Effect of Species Differences on Stromelysin-1 (MMP-3) Inhibitor Potency

For an animal model to predict a compound’s potential for treating human disease, inhibitor interactions with the cognate enzymes of separate species must be comparable. Rabbit and human isoforms of stromelysin-1 are highly homologous, yet there are clear and significant compound-specific differences in inhibitor potencies between these two enzymes. Using crystal structures of discordant inhibitors complexed with the human enzyme, we generated a rabbit enzyme homology model that was used to identify two unmatched residues near the active site that could explain the observed disparities. To test these observations, we designed and synthesized three chimeric mutants of the human enzyme containing the single (H224N and L226F) and double (H224N/L226F) mutations. A comparison of inhibitor potencies among the mutant and wild-type enzymes shows that the mutation of a single amino acid in the human enzyme, histidine 224 to asparagine, is sufficient to change the selectivity profile of the mutant to that of the rabbit isoform. These studies emphasize the importance of considering species differences, which can result from even minor protein sequence variations, for the critical enzymes in an animal disease model. Homology modeling provides a tool to identify key differences in isoforms that can significantly affect native enzyme activity.

MMP activity is essential for tissue remodeling or cellular migration and is required for physiological processes such as fetal development, wound healing, angiogenesis, or inflammatory cell trafficking. However, it is critical to these processes that the matrix degradation be locally confined and temporally limited. Consequently, MMPs are highly regulated enzymes and, in general, are induced only in response to specific stimuli such as cytokines and growth factors. When induced, they are transiently expressed as latent enzymes that require complex interactions for activation (3). Once activated they are subject to rapid autodegradation. In addition, enzyme activity is modulated in situ by potent and specific natural inhibitors, the TIMPs that are expressed under many of the same conditions that induce MMP activity (4). If the balance between induction, activation, and inhibition processes becomes even slightly dysregulated, the result can be chronic and debilitating diseases such as cancer, arthritis, inappropriate angiogenesis, autoimmune disease, aneurysm, atherosclerosis, or heart failure (5).
Because MMP expression in normal tissue is very low, enzyme activity can be associated with specific disorders through the correlation of disease severity with either mRNA or active protein expression. Stromelysin-1 (MMP-3) has been directly implicated in the pathologies of osteo-and rheumatoid arthritis and many types of cancers through these kinds of correlations (6 -8). In addition, this enzyme provides the initial step in the activation cascades for many members of the MMP family and as such may play a less direct role in other disease states as well (9). As a result, MMP-3 inhibition has been an attractive target for pharmaceutical drug discovery.
The MMPs are a homologous family of enzymes with a conserved multidomain structure that serves functionally to enable the proteolytic degradation of complex matrix macromolecules. The enzyme structure comprises a prodomain, which maintains enzyme latency, the catalytic domain that contains the active site, and a C-terminal domain, which is responsible for matrix interactions and substrate and TIMP recognition (10). The catalytic domain can be expressed independently and functions as a competent protease with substrate and inhibitor specificities typical of the full-length parent enzyme (11)(12)(13)(14). Several structures of MMP catalytic domains (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32) and one full-length enzyme (33) have been reported and describe a common protein fold for the MMP active-site domain. This consists of an open-face ␣-␤ sandwich made up of three ␣-helices packed against a 5-stranded ␤-sheet. The overall structure is ellipsoidal with a flat, surface-exposed substrate-binding cleft on one face of the surface. The zinc-binding active site is contained in the second helix and adjacent loop, followed by a characteristic 1-4 ␤-turn (Met-turn) which positions an invariant methionine near the catalytic zinc to support the active site * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  substructure. After a series of turns that define the S 1 Ј pocket, the structure terminates in the third helix.
The substrate-binding cleft is quite shallow, and most of the binding energy of synthetic inhibitors comes from two functional groups, a substituent that chelates the catalytic zinc and a hydrophobic portion that binds in the S 1 Ј pocket (34). The most variable region among MMP structures is this deep, hydrophobic pocket (35). Sequence differences within the S 1 Ј region define the shape and character of the pocket and appear to be responsible for much of the observed substrate specificity of a given MMP (32). These differences have been exploited to develop inhibitors with increased selectivity for (ideally) a single MMP, or (more realistically) a limited subset of family members. Targeting an MMP inhibitor to the specific enzyme(s) associated with a disease should limit the potential side effects of therapeutic use and allow safe, long-term drug treatment for chronic disorders.
