A unique substrate recognition profile for matrix metalloproteinase-2.

The catalytic domains of the matrix metalloproteinases (MMPs) are structurally homologous, raising questions as to the degree of distinction, or overlap, in substrate recognition. The primary objective of the present study was to define the substrate recognition profile of MMP-2, a protease that was historically referred to as gelatinase A. By cleaving a phage peptide library with recombinant MMP-2, four distinct sets of substrates were identified. The first set is structurally related to substrates previously reported for other MMPs. These substrates contain the PXX/X(Hy) consensus motif (where X(Hy) is a hydrophobic residue) and are not generally selective for MMP-2 over the other MMPs tested. Two other groups of substrates were selected from the phage library with similar frequency. Substrates in group II contain the L/IXX/X(Hy) consensus motif. Substrates in group III contain a consensus motif with a sequence of X(Hy)SX/L, and the fourth set of substrates contain the HXX/X(Hy) sequence. Substrates in Group II, III, and IV were found to be 8- to almost 200-fold more selective for MMP-2 over MMP-9. To gain an understanding of the structural basis for substrate selectivity, individual residues within substrates were mutated, revealing that the P(2) residue is a key element in conferring selectivity. These findings indicate that MMP-2 and MMP-9 exhibit different substrate recognition profiles and point to the P(2) subsite as a primary determinant in substrate distinction.

The catalytic domains of the matrix metalloproteinases (MMPs) are structurally homologous, raising questions as to the degree of distinction, or overlap, in substrate recognition. The primary objective of the present study was to define the substrate recognition profile of MMP-2, a protease that was historically referred to as gelatinase A. By cleaving a phage peptide library with recombinant MMP-2, four distinct sets of substrates were identified. The first set is structurally related to substrates previously reported for other MMPs. These substrates contain the PXX2X Hy consensus motif (where X Hy is a hydrophobic residue) and are not generally selective for MMP-2 over the other MMPs tested. Two other groups of substrates were selected from the phage library with similar frequency. Substrates in group II contain the L/IXX2X Hy consensus motif. Substrates in group III contain a consensus motif with a sequence of X Hy SX2L, and the fourth set of substrates contain the HXX2X Hy sequence. Substrates in Group II, III, and IV were found to be 8-to almost 200-fold more selective for MMP-2 over MMP-9. To gain an understanding of the structural basis for substrate selectivity, individual residues within substrates were mutated, revealing that the P 2 residue is a key element in conferring selectivity. These findings indicate that MMP-2 and MMP-9 exhibit different substrate recognition profiles and point to the P 2 subsite as a primary determinant in substrate distinction.
The matrix metalloproteinases (MMPs) 1 have important roles in several normal tissue remodeling events (1). The MMPs are also of interest as pharmaceutical targets because of their association with a number of pathological conditions (2).
One key area of interest is tumor angiogenesis, where MMP-2 and MMP-9 appear to have important roles. Mice lacking the gene for MMP-2 exhibit reduced tumor angiogenesis (3), an effect that may be related to the connection between MMP-2 and the ␣ v ␤ 3 integrin in angiogenesis (4). The role of MMP-9 in angiogenesis also appears to be significant, as this protease is reported to be part of the "angiogenic switch" that enacts the vascularization of tumors (5). Despite the structural similarity among these proteases, there are strong indications that their mechanisms of action are distinct in tumor angiogenesis. There are circumstances where both proteases are present within tumors, but only one participates in the angiogenic switch (5). Similar distinctions in the role of these MMPs have also been observed in platelet function (6) and in cell migration (7). At present there is no mechanistic explanation for these distinctions, but they raise the possibility that the two MMPs operate by cleaving distinct substrates.
