A Unique Substrate Binding Mode Discriminates Membrane Type-1 Matrix Metalloproteinase from Other Matrix Metalloproteinases*

In our study, we characterized the substrate recognition properties of membrane type-1 matrix metalloproteinase (MT1-MMP; also known as MMP-14), a key enzyme in tumor cell invasion and metastasis. A panel of optimal peptide substrates for MT1-MMP was identified using substrate phage display. The substrates can be segregated into four groups based on their degree of selectivity for MT1-MMP. Substrates with poor selectivity for MT1-MMP are comprised predominately of the Pro-X-X-↓-X Hy motif that is recognized by a number of MMPs. Highly selective substrates lack the characteristic Pro at the P3 position; instead they contain an Arg at the P4 position. This P4Arg is essential for efficient hydrolysis and for selectivity for MT1-MMP. Molecular modeling indicates that the selective substrates adopt a linear conformation that extends along the entire catalytic pocket of MT1-MMP, whereas non-selective substrates are kinked at the conserved P3 Pro residue. Importantly, the selective substrates can be made non-selective by insertion of a proline kink at P3, without significantly reducing overallk cat/K m values. Altogether the study provides a structural basis for selective and non-selective substrate recognition by MT1-MMP. The findings in this report are likely to explain several aspects of MT1-MMP biology.

soluble type and the membrane type. The membrane-type MMPs are tethered to the plasma membrane by either a transmembrane domain or a glycosylphosphatidylinositol linkage (6,7). Membrane type-1 matrix metalloproteinase (MT1-MMP; also known as  is the prototypic member of the MT-MMPs, and its expression has been associated with a variety of cellular and developmental processes, as well as multiple pathophysiologic conditions. Similar to most other MMPs, MT1-MMP cleaves a number of matrix proteins including collagen, fibronectin, and vitronectin (8 -10). Other work shows that MT1-MMP has substrates that extend beyond extracellular matrix proteins. MT1-MMP is known to convert pro-MMP-2 to active protease (6,(11)(12)(13), apparently through its trimolecular complex with tissue inhibitor of metalloproteinase-2 and the ␣ v ␤ 3 integrin (14,15). Recent work shows that cleavage of cell surface molecules such as CD44, pro-␣ v integrin, and transglutaminase by MT1-MMP modulates cell migration (16 -18).
The importance of the broad substrate recognition specificity of MT1-MMP is consistent with the complex phenotype of the MT1-MMP-deficient mice. These mice exhibit overt dwarfism, but close inspection also showed several additional defects including arthritis, delayed ossification of bone, and the inability to respond to angiogenic stimuli (19,20). Clearly, MT1-MMP is likely to have multiple physiologic and pathophysiologic substrates. Knowing the identity of these substrates, and the structural basis for their recognition, would provide an additional level of understanding of MT1-MMP. To this end, we used an unbiased substrate phage selection to identify optimal peptide substrates for MT1-MMP. Four groups of substrates with varying degrees of selectivity were identified. As anticipated, non-selective substrates comprised primarily of the Pro-X-X-2-X Hy sequence were identified. These substrates are collagen-like and have emerged as a canonical, and generally non-selective, MMP recognition motif (21)(22)(23). We also identified substrates that are recognized by MT1-MMP and both gelatinases (MMP-2 and MMP-9). Most significantly, a group of highly selective substrates for MT1-MMP was identified. Each of the highly specific substrates is devoid of a Pro residue at the P 3 position and contains a critical Arg at the P 4 position. Modeling of the selective and non-selective substrates bound to the catalytic pocket of MT1-MMP indicates two separate binding modes. These modes are distinguished by the degree of distortion at the P 3 position of substrate. These observations provide a solid structural basis for the recognition of distinct classes of substrates by MT1-MMP and for the design of highly specific inhibitors of this enzyme.