With the development of MMP inhibitors with improved enzyme selectivity, it has become increasingly important not only to identify the enzyme associated with a given disease, but also to characterize animal models to assure that they are dependent on that particular MMP. However, even with an established correlation for the disease model, cognate MMPs from different species may differ in the composition of the S 1 Ј pocket, and inhibitor activity in an animal model may not adequately predict efficacy for the human disease.
The rabbit and human isoforms of stromelysin-1 are highly homologous enzymes, and, in general, inhibitors demonstrate comparable potencies when tested against the purified enzymes. However, we have found that certain very selective human stromelysin inhibitors possess reduced potency against the rabbit enzyme. Using crystal structures of human stromelysin catalytic domain-inhibitor complexes as guides, we constructed a rabbit stromelysin catalytic domain homology model, which was used to identify specific amino acid interactions within the S 1 Ј pocket that could explain the observed disparity in inhibitor potencies. We then synthesized chimeric mutants of human stromelysin-1, which allowed us to characterize the inhibitor interactions predicted from homology modeling and to identify the specific region of the S1Ј pocket that was responsible for the observed differences in inhibitor selectivity of the wild-type isoforms.

EXPERIMENTAL PROCEDURES
Gene Constructs and Reagents-The synthetic gene for human MMP-3 CD was constructed using previously reported methodology (36). Codons were optimized for Escherichia coli expression, and silent mutations were introduced to maximize the number and distribution of unique restriction sites throughout the gene. The amino acid sequence of the expressed protein was identical to that previously reported for MMP-3 CD from this laboratory (37). The natural gene for rabbit MMP-3 CD was a gift from D. Kelner and S. Hunt (Parke-Davis Pharmaceutical Research, Ann Arbor, MI) and was designed to express a protein corresponding to the human catalytic domain. All constructs were expressed from the pGEMEX-1 plasmid (Promega, Madison WI) in E. coli strain BL21(DE3)pLysS (Novagen, Madison, WI). Oligonucleotides designed with the mutated sequences of human MMP-3 CD were purchased from Genosys (The Woodlands, TX). E. coli strain DM1 competent cells and ligation and phosphorylation enzymes were from Life Technologies, Inc. (Gaithersburg, MD). Restriction and kinase enzymes were obtained from New England Biolabs (Beverly, MA). LB ampicillin-agar plates and Superbroth with phosphates were from Digene (Beltsville, MD). Precast Tris glycine polyacrylamide gels (8 -16%) were obtained from Novex (San Diego, CA). Thiopeptolide substrate, Ac-Pro-Leu-Gly-[2-mercapto-4-methyl-pentanoyl]-Leu-Gly-OEt, was from BACHEM Bioscience (King of Prussia, PA). All other chemicals were from Sigma.
Homology Modeling-The amino acid sequence of rabbit MMP-3 (accession number P28863) was obtained from the Swiss-Prot data base. The coordinates of SCD complexed with a diphenylpiperidine inhibitor (31) (Brookhaven code 1caq) were used to generate the homol-ogy model. The homology model was built using COMPOSER (38) as implemented in SYBYL (version 6.22) (commercially available from Tripos Associates, St. Louis, MO). The sequence corresponding to the catalytic domain was generated by removing residues 1-118 and 462-731 from P28863. The resulting sequence was 86.2% identical to 1caq as determined by COMPOSER. Hydrogens were added to the model that emerged from COMPOSER, and the protein was subjected to minimization within SYBYL. 2 No substantial change in overall geometry occurred during minimization beyond changes in local geometries about individual residues. With the exception of arginine 148 (-in a disallowed region), the resulting structure passed all stereochemical and geometric checks for a 2.0-Å resolved structure in PROCHECK (version 3.0.1) (41). This arginine occurs in a loop on the other side of the protein 18.4 Å from the catalytic zinc, and was not manipulated further. The resulting structure is quite similar to 1caq (Fig. 4), with the largest differences seen in the loop containing residues 224 -226. These differences were expected, since the minimization was run on the apo homology model and the loop collapsed somewhat to achieve closer packing with the rest of the protein. Such loop motion in this region has been observed previously in other MMP structures (31,42).