The factors that govern substrate recognition and substrate distinction by the MMPs have not been fully elucidated. The structural features of the catalytic clefts of all MMPs are generally similar. All of the catalytic clefts contain a zinc ion and glutamic acid residue involved in catalysis (8). Furthermore, the MMPs contain a deep S 1 Ј pocket (9 -14). This subsite has been exploited as a docking point for the vast majority of pharmaceutical inhibitors of the MMPs. Perhaps not surprisingly then, many of these antagonists are broad spectrum inhibitors (15), and show evidence of unwanted side effects (16,17). There is also evidence to indicate that the common structural features of the catalytic cleft lead to an overlap in substrate recognition (18). The vast majority of known MMP substrates contain a large hydrophobic residue, which is frequently leucine, at the P 1 Ј position. This is consistent with the deep S 1 Ј pocket. Further similarity is observed at the P 3 position, where proline is often preferred. In fact, the PXX2X Hy motif appears to be an excellent substrate for a wide range of MMPs (19 -21). However, recent work counters the notion that MMPs have an overlapping substrate recognition profile. Recent studies on both MMP-13 and MMP-9 show that a high degree of selectivity can be obtained for individual MMPs, even among substrates comprised of the PXX2X Hy motif (20,21).
Given these recent observations, we wondered whether very closely related MMPs would exhibit distinctions in substrate recognition. Thus, we focus on substrate recognition by MMP-2 and its closely related homolog, MMP-9. These two MMPs are unique in that they contain three type II fibronectin domains intercalated within their catalytic domains. Although these fibronectin domains are oriented away from the catalytic cleft, they are believed to mediate docking interactions with substrates like gelatin and with the natural inhibitors, TIMPs. Until now it was assumed that MMP-2 and MMP-9 have over-lapping substrate recognition profiles. Here we use an unbiased substrate phage approach to obtain the substrate recognition profile of MMP-2. We find that like other MMPs, MMP-2 can cleave peptides containing the PXXX Hy sequence. However, three novel substrate motifs were also identified. These novel substrates are highly selective for MMP-2 over MMP-9. Consequently, the results of the present study challenge the idea that the two enzymes are functionally similar and provide a potential basis for their distinct biological roles.

EXPERIMENTAL PROCEDURES
Source of Commercial Proteins and Reagents-MMP-7 (active enzyme), MMP-13 (pro-enzyme), and TIMP-2 were purchased from Chemicon (Temecula, CA). Ilomastat was purchased from AMS Scientific (Concord, CA). Restriction enzymes were from Roche Biosciences or 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.
Construction of the Substrate Phage Library-The substrate phage library used in this study was generated using a modified version of the fUSE5 phagemid (22), as we have previously described (21). The library's primary features are a random peptide hexamer on the N terminus of the gene III protein and a FLAG epitope positioned to the N terminus of the hexamer. The substrate phage library represents 2.4 ϫ 10 8 individual sequences, ensuring with 75% confidence that all possible sequences are represented (22).
Expression and Purification of the Catalytic Domain of Human MMP-2 and 9 -Recombinant MMP-2 was generated in a manner similar to that we have previously reported for MMP-9 (21). Briefly, the cDNA encoding the catalytic domain of MMP-2 was generated by PCR, cloned into the pCDNA3 expression vector (Invitrogen), and used to transfect HEK 293 cells. Individual antibiotic-resistant clones were isolated with cloning rings, expanded, and then screened by reverse transcription-PCR and zymography. The catalytic domains of MMP-2 and 9 were purified from the conditioned medium by gelatin-Sepharose chromatography as described previously (23,24). We also employed an additional purification by ion exchange on Q-Sepharose to obtain greater purity. Fractions containing MMP-2 or 9 were concentrated in a dialysis bag against Aquacide II (Calbiochem). The purity of both enzymes was greater than 90% judging by silver-stained acrylamide gels. The purified zymogens of MMP-2 and 9 were stored at Ϫ70°C at concentrations ranging from 0.4 -1.3 mg/ml. Activation and Active Site Titration of Proteases-MMPs were activated by 2 mM p-aminophenylmercuric acetate(APMA) at room temperature as previously described (23). APMA was used to activate the two MMPs to avoid the inclusion of additional contaminating proteases in our phage selections. After activation, the activities of MMP-2 and MMP-9 were titrated using the hydroxamate inhibitor ilomastat as previously described (21,25). The active sites of full-length MMP-13 and MMP-7 were titrated with human TIMP-2. Briefly, 5-15 nM of each protease was pre-incubated with a range of TIMP-2 for 5 h at room temperature. Residual MMP-13 activity was monitored by cleavage of MCA-Pro-Cha-Gly-Nva-His-Ala-Dpa-NH 2 (Calbiochem). Residual MMP-7 activity was monitored by cleavage of MCA-Pro-Leu-Gly-Leu-Dpa-Ala-Arg-NH 2 (Calbiochem).
