues are likely to have an influence on substrate recognition by each MMP.
Certainly an analysis of the analogous substitution within MMP-9 would be informative. However, we were unable to express this mutant, although we tested a number of different host cell types. Although the inability to express this mutation may seem surprising given the extensive homology between MMP-2 and MMP-9, there may be a mechanistic basis for this observation. It is well established that the MMPs require the propeptide, which inserts into the catalytic cleft via a cysteine switch, for proper folding (27,28). We suspect that this mutant of MMP-9 cannot be expressed because it interferes with the insertion of the propeptide into the catalytic cleft during synthesis. This hypothesis is based on inspection of the crystal structure of MMP-2 with the propeptide inserted into the cleft. In MMP-2, Glu 412 extends up from the floor of the cleft to within 3-4 angstroms of the side chain of Asn 109 . In MMP-9, the corresponding residues are Arg (longer than Asn) and Asp (shorter than Glu). The Arg in the propeptide of MMP-9 would be expected to extend further into the catalytic cleft than Asn. Such positioning would likely be favorable with Asp in the S 2 subsite. However, mutation of Asp 410 to Glu, which would extend further into the cleft, is likely to interfere with insertion of Arg in the MMP-9 propeptide. Such interference would be likely to prevent proper folding of the mutated MMP-9 and cause its degradation.