Site-directed Mutagenesis-The following pairs of oligonucleotides were used to introduce mutations into human MMP-3 CD by cassette mutagenesis (mutant codons are underlined): Unique restriction sites were introduced into each cassette through silent mutations to facilitate the identification of mutant clones (sites in italics). All oligos were designed with the MluI restriction site at the 5Ј-site and AvrII at the 3Ј-site. Forward and reverse strands were annealed at 95°C for 10 min, then slowly cooled to room temperature. The resulting oligo duplexes were phosphorylated with T4 polynucleotide kinase and ligated into pGEMEX-1 containing the synthetic human MMP-3 CD gene which had been previously digested (MluI and AvrII) and dephosphorylated (calf intestine alkaline phosphatase). Plasmids were transformed into DM1 competent cells and screened by restriction mapping for the unique sites introduced with the oligo cassettes. Mutant gene sequences were confirmed by dideoxy sequencing of the complete coding sequence.
Protein Expression and Purification-MMP-3 CD plasmids were transformed into BL21(DE3)pLysS competent cells. Two-liter cultures of the freshly transformed expressing strains were grown overnight in shaking flasks at 37°C in Superbroth with phosphates supplemented with ampicillin (50 g/ml) and chloramphenicol (50 g/ml) and induced with 1 mM isopropyl-thio-1-␤-D-galactopyranoside for 2 h before harvesting by centrifugation. Mutant and wild-type MMP-3 CDs were all strongly expressed as insoluble proteins under these conditions and represented about 30% of the cellular protein. Active catalytic domains were refolded and purified as described previously for human MMP-3 CD (37), except that refolded material was centrifuged, concentrated, dialyzed, and applied directly to a Q-Sepharose column (Pharmacia LKB, Piscataway, NJ) without ammonium sulfate fractionation.
Enzyme Assays and Steady State Kinetic Determinations-Initial rates for hydrolysis of the thiopeptolide substrate coupled to a reaction with 5,5Ј-dithiobis(2-nitrobenzoic acid) were used to assess the catalytic activity of mutant and wild-type enzymes. The change of absorbance at 405 nm was monitored continuously at room temperature using a SPECTRAmax 340 microplate reader (Molecular Devices, Sunnyvale, CA). A typical 100-l reaction contained 50 mM MES, pH 6.0, or 50 mM HEPES, pH 7.0, with 10 mM CaCl 2 , 100 M substrate, 1 mM 5,5Јdithiobis(2-nitrobenzoic acid), and 5 nM human or 10 nM rabbit enzyme. For K i determinations, inhibitors were included in the assay at appro-priate concentrations resulting in a final reaction mixture concentration of 2% dimethyl sulfoxide. Under these conditions, K i values typically vary from 5 to 10% in replicate determinations. Estimations of K m and k cat were obtained from plots of reaction velocity versus substrate concentration assuming a Michaelis-Menten mechanism, using substrate concentrations ranging from 0 to 2 mM. The least squares fit of the data to a hyperbola was calculated using the equation Because it was known that these inhibitors are bound in the enzyme active site, values for binding affinity, K i , could be calculated from the relationship K i ϭ IC 50 /(1 ϩ [S]/K m ) using experimentally derived IC 50 values (43).
Inhibitors-The compounds used in this study were synthesized as a part of a medicinal chemistry program supporting the development of small molecule inhibitors selective for MMP-3 (39,40,44). The specific compounds comprising each of the four subsets described in Fig. 3 were selected to represent a broad range of inhibitor potencies against human stromelysin-1. The K i values within each subset of the non-peptidic compounds ranged from low nanomolar to high micromolar. The reference compounds were consistently more potent, demonstrating low nanomolar potency. The carboxylate zinc-chelating group in the generic template was replaced in some cases with hydroxamic acid. Other modifications included the substitution of a sulfonamide or oxime group for the ␥-keto functionality, the addition of R-groups at the ␤-position of the side chain, or the variation of the length of the side chain. These modifications were reasonably well represented among all three small molecule subsets.