Selection of MMP-2 Substrates from the Phage Library-The substrate phage library (2 ϫ 10 10 phage) was incubated with 2.5 g/ml of MMP-2 in 50 mM Tris, pH 7.4, 100 mM NaCl, 10 mM CaCl 2 , 0.05% Brij-35, and 0.05% BSA for 1 h at 37°C. A control reaction was performed without protease. The cleaved phage were separated from the non-cleaved phage by immuno-depletion. 100 g of an anti-FLAG monoclonal antibody (Sigma) was added to the phage samples and then incubated for 18 h with rocking at 4°C. The phage-antibody complexes were twice precipitated by the addition of 100 l of pansorbin (Calbiochem). The cleaved phage remaining in the supernatant were amplified using K91 Escherichia coli and were then used for one additional round of substrate selection.
Monitoring Phage Hydrolysis by ELISA-Hydrolysis of individual phage substrates was measured using a modified ELISA that we have previously described (21). Briefly, phage from overnight cultures were captured into microtiter plates coated with anti-M13 antibody (Amersham Biosciences, Inc., 2.5 g/ml). The captured phage were incubated with MMP-2 (2.5 g/ml) in incubation buffer (50 mM Tris, pH 7.5, 100 mM NaCl, 10 mM CaCl 2 , 0.05% BSA, 0.05% Brij-35, 50 M ZnCl 2 ) for 2 h at 37°C. Control wells lacked protease. Following hydrolysis and ex-tensive washing, anti-FLAG polyclonal antibody (1.8 g/ml in TBS-T with 1 mg/ml BSA) was added to the wells and incubated for 1 h. Following additional washing, the level of bound anti-FLAG antibody was quantified with an horseradish peroxidase-conjugated goat antirabbit IgG antibody (Bio-Rad) followed by detection at 490 nm. The extent of hydrolysis of each phage was calculated by the ratio of the A 490 of the protease-treated samples versus samples lacking protease.
Mapping the Position of Scissile Binds within Peptide Substrates-Peptides representing the phage inserts were synthesized to our specifications by Annaspec Inc. The N termini were acetylated and C termini were synthesized as amides. The cleavage site within peptide substrates was determined using MALDI-TOF mass spectrometry. MMP-2 (25 nM) was incubated with 200 M of each peptide independently in 50 mM Tris, pH 7.5, 100 mM NaCl, 10 mM CaCl 2 for 2 h at 37°C. The mass of the cleavage products was determined using a Voyager DE-RP MALDI-TOF mass spectrometer (PerSeptive Biosystems, Framingham, MA). Following hydrolysis, the peptide samples were prepared for MALDI analysis according to methods described previously (26), and subsites within the peptide were designated according to the nomenclature of Schechter and Berger (27).
Quantifying the Kinetic Parameters of Peptide Hydrolysis-The kinetic parameters of substrate hydrolysis were measured using a fluorescamine incorporation assay that has been previously described (28 -31). Briefly, MMP-2, MMP-9, MMP-7, or MMP-13 were incubated with individual peptide substrates at concentrations ranging from 100 -800 M in 50 mM Tris, pH 7.5, 100 mM NaCl, 10 mM CaCl 2 , and 50 M ZnCl 2 . At selected time points the reactions were stopped by the addition of 1,10-phenanthroline. Peptide hydrolysis was determined by the addition of fluorescamine followed by detection at ex 355 nm and em 460 nm. The data were transformed to double reciprocal plots (1/[S] versus 1/V i ) to determine K m and k cat (28 -31). For some substrates, K m and k cat could not be determined individually, but the specificity constant, k cat / K m , was derived by the equation: k cat /K m ϭ v i /(E 0 )(S 0 ) and with the assumption that (S) is significantly lower than the K m (18).