EXPERIMENTAL PROCEDURES
Construction of Substrate Phage Display Libraries-Substrate phage libraries were generated using a modified version of the fUSE5 phagemid (24 -26). A FLAG epitope was engineered at the N terminus of the geneIII protein by annealing oligonucleotides 5Ј-CCGGGTTT-GTCGTCGTCGTCTTTGTAGTCGGTAC-3Ј and 5Ј-CGACTACAAA-GACGACGACGACAAAC-3Ј and ligating them into fUSE5 at the KpnI and XbaI restriction sites. The random hexamers were generated by PCR extension of the template oligonucleotide 5Ј-GGGGAGGCCG-ACGTGGCCGTCATCAGGCGGCTCAGGC(NNK) 6 ACGGCCTCTGGG-GCCGAAAC-3Ј, where N is any nucleotide and K is either G or T. The template oligonucleotide also encodes an SGGSG linker positioned in between the FLAG epitope and the random hexamer. A primer oligonucleotide 5Ј-AATTTCTAGTTTCGGCCCCAGAGGC-3Ј and the template oligonucleotide were mixed and heated at 65°C for 2 min. The heating block was switched off and allowed to cool passively to 40°C to allow annealing of the extension oligonucleotide to the template oligonucleotide. Elongation of the template oligonucleotide was performed using Sequenase (United States Biochemical, Cleveland, OH) (27). The final cDNA product was precipitated with ethanol, resuspended in water, and digested with SfiI. The DNA insert and fUSE5 were mixed and ligated at a 5:1 molar ratio and electroporated into Escherichia coli MC1061(FϪ). Several phages were selected for sequencing to confirm randomness in the insert sequences and the correct reading frame.
Cloning and Expression of Recombinant MT1-MMP, MMP-2, and MMP-9 Catalytic Domains-Recombinant versions of the catalytic domains of MMP-2 and MMP-9 were expressed, purified, and activated as described previously (21,28). The catalytic domain of MT1-MMP was amplified from a human Universal QUICK-Clone cDNA library (CLON-TECH, Palo Alto, CA) by PCR using the primers 5Ј-catatgTACGC-CATCCAGGGTCTCAAATGG-3Ј and 5Ј-gaattcttaCCCTGACTCAC-CCCCATAAAGTTGC-3Ј. The NdeI and EcoRI restriction sites and TAA stop codon are shown in lowercase. The amplified PCR product was digested with NdeI and EcoRI and cloned into the corresponding sites of pFLAG-ATS (Sigma). The resulting plasmid (pGM-MT1) was transformed into BL21(DE3) E. coli and grown in 2YT supplemented with ampicillin (50 g/ml). Protein expression was induced by the addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 0.1 mM. After 3 h of induction, the bacteria were collected and lysed, and inclusion bodies containing the catalytic domain of MT1-MMP were isolated by centrifugation.
Recombinant MT1-MMP was purified from inclusion bodies by anion-exchange chromatography on a MonoQ column (Amersham Biosciences) in the presence of 8 M urea. The purified material was refolded by dilution to 100 g/ml in 50 mM HEPES, pH 7.0, 5 mM CaCl 2 , and 50 M ZnCl 2 , followed by an incubation at 12°C for 1 h. The refolded MT1-MMP was used immediately in activity assays. The concentration of active enzyme was determined by active site titration. Active site titration of MT1-MMP was performed with the hydroxomate inhibitor AG3340 (K i ϭ 3 ϫ 10 Ϫ10 M) and tissue inhibitor of metalloproteinase-2. The titrations of active MT1-MMP were equivalent; thus AG3340 was used to titrate the enzyme for this study. After pre-incubation, the steady-state rate of hydrolysis of the fluorogenic substrate Mca-PLGL-Dnp-AR-NH 2 (Bachem, King of Prussia, PA) was determined at ambient temperature using an fMax fluorescence microplate reader (Molecular Devices, Sunnyvale, CA). The steady-state rate was plotted as a function of inhibitor concentration and fitted with the equation where V is the steady-state rate of substrate hydrolysis, SA is specific activity (rate per unit enzyme concentration), E 0 is enzyme concentration. I is inhibitor concentration, and K I is the dissociation constant of the enzyme-inhibitor complex (29).
Phage Selection of MT1-MMP Substrates-An aliquot (2 ϫ 10 10 phage) of the substrate phage library was incubated with 5 g/ml of MT1-MMP 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 selection was performed without protease. The cleaved phages were separated from the non-cleaved phages by immunodepletion. The M1 anti-FLAG monoclonal antibody (100 g) was added to the phage samples and then incubated for 18 h with rocking at 4°C. The phage-antibody complexes were precipitated by the addition of 100 l of Pansorbin (Calbiochem, San Diego, CA). The cleaved phages remaining in the supernatant were amplified using K91 E. coli as described previously (30,31) and were then used for subsequent rounds of substrate selection.