Homology Modeling Predicts S 1 Ј Pocket Interactions That May Be Responsible for Differences in Inhibitor Potency between
Human and Rabbit MMP-3 CD-Human and rabbit stromelysin-1 are extraordinarily homologous enzymes, sharing 91% identity of the primary amino acid sequence for the catalytic domains ( Fig. 1) (45). Human MMP-3 has a sharp acidic optimum for catalytic efficiency, resulting from an increasing K m at neutral pH (46 -49). Because this seems to be a physiologically relevant feature of this enzyme, we have routinely characterized stromelysin-1/inhibitor interactions at acidic pH (13). Despite its high degree of homology to the human isoform, rabbit stromelysin has a much broader pH profile, more characteristic of other MMP family members, with a general shift to neutrality of the optimal catalytic activity (Table I). At pH 6.0, the k cat /K m value for the rabbit enzyme is about half that of the human isoform, mainly due to differences in K m . It has been shown that the specific architecture of the stromelysin-1 S1Ј pocket can affect substrate binding (46), and the observed difference in catalytic efficiency for these two enzymes at acidic pH probably reflects diverging substrate preference resulting from the pH dependence of K m . Nevertheless, the catalytic efficiencies of human and rabbit stromelysin-1 at pH 6 are reasonably similar, and most synthetic inhibitors show comparable inhibition profiles when tested against these two species isoforms. However, some very selective human MMP-3 inhibitors were, as a class, much less active against the rabbit enzyme. Fig. 2 shows the correlation of K i values determined for the rabbit and human enzymes of a cross-section of MMP inhibitors. The test group included the four general classes of synthetic, substrate-competitive inhibitors described in Fig. 3: (a) dibenzofuran, (b) biphenyl, (c) diphenylpiperidine, and (d) reference peptidic compounds. The functional groups interacting with the active site zinc were varied to provide each set with compounds that demonstrated a fairly wide range of K i values and included both strong and weak zinc-chelating groups.
As a class, diphenylpiperidines segregate from the other inhibitors demonstrating significantly less activity against the rabbit enzyme. We had previously shown that the tertiary phenyl ring in this series was crucial to activity against the human enzyme (50) and expected that the differences between rabbit and human enzyme potencies were due to interactions deep within the S 1 Ј binding pocket. Analysis of the primary sequence of the S 1 Ј region identified 5 residues (Fig. 1) that differed between the two isoforms, yet these differences represented conservative mutations which, on the surface, did not explain the observed disparity of inhibitor potencies.
Using the crystal structures of inhibitor-complexed human MMP-3 CD, we constructed a homology model of the rabbit enzyme and used it to identify a 3-residue sequence that could affect inhibitor binding within the S 1 Ј pocket (residues 224 through 226 in Fig. 1). Because the side chain of alanine 225 (serine in the human sequence) was solvent exposed and positioned outside the pocket, it was not expected to interact significantly with the diphenylpiperidine inhibitors and was not included in the mutagenesis studies. In contrast, asparagine 224 and phenylalanine 226 both had side chains that extended into the S 1 Ј pocket (Fig. 4).
The rabbit homology model was constructed to represent an apoenzyme and is shown here with the "collapsed" S 1 Ј subsite characteristic of MMP-3 CD structures where the P 1 Ј side chain does not penetrate deeply into the specificity pocket (26). Although asparagine 224 in this model is stabilized by a hy-

TABLE I Steady-state kinetic parameters for wild-type and mutant MMP-3CD
Kinetic constants were determined for the hydrolysis of the thiopeptolide substrate, Ac-Pro-Leu-Gly-[2-mercapto-4-methyl-pentanoyl]-Leu-Gly-OEt, at pH 6.0 (or pH 7.0 as noted) in the presence of 10 mM CaCl 2 , using either 10 nM rabbit or 5 nM human enzyme.

. Primary amino acid sequence of stromelysin-1 catalytic domain (human and rabbit isoforms).
Amino acid variations are highlighted in bold. The underlined sequence in the S 1 Ј pocket region represents the target of these mutagenesis studies. S designates individual strands of the ␤-sheet, and H, the ␣-helices. Asterisks indicate the zincbinding histidine residues in the active site. The # marks the conserved methionine of the 1-4 ␤-turn (Met-turn). drogen bond across the empty pocket with the carbonyl oxygen of alanine 217, we have previously demonstrated for the human catalytic domain that this region of the S 1 Ј loop is flexible and can relax to accommodate a larger P 1 Ј group (31). When complexed with the diphenylpiperidine class of inhibitors, the rabbit enzyme is expected to adopt the more open structure of the human enzyme in Fig. 4. However, in the human enzyme, the ring structure of histidine 224 allows asandwich stacking interaction with the terminal phenyl ring of the inhibitor (31). This interaction would be eliminated with the substitution of asparagine in the rabbit sequence, which could result in a loss of inhibitor potency for the rabbit enzyme.