Assessing Cleavage of Recombinant Eph Receptors by MMP-2 and MMP-9 -Recombinant fusion proteins between EphB1, EphB2, and the Fc domain of IgG were purchased from R&D Systems Inc. In these constructs, the extracellular domain of rat EphB1 (amino acid residues 1-538) and mouse EphB2 (amino acid residues 1-548) are fused to the Fc region of human IgG via a short polypeptide linker (32). The fusion proteins (1.8 M) were incubated for 4 h at 37°C with 280 nM of either MMP-2 or MMP-9. Following incubation, samples were resolved by 10% SDS-PAGE, and samples were visualized by Coomassie staining. The N terminus of the cleaved Eph B1 was determined by automated Edman degradation of protein blotted to polyvinylidene difluoride membranes.

RESULTS
Activated MMP-2 was used to select optimal substrates from a phage display library that we have previously described (21). Following two rounds of phage hydrolysis and subsequent amplification, several individual phage clones were selected randomly and assessed for hydrolysis by MMP-2. The phage vector encodes a FLAG epitope that is positioned to the N-terminal side of the randomized peptide hexamer. Therefore, cleavage within the hexamer by MMP-2 was gauged by measuring the liberation of the FLAG epitope using an ELISA. Thirty individual clones were selected for sequencing based on the fact that they were cleaved by more than 25% when incubated for 2 h with activated MMP-2 (60 nM). Four distinct groups of substrates are found among these clones (Table I).
The first group of substrates contains the PXXX Hy motif, where X Hy represents a large hydrophobic residue. This motif appears to be a substrate for a number of different MMPs (19 -21). Substrates in group II-IV represent novel recognition motifs for MMP-2. Group II substrates contain the I/LXXX Hy motif, in which the last hydrophobic residue is usually Ile or Leu. Group III substrates contain the X Hy SXL motif, where Ser and Leu are invariant. Group IV substrates are comprised of the HXXX Hy motif, which is similar to the cleavage site for MMP-2 within laminin-5 (7).
Assessing the Selectivity of the MMP-2 Substrates-We compared the rate of hydrolysis of a set of representative phage substrates by MMP-2 and MMP-9 (Fig. 1). With the exception of one clone, A45, the phage from group I lacked selectivity for MMP-2 over MMP-9. In contrast however, all of the phage substrates from groups II and III were highly selective for MMP-2 over MMP-9. The extent of hydrolysis of phage substrates from group IV was not compared using this assay because their rate of hydrolysis by MMP-2 was generally low. These substrates were characterized in more detail with the aid of synthetic peptides (see below).
Characterization of Synthetic Peptide Substrates-Representative peptides from groups I-IV were synthesized and used to characterize substrate hydrolysis in greater detail. The peptides were initially used to determine the position of the scissile bond within each motif by analyzing the cleavage products by MALDI-TOF mass spectrometry (Table II). In virtually all cases, the scissile bond directly precedes a large hydrophobic residue, which is frequently Ile or Leu (Table I). This feature is consistent with the presence of a deep binding pocket at the corresponding S 1 Ј subsite within MMP-2 (13).
The synthetic peptide substrates were also used to gain more detailed information on the selectivity of the substrate motifs for individual MMPs. We compared the rate of hydrolysis of each peptide by MMP-2, MMP-9, MMP-7, and MMP-13. Each enzyme was quantified by active site titration to ensure that accurate comparisons were obtained. The initial velocity of hydrolysis was measured across a range of peptide concentrations, and reciprocal plots were used to derive K m and k cat for MMP-2 and the k cat /K m ratio for each enzyme (Table III). A double reciprocal plot of the hydrolysis of peptide B74 is shown as a representative plot (Fig. 2). When cleaved by MMP-2, most of the substrates exhibited Michaelis constants in the mM range, with turnover rates ranging from 1 to 750 s Ϫ1 . Correspondingly, the k cat /K m ratios were all between 1.6 ϫ 10 4 M Ϫ1 s Ϫ1 and 2 ϫ 10 5 M Ϫ1 s Ϫ1 . Most significantly, substrates from groups II, III, and IV exhibit k cat /K m ratios that are 8-to 350-fold higher for MMP-2 than for the other MMPs. As a group, the substrates within group II show the most selectivity. The selectivity of these substrates for MMP-2 is conveniently illustrated by the ratio of k cat /K m for MMP-2 divided by the same value for the other MMPs (Table IV).