Substrate Phage ELISA-Hydrolysis of individual phage substrates was measured using a modified ELISA. Wells of a 96-well microtiter plate were coated with anti-M13 antibody (Amersham Biosciences) at 2.5 g/ml in phosphate-buffered saline overnight at 4°C. The wells were blocked for 1 h at room temperature in TBST (50 mM Tris, pH 7.8, 150 mM NaCl, 0.2% Tween 20) containing 10 mg/ml BSA. After blocking, 150 l of supernatant from an overnight phage culture was added to each well and incubated for 2 h at 4°C to allow for phage capture. Unbound phage were removed with three washes of ice-cold TBST. To assess hydrolysis, MT1-MMP (3.0 g/ml) was added to the appropriate wells in incubation buffer (50 mM Tris, pH 7.4, 100 mM NaCl, 10 mM CaCl 2 , 0.05% BSA, 0.05% Brij-35) for 2 h at 37°C. Control wells lacked protease. The protease solution was removed, and the wells were washed four times with ice-cold TBST. To determine the extent of hydrolysis of the peptides on phage by proteinase, anti-FLAG polyclonal antibody (1.8 g/ml in TBST with 1 mg/ml BSA) was added to each well, and the plates were incubated at 4°C for 1 h. Binding of anti-FLAG antibody to FLAG epitope was measured with a horseradish peroxidase-conjugated goat anti-rabbit IgG antibody (Bio-Rad) followed by detection at 490 nm. The extent of hydrolysis, taken as a measure of substrate hydrolysis, was calculated by the ratio of the optical density at 490 nm of the protease-treated samples versus controls.
Determination of Scissile Bonds-The cleavage site for MT1-MMP within peptide substrates was determined using MALDI-TOF mass spectrometry. MT1-MMP (23 nM) was incubated with 100 M of each peptide sample in 50 mM Tris, pH 7.4, 100 mM NaCl, 10 mM CaCl 2 for 2 h at 37°C. Following hydrolysis, the peptide samples were prepared according to methods described previously (32,33). The mass of the cleavage products was determined using a Voyager DE-RP MALDI-TOF mass spectrometer (PerSeptive Biosystems, Framingham, MA). In all cases, the observed fragments corresponded to a single cleavage site.
Kinetic Measurements of Peptide Hydrolysis-The kinetic parameters of substrate hydrolysis were measured using a fluorescamine incorporation assay that has been described previously (34). Briefly, MT1-MMP, MMP-2, and MMP-9 were incubated with individual peptide substrates at concentrations ranging from 100 to 800 M in 50 mM Tricine, pH 7.5, 100 mM NaCl, 10 mM CaCl 2 , 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 (34). Similar results were obtained using different batches of protease. 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 Modeling of Substrate Peptides with MT1-MMP-The complexes of MT1-MMP and cognate substrates were modeled using the Sybyl software package available from Tripos (www.tripos.com) using the MT1-MMP spatial structure taken from PDB file 1bqq (36 -38).

Selection of Peptide Substrates for MT1-MMP-
The catalytic domain of MT1-MMP was used in three successive rounds of selection to isolate optimum substrates from a phage library of 2.4 ϫ 10 8 independent clones. Two-hundred clones were screened for their ability to be hydrolyzed by MT1-MMP using the substrate phage ELISA we reported previously (21,28). Forty of the two-hundred clones were cleaved efficiently by MT1-MMP and were carried forward for additional analysis.
In a second round of analysis, substrate phage ELISAs were used to compare relative rates of hydrolysis of phage peptide substrate by MT1-MMP, MMP-2, and MMP-9 catalytic domains. Based on the activity profiles resulting from this analysis, the 40 phage substrates could be segregated into four distinct groups as follows: (i) those recognized by all three MMPS, (ii) those recognized equally by MT1-MMP and MMP-2, (iii) those recognized equally by MT1-MMP and MMP-9, and (iv) those selective for MT1-MMP (Table I).