Slightly deeper in the pocket, phenylalanine 226 (rabbit sequence) corresponded to a leucine in the human sequence. It was possible that the bulkier residue in the rabbit enzyme formed a narrower channel and interfered with the tight packing observed for structures of diphenylpiperidine inhibitors complexed with the human catalytic domain (31). Alternatively, the aromatic ring of phenylalanine might enhance the binding of these inhibitors by providing stacking interactions with the terminal ring of the inhibitor, thus compensating, at least partially, for the loss of interactions with histidine 224.

Site Mutants of Human MMP-3 CD Assess Inhibitor Interactions Predicted by Homology Modeling of the Rabbit Enzyme-
The synthetic gene coding for human MMP-3 CD facilitated the construction of chimeric mutants to test the predictions from our computer modeling studies. It was unclear from the model alone whether we were looking at the positive effect on inhibitor binding of His-224 in the human enzyme or the negative (or compensatory) effect of Phe-226 in the rabbit protein. We designed and synthesized both the single (H224N and L226F) and the double (H224N/L226F) chimeric mutants. Mutant and wild-type proteins were expressed and purified from bacterial cultures. All mutants were purified to homogeneity as defined by a single band on Coomassie Blue-stained protein gels (data not shown). The isolated enzymes were characterized for k cat and K m using the thiopeptolide substrate (Table I). K m values for the mutants were almost identical to the wild-type human enzyme, indicating that these mutations did not affect substrate binding or significantly disrupt the critical folding of the enzymes. The main differences in catalytic efficiency resulted from decreased k cat . This may reflect a slight repacking of the residues around the active site that support and position the catalytic zinc, since it is difficult to design totally benign site mutations. Overall, there is a good agreement of catalytic efficiencies among the mutant and wild-type enzymes.
K i values were determined for the mutants using the group of inhibitors described in Fig. 3, and these values were compared with the inhibitor activities against the parent enzymes (Fig. 5). For all mutants, there was no effect on the observed potencies of the dibenzofuran, biphenyl, or peptidic inhibitors. The K i values for these classes of inhibitors correlated well with values obtained for both the human and rabbit wild-type enzymes, further supporting the conclusion that the S 1 Ј pocket mutations did not interfere with either the overall integrity of the active site substructure or the ligation of the catalytic zinc. However, diphenylpiperidine inhibitors were 10 -100-fold less potent versus the H224N mutant compared with human MMP-3 CD exhibiting an inhibitor profile identical with the rabbit enzyme (Fig. 5, Panel d). The mutation of L226F had no effect on inhibitor interactions, either as a single mutation (profile matched the human enzyme) or in combination with H224N (profile matched the single H224N mutant). These correlations demonstrate that the mutation of a single amino acid in MMP-3 CD, H224N, predicted by homology modeling of the rabbit isoform, was sufficient to change the inhibitor selectivity of the human enzyme to that observed for the rabbit catalytic domain.

DISCUSSION
The recent burst of structures for matrix metalloproteinases complexed with small molecule inhibitors has shown that the S 1 Ј pocket is critical in defining inhibitor selectivity. Minor sequence differences within this region can change the shape and hydrophobic character of the pocket resulting in wide variations in inhibitor potencies for a particular compound among members of this family of homologous enzymes (32). These binding interactions have been exploited to develop increas-

FIG. 3. Representative classes of synthetic MMP inhibitors.
Four sets of inhibitors were used to characterize the effect of species differences on inhibitor potencies for stromelysin-1. Examples within each set were chosen so that the portion of the molecule that interacted with the active site zinc would vary to describe a wide range of inhibitor activities, including both strong and weak zinc chelators.
ingly potent and selective inhibitors through structure-based drug design processes. Improving selectivity for the human target, however, may result in altered activity for the cognate enzyme in a given animal model of disease. Since it is impractical to test all inhibitors against the purified enzymes from all species used for animal models, it may be possible to predict the species-specific activity of synthetic inhibitors using computergenerated homology models of the animal enzymes.
The rabbit and human isoforms of stromelysin-1 are highly homologous enzymes, demonstrating over 90% identity for the primary amino acid sequence of the catalytic domains. Catalytic efficiencies for synthetic substrates are similar, and, in general, inhibitors show comparable potencies when tested against the purified enzymes. Unexpectedly, one class of very selective inhibitors for human stromelysin-1 was much less potent against the rabbit enzyme. Using crystal structures of these inhibitors complexed with the human catalytic domain, we generated a rabbit homology model that was used to identify specific amino acids within the S 1 Ј pocket that might be responsible for the disparity in inhibitor potencies. The model showed that of the several amino acid differences between the two enzymes, only asparagine 224 and phenylalanine 226 of the rabbit enzyme were positioned within the S 1 Ј pocket, and both could potentially impact the binding efficiency of the diphenylpiperidine class of inhibitors. Using human to rabbit chimeric mutants, we showed that the mutation of a single amino acid, histidine 224 to asparagine, changed the inhibitor selectivity of the human enzyme to that of the rabbit isoform. This indicated that the disparity of inhibitor potencies observed for the diphenylpiperidine inhibitors was due to the loss of positive stacking interactions with histidine 224 of the human enzyme, as predicted by the homology model.