The Role of Residues at P 3 and P 2 in Conferring Substrate Selectivity-Several experiments were conducted to gain a better understanding of the structural basis for the selectivity of the substrates in group II-IV. Two hypotheses were tested. The first centered on the fact that neither group II nor group III substrates contain a Pro at the P 3 position. Since this residue is frequently found in substrates for other MMPs, and is also present at this position in group I substrates, we reasoned that its absence may be a defining feature of the selective substrates. To test this hypotheses, variants of three peptides from groups II and III were synthesized to contain a Pro at P 3 , and their hydrolysis by each MMP was measured (Table III). Although the inclusion of Pro at P 3 increased the k cat /K m ratio for MMP-9 between 3-and 11-fold, the proline-containing peptides remained better substrates for MMP-2. Consequently, the absence of Pro at the P 3 position is not the only determinant in the selectivity of these substrates.
The second hypothesis was based on our recent characterization of the substrate recognition profile of MMP-9 (21). That study revealed a critical role or preference for Arg at the P 2 position. Since Arg is rarely present at P 2 in the substrates selected for MMP-2, we hypothesized that substitution of Arg into P 2 might shift selectivity away from MMP-2 and toward MMP-9. Indeed, in the three peptides tested, B74, A13, and C9, the substitution of Arg into the P 2 position dramatically increased hydrolysis by MMP-9. This substitution also significantly decreased hydrolysis by MMP-2. In combination, these Phage substrates are selective for MMP-2 over MMP-9. The ability of MMP-2 (dark bars) and MMP-9 (open bars) to cleave substrate selected from the phage library were compared using the phage ELISA procedure described under "Experimental Procedures." Immobilized anti M13 antibody was used to capture individual phages onto 96-well microtiter wells. The captured phage were cleaved with 2.5 g/ml MMP-2 or MMP-9. The extent of cleavage within the phage insert was assessed by measuring the release of the FLAG epitope. Results are presented as the percentage of hydrolysis compared with non-treated control phage. This experiment was repeated three times, yielding nearly identical results in each repetition. effects completely switch the selectivity ratio of the mutated peptides (Table IV). These findings underscore the significance of Arg at P 2 in facilitating substrate recognition by MMP-9 and also point to the important role of the S 2 subsite in distinguishing the activity of MMP-2 and MMP-9. Interestingly, the substitution of Arg at P 2 had minimal effects on the k cat /K m ratio of MMP-7 or MMP-13 (Table III), indicating that other features that are still not understood, confer selectivity of these peptides for MMP-2 over MMP-7 and MMP-13.
We noticed that the sequence of substrate A21 is similar to an MMP-2 selective cleavage site within laminin-5 (7). Peptide A21 has a k cat /K m ratio that is 8-fold higher for MMP-2 over MMP-9. Interestingly though, it contains a rather large residue, Lys, at P 2 . Because smaller residues are favored at P 2 by MMP-2, we mutated this Lys to Ala as an additional test of the P 2 residue in conferring selectivity. In addition, this mutation makes the sequence of the peptide match more closely with the cleavage site within laminin-5, which contains the HAAL sequence (7). The mutated peptide is hydrolyzed by MMP-2 more than 10-fold better than the A21 parent peptide. This mutation was without significant effect on hydrolysis by MMP-9. These observations support the idea that the P 2 /S 2 interaction is key to distinguishing substrate recognition by MMP-2 and MMP-9.