The amino acid sequence of each phage peptide was determined by sequencing the corresponding DNA inserts of the selected phage, and the peptides are presented in Table I. These sequences reveal the structural features common to each group of substrates. The non-selective substrates of group I all contain Pro residues. Interestingly, all of the Pro residues were expressed at the same site in the random peptides. Upon first inspection, these substrates appear similar to the Pro-X-X-2-Hy motif that is commonly cleaved by many MMPs (21)(22)(23). This motif represents a canonical collagen-like MMP recognition motif (9). The substrates in group II, which are susceptible to cleavage by MT1-MMP and MMP-2, were comprised primarily of substrates that appeared to be similar in primary sequence to an MMP-2 recognition motif we identified recently (39). The substrates in group III, which are susceptible to cleavage by MT1-MMP and MMP-9, were relatively small in number however. Consequently, no consensus could be defined. The substrates that were highly selective for MT1-MMP are in group IV and appeared to give little initial indication of consensus. It is noteworthy that Pro is absent from these peptides, however. Given the contrast in substrate selectivity between groups I and IV, we focused on trying to determine the structural basis for their distinct reactivity with MT1-MMP.
Identification of Scissile Bonds within MT1-MMP Peptide Substrates-Although the position of the scissile bond in the phage substrates could generally be surmised by the position of hydrophobic residues (Leu/Ile) that normally occupy the deep S 1Ј pocket, we confirmed the placements by analyzing the hydrolysis of synthetic peptides. Peptides were synthesized to mimic the peptides displayed by the individual phage clones. Following exposure to the MT1-MMP catalytic domain, the masses of the cleavage products were determined with MALDI-TOF mass spectrometry (data not shown). The MALDI analysis confirmed that hydrophobic amino acids occupied the P 1Ј position of the MT1-MMP substrates (Table II). In addition, a polar amino acid appears to be preferred at the P 2Ј position. Furthermore, an alignment of the group IV peptides based on their scissile bonds demonstrates a preference for Arg at the P 4 position. Two substrates, A20 and A167, from group IV that lacked Arg at P 4 were also examined. The masses of cleavage products of these two peptides indicate that the P 4 position is Phe or Lys, respectively, which both contain long side chains like Arg that can participate in hydrophobic interactions (see below).
Kinetic Analysis of Substrate Hydrolysis by MT1-MMP-Results from the substrate phage ELISA (Table I) indicate a rank order of substrate preference within each group of MT1-MMP substrates. To quantify precisely the rate of hydrolysis of the peptide substrates in group IV by MT1-MMP, the specificity constant (k cat /K m ) of selected peptides was determined (Table II). In these studies only MMP-9 was used for comparisons with MT1-MMP, because MMP-9 and MMP-2 share similar substrate specificity and cleavage efficiency. Each proteinase was incubated with increasing concentrations of peptide. Peptide hydrolysis was measured by incorporation of fluorescamine onto newly formed N termini. From these measurements, k cat and K m values were derived for each peptide using double reciprocal plots of 1/[S] versus 1/v i . Among the peptides, overall k cat /K m values ranged from 8,200 M Ϫ1 s Ϫ1 to 777,200 M Ϫ1 s Ϫ1 , and the selectivity of these peptides ranged from 2.2-to 83-fold when compared with hydrolysis by MMP-9. Because nearly one-half of the MT1-MMP-selective substrates lack Arg at P 4 , there may be other modes of selective recognition. Nevertheless, we suspect that some of these other peptides are related to the Arg-containing peptides. For example, peptides A167 and A20, which lack Arg at P 4 , showed the lowest k cat /K m values and the lowest selectivity ratios. Yet, Phe and Lys, which occupy P 4 in these peptides, can participate in similar hydrophobic interactions.