Murine and rat stromelysin-1 have a lysine residue at position 224 (45), and those isoforms would be expected to have diminished activity against the diphenylpiperidine inhibitors as well. Since both rabbit and rodent models are commonly used in preclinical studies, it is noteworthy that inferences about stromelysin inhibition in the corresponding human disease may not be valid for at least some classes of compounds.
As a family the MMPs are neutral endopeptidases, demonstrating broad bell-shaped curves for the dependence of catalysis on pH with generally uniform and robust activity from pH 5-9. Human stromelysin is unique among the MMP family members in that it has a sharp catalytic optimum between pH 5.5 and 6.0 (46,47). The fall-off of catalytic efficiency at neutral pH is dramatic with proteolytic activity at pH 7.0 about 32% of the activity at pH 6.0 (Table I). This appears to be an intrinsic property of stromelysin-1 since this effect is observed for both peptide and natural substrates, using either the catalytic domain or full-length enzyme (37, 46 -49). Latent stromelysin-1 activates spontaneously at pH 5.5 (51) and is subject to much more rapid autodegradation (37). In addition, TIMP interactions with active MMP-3 are weakened at acidic pH (52), indicating that the exceptional pH dependence of stromelysin activity may serve as an additional layer of post-translational regulation in vivo.
Holman and co-workers (46) constructed the H224Q mutant of human MMP-3 CD and showed that this single mutation released the enzyme from its acidic pH optimum resulting in strong catalytic activity over a broad pH range. In this case, the effect of the mutation was believed to be the stabilization of a pH-dependent S 1 Ј substructure that is required in the wildtype enzyme for efficient substrate binding and catalysis. His-224 has been shown to form a hydrogen bond across the S 1 Ј pocket with the alanine 217 carbonyl, an interaction that tethers the flexible outer loop of the S 1 Ј pocket and, in effect, serves to stabilize the S 1 Ј subsite (26,31). The unprotonated form of His-224 cannot form this bond, and, above its pK a (predicted to be near pH 6), this interaction would be lost in the wild-type but not the mutant enzyme. In our characterization of the rabbit isoform (asparagine at position 224), we have also observed a broad pH dependence for catalytic activity that is, not surprisingly, very similar to the H224Q mutant activity. Considering these results, it is possible that human stromelysin-1 FIG. 4. Superposition of rabbit stromelysin catalytic domain homology model (cyan tube, orange residues) and the crystal structure of the human isoform (yellow tube, yellow residues). Zinc and calcium ions (magenta and green spheres, respectively) from both structures have been added. The catalytic zinc-ligating histidines and a conserved methionine are shown, as are the residues that were mutated. A diphenylpiperidine inhibitor from 1 caq in the Protein Data-Bank is shown in its crystallographically determined orientation, in a CPK representation color coded by atom type. Note thestacking interaction that occurs between histidine 224 and the terminal phenyl ring of the inhibitor that is lost when this residue is mutated to an asparagine.
has somewhat different physiologic regulation and substrate activity than the corresponding rabbit and rodent enzymes, and disease models in these species may not accurately describe the underlying mechanism of human disease if indeed stromelysin-1 plays a key role in the pathology.
The studies described here emphasize the importance of understanding inhibitor and substrate interactions with species isoforms of the enzymes believed to be active in a given model of human disease. This goes beyond a simple accounting of primary amino acid sequence similarity or identity, since even a single amino acid difference can result in significant differences in potency between the human and animal enzymes. Using a homology model of rabbit stromelysin-1 based on the crystal structure of the human enzyme, we predicted specific amino acid interactions within the S 1 Ј pocket that could be responsible for the differences in inhibitor potency between the rabbit and human enzymes. The biochemical analyses of chimeric mutants confirmed and clarified the model predictions. These data support the use of homology modeling coupled with site-directed mutagenesis to assess inhibitor interactions that may be due to minor sequence differences between cognate enzymes from separate species.