Selective Hydrolysis of a Protein Substrate Containing the SXL Motif-Ultimately, one would hope to be able to use the substrate recognition profiles obtained from substrate phage and other substrate profiling strategies (33) to generate hypotheses regarding physiologic substrates. However, there have not been enough test cases to establish the rules for such extrapolations. As an initial step in this direction, we compared the ability of MMP-2 and MMP-9 to cleave Eph B1 and Eph B2, tyrosine kinase receptors that are responsible for cell-cell signaling in neuronal development (34). These proteins contain putative cleavage sites that correspond rather closely to the substrates selected from the phage library by MMP-2. There are two potential cleavage sites within Eph B1. The first motif contains the sequence is SISSLW, which matches well with the X Hy SX2L motif of the group III substrates. This motif is positioned within a predicted ␤-strand within a fibronectin repeat in Eph B1 (35). A second potential cleavage site with a sequence of KSEL, is located in a 10-residue linker between the membrane-spanning segment and the second type III repeat of Eph B1. This sequence does not precisely match the motif of the group III substrates, but it does contain the core SXL motif. In

TABLE II
Identification of the position of scissile bonds within MMP-2 substrates Representative peptides were synthesized from each group of substrates. Using MALDI-TOF mass spectrometry the position of scissile bonds was determined by analyzing the mass of cleaved peptide fragments generated by MMP-2. The position of the scissile bond is noted by placing the P 1 and P 1 Ј residues in bold. a Denotes the mass of Na ϩ adducts of the peptide fragments.

TABLE III Measuring peptide hydrolysis by a panel of MMPs
The hydrolysis of synthetic peptides by different MMPs was quantified using procedures outlined in "Experimental Procedures." For MMP-2, hydrolysis was measured over a concentration range of peptide. Values for k cat and K m were derived from Lineweaver-Burk plots. For other proteases only the k cat /K m ratio was measured (see "Experimental Procedures"). Additional synthetic peptides were synthesized with substitution of Pro at P 3 and Arg at P 2 to test the influence of these two subsites on substrate specificity. Each measurement was repeated three times and the relative error of k cat /K m in these measurements was less than 10%. ND, not determined. The Eph B1 and Eph B2 fusion proteins were incubated with equimolar amounts (280 nM) of MMP-2 and MMP-9 for 4 h. The extent of hydrolysis was gauged by SDS-PAGE (Fig. 3). The Eph B1-Fc fusion protein was almost quantitatively cleaved by MMP-2 (Fig. 3, lane 2). The extent of cleavage by MMP-9 was far lower (Fig. 3, lane 3), which is not surprising considering the selectivity exhibited by the group II substrates. Neither protease cleaved the Eph B2 fusion protein. The site of hydrolysis within Eph B1 was determined by sequencing the N ter-minus of one of the released fragments. The amino acid sequence of this fragment indicates that the protein was cleaved at the sequence DDYKSE2LRE, which is found within the 10-residue linker of Eph B1. This is one of the predicted cleavage sites in EphB1. These findings illustrate that motifs found to be selective for MMP-2 by substrate phage display can act as selective substrates within the context of whole proteins. Equally as important, however, this experiment illustrates that the three-dimensional conformation of the putative cleavage site will, to a large degree, control the extent of hydrolysis. Therefore, we conclude that meaningful genome-wide predictions on putative physiologic substrates will require the incorporation of additional information and computational filters (see "Discussion").

DISCUSSION
Despite their high structural homology, MMP-2 and MMP-9 are reported to have distinct biological roles. For example, in tumors in the RIP-Tag mouse where both proteases are present, only MMP-9 seems to be causally involved in the angiogenic switch of these tumors (5). Similarly, in the process of platelet aggregation, MMP-2 promotes aggregation, but MMP-9 inhibits aggregation (6). The reasons for these differences in function are not clear. One possible explanation could be that the two enzymes have distinct substrate recognition profiles. Until now, however, there has been no systematic and unbiased comparison of these profiles. Here, we defined the substrate recognition profile of MMP-2 and compared it to that of the closely related MMP-9. The findings of the study show that although two MMPs can cleave a common set of substrates with the PXXX Hy sequence, MMP-2 hydrolyzes an additional array of peptide substrates. These observations support the idea that differences in biological function of MMP-2 and MMP-9 could stem from their action on distinct physiologic substrates.