Assessing the Role of Arg at P 4 in Substrate Recognition by MT1-MMP-Because of the preponderance of Arg at the P 4 position among the group IV substrates, we sought to determine the requirement for this residue in hydrolysis by MT1-MMP. Three peptides that contain this P 4 Arg, A42, B175, and A176, were selected for mutational analysis. The P 4 position in each synthetic peptide was changed from Arg to Ala, and k cat /K m values were measured (Table II). This substitution decreased k cat /K m of peptide A176 from 49,100 M Ϫ1 s Ϫ1 to 5,900 M Ϫ1 s Ϫ1 . Similarly the same substitution in B175 reduced k cat /K m from 63,600 M Ϫ1 s Ϫ1 to 10,400 M Ϫ1 s Ϫ1 . The Arg to Ala substitution had a less significant effect in the A42 peptide (3-fold reduction). We suspect that the very high k cat /K m of this peptide (777,200 M Ϫ1 s Ϫ1 ) indicates the presence of several additional contacts that favor substrate recognition by MT1-MMP. Consequently, substitution of the Arg at P 4 has less of an effect on A42. Altogether, the mutational analyses support our  hypothesis that Arg at P 4 is a key contact residue in the group IV substrates.
Making Selective Substrates Non-selective-An obvious distinction between the selective and non-selective substrates is the presence of Pro at the P 3 position. We sought to determine whether the presence of a Pro at this position of the group IV substrates would be sufficient to override their selective docking. For this purpose, we chose the peptide with the highest selectivity for MT1-MMP (A176). When the P 3 Ser of A176 was replaced with Pro, the k cat /K m by MT1-MMP was reduced by only 32%. Conversely, the hydrolysis of this peptide by MMP-9, a measure of selectivity, was increased 58-fold. Hence, substitution of Pro at the P 3 position converted a substrate selective for MT1-MMP to a substrate that is recognized equally well by both MMPs. A similar substitution was made into the A176A peptide, in which the Arg at P 4 of A176 was changed to Ala. This substitution caused a 9-fold reduction in k cat /K m ratio for MT1-MMP compared with the parent peptide. Yet, when Pro was substituted into the P 3 position of this peptide the k cat /K m ratio for MT1-MMP increased by ϳ6-fold. Consequently, the deleterious effect of the Arg to Ala substitution at P 4 was rescued by substitution of Pro at P 3 . Equally as significant, this substitution reversed the selectivity ratio to favor MMP-9. These findings support the idea that MT1-MMP recognizes substrates in two distinct modes. One mode, demonstrated by the group I peptides, makes use of the P 3 and P 1Ј positions as dominant contact points. The other mode, demonstrated by the selective group IV peptides, appears to use primarily the P 4 and P 1Ј subsites as key contacts.
Modeling Substrates in the Catalytic Pocket of MT1-MMP-To visualize the differences in these two binding modes, molecular modeling studies were conducted. A three-dimensional model of the catalytic domain of MT1-MMP was generated from the coordinates of its known crystal structure (36). The catalytic domain was then docked with peptides A176 (Ac-SGRSEN2IRTA-NH 2 ) and A176P (Ac-SGRPEN2IRTA-NH 2 ) using Sybyl. Because interactions between MMP and substrate at the S 1Ј subsite are known to make a major contribution to substrate binding, interactions around this site were kept constant to constrain the modeling procedure. Two restrictions were applied. First, the hydrogen bond between the NH group of the P 1Ј Ile residue and the carbonyl oxygen of Ala 200 in MT1-MMP was set as a constant. Second, the interaction between the side chain of the P 1Ј Ile residue of substrate and the hydrophobic patch made up of Val 236 and the surrounding atoms in MT1-MMP was held constant. The initial docking conformation of the substrate was essentially a ␤-strand running anti-parallel to the neighboring ␤-strand in MT1-MMP (Ala 200 to Ala 202 ). Conformation searches of the N-terminal and C-terminal halves of the substrate were run separately. To reduce the computation, the conformations were searched in torsion angle space, in which bond length and bond angles of the whole substrate remained fixed. These computational searches led to the hypothetical docking modes illustrated in Fig. 1.