We found that like other MMPs, MMP-2 efficiently cleaves peptide substrates that contain the PXXX Hy motif. This observation is consistent with the idea that this motif is recognized universally by the MMP family. The sequence of this motif is generally complementary with the structural features within the catalytic clefts of the MMPs, including a deep S 1 Ј binding pocket, and a general deviation from linearity extending from The selectivity of each of the peptide substrates for MMP-2 compared to other MMPs is represented by dividing the k cat /K m ratio for MMP-2 by the same value for the other MMPs that were tested. In several cases mutation of the P 2 residue gave rise to dramatic shifts in the selectivity ratio between MMP-2 and MMP-9. These values are in bold text. ND, not determined.   the S 3 position, which is occupied by the invariant Pro in this group of substrates. Even within the context of this commonly recognized motif, some substrate selectivity is evident. Substrates for MMP-13 that contain this motif exhibit good selectivity for MMP-13 over three other MMPs (20). Similarly, the MMP-9 substrates with this motif that we recently characterized are selective for MMP-9 over MMP-7 and MMP-13. Interestingly, those substrates were found to be cleaved equally well by MMP-9 and MMP-2 (data not shown). Here we observed one substrate with the PXXX Hy sequence, phage A45, that is selective for MMP-2 over MMP-9, but such selectivity among the two gelatinases does not appear to be a general property of substrates containing PXXX Hy .
Three of the novel groups of substrates for MMP-2 were characterized in detail and were found to be selective for MMP-2 over the other MMPs tested. These substrates contain consensus motifs of L/IXXX Hy , X Hy SXL, and HXXX Hy . All three sets of substrates have Michaelis constants in the low mM range and turnover rates of several hundred per second. They also contain large hydrophobic residues at the P 1 Ј position, a feature that is consistent with the depth of the S 1 Ј pocket. The dominant role of this interaction to substrate recognition by the MMPs is underscored by presence of a hydrophobic residues at this position in virtually all substrates selected for MMPs out of randomized libraries (19 -21). The P 3 residue of group II and III substrates also has a hydrophobic character. In group II this position is an invariant Ile or Leu. Interestingly, an invariant Ser at P 2 is the feature that distinguishes substrates in group II and III. The group IV substrates are unique in that the P 3 residue is occupied by histidine. We did not identify a sufficient number of substrates within this group to arrive at a consensus at other positions.
The findings presented here also begin to provide a structural basis for substrate selectivity between MMP-2 and MMP-9. In our selection of substrates for MMP-9 (21), we found Arg to be preferred at P 2 , whereas Arg was rarely observed at this position in the substrates selected for MMP-2. Rather, the MMP-2 substrates frequently display a relatively small residue at P 2 . For example, this position occupied by an invariant Ser in the group III substrates. When Arg is substituted into the P 2 position in the MMP-2 substrates, a marked shift in substrate recognition was observed. In all cases the rate of hydrolysis by MMP-2 was decreased, and hydrolysis by MMP-9 was increased. In fact, this substitution increased the k cat /K m ratios for MMP-9 between 25-and 60-fold. Further support for the role of the P 2 residue in substrate recognition comes from analysis of the group IV substrate A21, which contains Lys at P 2 . We converted this Lys to Ala in order to more closely mimic the natural sequence in laminin-5, which is known to be selectively cleaved by MMP-2. Substitution of this Lys to Ala dramatically increased the k cat /K m ratio for MMP-2, providing further support for the idea that the S 2 subsite within MMP-2 is sterically hindered. It is interesting to note that the substitution of Arg into P 2 of the substrate had little effect on the already poor recognition of these motifs by MMP-7 and MMP-13. Hence, the primary effect of substitutions at P 2 is a switch in substrate recognition by the closely related gelatinases.
Analysis of the structure of the S 2 subsite within MMP-2 (13) and MMP-9 reveals a potential structural basis for this distinction in substrate recognition at the P 2 position (Fig. 4). The S 2 pocket is remarkably similar in both enzymes, save the presence of Glu-412 in MMP-2, which is replaced by an Asp in MMP-9. Our prior docking of substrates containing Arg at P 2 into the catalytic cleft of MMP-9 indicated a favorable interaction between the positively charged guanidino group of Arg with the acidic side chain of Asp-410 (21). Since Glu contains an additional methylene group in its side chain, Glu-412 would be expected to extend further into the S 2 pocket of MMP-2 (orange arrow). Thus Glu-412 is expected to occlude the S 2 subsite and hinder the docking of substrates that contain larger residues, like Arg, at the P 2 position. Another distinction between the two proteases is observed at Ala-196 of MMP-2, which is replaced by Pro-193 in MMP-9. The pyrrolidine ring of Pro-193 of MMP-9 extends further out into the S 3 space. This Pro could potentially interfere with docking of extended residues like Leu and Ile that are often found at P 3 in the MMP-2 substrates.
Aside from the differences in structure within the catalytic clefts of MMP-2 and MMP-9, we must also consider the possibility that these proteases can assume distinct conformations that are not revealed by the existing crystal structure (9 -12, 36) and models (37). This possibility is supported by the fact that unbiased searches reveal four separate sets of substrates for MMP-2 (this report) and three families of substrates for MMP-9 (21). Thus, it is conceivable that the two pockets, although containing virtually the same resides, could adopt different conformations because of constraints imposed at ancillary regions of the protease. Although there is little direct evidence for this possibility, it does provide an alternative hypothesis to explain their differences in substrate recognition. Such conformational modulation of the catalytic pocket could also relate to the biology behind the two MMPs because their binding to other proteins, like the ␣ v ␤ 3 integrin (4), could Models of the catalytic domain of MMP-2 (top panel) and MMP-9 (bottom panel) were constructed using the Modeler4 software from Rockefeller University (40). Models were constructed using the coordinates of the catalytic domain of MMP-2 from the reported crystal structure (13)(PDB accession number 1qibA). The model of MMP-9 is based on the crystal structure of MMP-2. A view looking into the catalytic cleft is illustrated with the zinc ion colored orange. The side chains of the three His residues that ligand with zinc are evident. The S 2 and S 3 subsites fall within the cleft to the left of the zinc ion. Glu-412 protrudes into the S 2 subsite of MMP-2 (orange arrow). In MMP-9 the corresponding residue is Asp-410 (orange arrow). One other notable difference among the two proteases is Pro-193 in MMP-9, whose pyrrolidine ring protrudes away from the upper rim of the catalytic cleft and into the interface between the S 2 and S 3 subsites (white arrow). In MMP-2 this Pro is substituted by Ala, creating a relatively unobstructed surface across the S 3 space.
conceivably regulate substrate recognition. One might also hypothesize that the binding of either protease to collagen through its type II fibronectin repeats could provide similar conformational regulation.
Ideally, one would like to use the information obtained from substrate profiling approaches, like the one reported here and elsewhere (33,38), to predict the physiological and pathophysiological substrates for proteases. Our findings indicate that such predictive strategies have merit. For example, it is encouraging that the differences in the substrate phage profiles for MMP-2 and MMP-9, are reflected by the selective cleavage of Eph B1 by MMP-2. It is equally encouraging that the closely related homologue Eph B2, which lacks the predicted MMP-2 cleavage site, is not cleaved by MMP-2. The group IV substrates that contain the HXXX Hy motif match closely with the selective MMP-2 cleavage site in laminin-5 (sequence of HAA2LTS) (7), an observation that provides additional support for the idea that substrate profiling could ultimately have predictive utility.
However, it is evident that if they are to be applied on a genome-wide scale these predictive methods require further refinement. Even though the sequences within Eph B1 and laminin-5 are similar to the substrate motifs from phage, they are not identical. Consequently, using the phage profiles to arrive at a consensus recognition motif that is based on the physical properties of preferred residues (e.g. large hydrophobic versus small hydrophilic) rather than actual residues, is worthy of exploration. Building from here though, additional filters or constraints will need to be applied to narrow the number of putative substrates. For example, one could limit searches for putative substrates to proteins expressed in the appropriate cellular compartment or extracellular space. Further constraints could be imposed based on information obtained from gene expression profiling, which will reveal all genes that are co-expressed with any given protease. Our findings also indicate that the degree to which a predicted cleavage site is exposed to solvent should be taken into consideration. Since automated medium resolution structural predictions can now be made across entire genomes (39), it is not unreasonable to suggest that this information could be used as an additional filter to identify proteins with accessible cleavage sites.