Interestingly, peptide A176 fits in the catalytic cleft of MT1-MMP with very little deviation from linearity (panel A). The Arg at the P 4 position in this peptide is distorted slightly, apparently facilitating hydrophobic contact with Gly 116 and Phe 204 of MT1-MMP. It would appear that these hydrophobic contacts are the dominant interactions within the S 4 pocket of MT1-MMP. In contrast to the extended linear conformation of peptide A176, a variant of this peptide, with Pro at P 3 , bound with a pronounced kinked conformation. The effect of this deviation from linearity is that the P 4 Arg is not involved in hydrophobic contacts described for peptide A176. Instead, the presence of Pro at the P 3 position changes contacts such that the P 3 Pro residue interacts with Phe 204 of MT1-MMP, in effect suppressing the ability of the P 4 Arg to make its contacts. These models illustrate the likely structural distinctions between the selective and non-selective substrate binding modes of MT1-MMP, thereby representing the apparent distinction between collagen and non-collagen substrates by MMPs. DISCUSSION The absolute level of MMP activity is regulated at three key steps. First, virtually all MMPs are regulated at the level of transcription (1, 2). Second, inactive MMP zymogens are acti- vated by a proteolytic removal of a segment of the pro-domain. Third, their natural inhibitors, the tissue inhibitors of metalloproteinases, can inhibit the activated proteinases. Another key aspect of MMP biology is the control at the level of substrate recognition. Substrate recognition can be modulated strongly by domains ancillary to the catalytic cleft, like the hemopexin domain, or the type II fibronectin repeats in gelatinases (40 -42). These domains guide proteolysis by binding to specific substrates. The control of substrate hydrolysis at the catalytic cleft is generally thought to play a lesser role. This perception is driven partly by the overall structural similarity of the catalytic clefts of MMPs, which all contain a defining zinc ion, zinc-coordinating His residues, a catalytic glutamic acid, and a rather deep, S 1Ј pocket (43,44). Indeed, it is the S 1Ј pocket that has been exploited to design small molecule antagonists of MMPs. The fact that many of these antagonists have a broad inhibition spectrum also supports the contention that substrate recognition is similar at the catalytic cleft. Furthermore, many MMPs recognize collagen and collagen-like peptides, like those containing the Pro-X-X-X Hy motif described here. Despite these indications of similarity, however, we found that even the closely related gelatinases, MMP-2 and MMP-9, show distinct recognition profiles for small peptide substrates. Others have shown that MMP-3, MMP-7, and MMP-13 also have a unique substrate recognition capability (22,23). These findings lead to questions about how encompassing, and how distinct, the substrate recognition profiles of the MMP family may be.
The objective of this study was to define the scope of substrate recognition by MT1-MMP and to compare its recognition patterns with that of MMP-2 and MMP-9. We were interested particularly in knowing whether MT1-MMP can recognize substrates that other MMPs cannot. Using substrate phage display, we were able to identify four distinct groups of substrates with the catalytic domain of MT1-MMP. Although three of the groups did not exhibit any appreciable selectivity, one of the substrate groups (group IV) is highly selective for MT1-MMP. The defining feature of these substrates is the presence of an Arg residue at the P 4 position. The P 4 Arg residue makes major contributions to the k cat /K m ratio and to the selectivity of the substrates for MT1-MMP (Table II). To our knowledge, this is the first description of a key docking point to the left of the S 3 subsite for MMP substrates. The selective substrates have two other common features. First, they have a hydrophobic residue (often Leu or Ile) at the P 1Ј position. The occupation of this position by hydrophobic residues is not unexpected, because all MMPs contain a deep S 1Ј pocket that has a dominant role in substrate recognition. Second, the selective substrates lack Pro at the P 3 position. The lack of this residue is striking, because the vast majority of MMP peptide substrates selected from this phage library contain such Pro residues (21,28).
The substrates for MT1-MMP illustrate an emerging picture of substrate recognition at the catalytic pocket. The nature of the substrates identified here indicates strongly that cooperativity between subsites is key to recognition and hydrolysis. That is, the binding of substrate at one subsite influences the binding of substrate at another. Both types of substrates contain hydrophobic residues that fit into the deep S 1Ј pocket of MT1-MMP. Beyond this contact point though, the two sets of substrates access different subsites within MT1-MMP. Our studies show that the P 4 Arg residue of the Group IV substrates has a key role in recognition. Molecular modeling indicates a significant role for the methylene groups of this Arg side chain in making hydrophobic contacts with Gly 116 and Phe 204 of MT1-MMP (see Fig. 1). In contrast, the non-selective substrates of group I primarily use the P 3 subsite as key contact point. In agreement, the insertion of a Pro into the P 3 position of the selective substrates can completely override the effects of the Arg at P 4 , presumably by causing a kink in the peptide bond that disallows favorable docking of the P 4 residue.
The information on the structure of the substrates in group IV provides a structural basis for the recognition of at least one know protein substrate for MT1-MMP. Recent work from one of our groups shows that MT1-MMP will cleave transglutaminase (16). The cleavage site within transglutaminase contains the two key features of the group IV substrates, an Arg at P 4 and a hydrophobic residue at P 1Ј . Moreover, there is no Pro residue at P 3 . It is interesting to speculate that other novel substrates for MT1-MMP could be identified through a combination of structural data from group IV substrate and genomic databases.
Although we focused primarily on the contrast between the group I and IV substrates in this study, it should also be emphasized that substrate recognition by MT1-MMP is not limited to just these two binding modes. The substrate phage selections show that MT1-MMP recognizes at least two other sets of substrates (Table I). The group II substrates, which are recognized by MT1-MMP and MMP-2, provide additional support for the idea that cooperative interaction among subsites is important. The consensus sequence of this family indicates that the P 3 and P 1Ј subsites are dominant. Interestingly, most of the peptide substrates that have been selected for the MMPs appear to have key contact points on both sides of the scissile bond. Given that no active MMP has been crystallized with bound substrate, attempts to co-crystallize active site mutants of MMPs with the selective substrates are certainly warranted.
The nature of the substrates identified here suggests unexpected relationships among the MMPs. The substrates selected by MT1-MMP are often recognized by MMP-2 or MMP-9. The Venn diagram in Fig. 2 can describe the nature of the substrates. For example, substrates in group I are recognized equally well by all three MMPs. Substrates in group II are recognized equally well by MT1-MMP and MMP-2. Yet, these peptides are not cleaved efficiently by MMP-9. In contrast, substrates in group III are recognized by MT1-MMP and MMP-9 but not by MMP-2. This outline of substrate recognition could ultimately be expanded to include the entire MMP family. Such an expanded view might be better described by the clustering algorithms popularized for gene array analysis (45,46) rather than Venn diagrams. It is conceivable that matching such clusters with structural information on the subsites of each MMP might give clues to detailed structural basis for the overlap and distinction in substrate recognition.
The potential for distinct methods of recognizing substrate within the catalytic cleft suggests an interesting hypothesis about another potential level of regulation of MMP activity, substrate switching. Although we have investigated active site interactions in this report, the influence on substrate recognition by exosite interactions remains unknown. Protein-protein interactions at domains beyond the catalytic cleft, like the hemopexin domain, could influence allosterically the structure of the catalytic pocket. One might hypothesize a similar role for the cytoplasmic tail of MT1-MMP, which is known to localize MT1-MMP to invasive fronts and regulate trafficking and internalization (47). Inside-out signaling is a well known phenomena in the integrin field (48 -50) and could be at play in regulation of substrate hydrolysis by the membrane-type MMPs. The membrane-type MMPs also contain an additional loop in the catalytic domain, called the MT-specific loop. Recent work shows that this loop regulates the ability of MT-specific loop to activate MMP-2 (51). Although mutations within the MT-loop were without effect on peptide substrates with the Pro-X-X-X Hy motif, the effect of such mutations on the substrate binding mode represented by the group IV peptides is worthy of exploration. Finally, it is tempting to hypothesize that a single substrate could switch dynamically between the two binding modes during catalysis. Such a mechanism could be invoked in the hydrolysis of collagen, which is unwound during hydrolysis by the MMPs (52).
The results of the present study also suggest a path for defining the biological significance of each of the substrate binding modes for MT1-MMP in vivo. We anticipate that point mutations can be made to ablate selectively individual binding modes, without affecting significantly the other modes. The hypothesized residues that form hydrophobic contacts with the P 4 Arg side chain (group IV substrates) or that disallow the kink imposed by Pro at P 3 (group I substrates) are good initial targets for mutation. If such mutations can be made without altering recognition of other classes of substrates, the biological role of each binding mode can be determined. Consequently, future work will aim to determine whether the wide-ranging phenotype of the MT1-MMP-deficient mice can be segregated into effects attributable to each individual substrate binding mode.