The Intermediate S1′ Pocket of the Endometase/Matrilysin-2 Active Site Revealed by Enzyme Inhibition Kinetic Studies, Protein Sequence Analyses, and Homology Modeling*

Human matrix metalloproteinase-26 (MMP-26/endometase/matrilysin-2) is a newly identified MMP and its structure has not been reported. The enzyme active site S1′ pocket in MMPs is a well defined substrate P1′ amino acid residue-binding site with variable depth. To explore MMP-26 active site structure-activity, a series of new potent mercaptosulfide MMP inhibitors (MMPIs) with Leu or homophenylalanine (Homophe) side chains at the P1′ site were selected. The Homephe side chain is designed to probe deep S1′ pocket MMPs. These inhibitors were tested against MMP-26 and several MMPs with known x-ray crystal structures to distinguish shallow, intermediate, and deep S1′ pocket characteristics. MMP-26 has an inhibition profile most similar to those of MMPs with intermediate S1′ pockets. Investigations with hydroxamate MMPIs, including those designed for deep pocket MMPs, also indicated the presence of an intermediate pocket. Protein sequence analysis and homology modeling further verified that MMP-26 has an intermediate S1′ pocket formed by Leu-204, His-208, and Tyr-230. Moreover, residue 233 may influence the depth of an MMP S1′ pocket. The residue at the equivalent position of MMP-26 residue 233 is hydrophilic in intermediate-pocket MMPs (e.g. MMP-2, -8, and -9) and hydrophobic in deep-pocket MMPs (e.g. MMP-3, -12, and -14). MMP-26 contains a His-233 that renders the S1′ pocket to an intermediate size. This study suggests that MMPIs, protein sequence analyses, and molecular modeling are useful tools to understand structure-activity relationships and provides new insight for rational inhibitor design that may distinguish MMPs with deep versus intermediate S1′ pockets.

Matrix metalloproteinases (MMPs, 1 matrixins) are believed to participate in angiogenesis, embryonic development, morphogenesis, reproduction, tissue resorption and remodeling, and tumor growth, progression, invasion, and metastasis through breakdown of the extracellular matrix, cell surface proteins, and processing growth factors, cytokines, and chemokines (1)(2)(3). Recently, human MMP-26 (endometase/matrilysin 2) was identified and its mRNA expression was detected in normal tissues of the human uterus and placenta, and in many types of malignant tumors (4 -7). Characterization of the MMP-26 promoter suggests that this proteinase may be expressed in cancer cells of epithelial origin (8). MMP-26 may play an important role in human prostate and breast cancer invasion (9 -10).
MMP-26 cleaves type I gelatin, ␣ 1 -proteinase inhibitor, fibrinogen, fibronectin, vitronectin, type IV collagen, and insulin-like growth factor binding protein-1 (4,7,11). Studies of MMP-26 indicate that it has substrate specificity similar to other MMPs, with the exception of a preference for Ile at the P 2 and P 2 Ј positions, for small residues at the P 3 Ј and P 4 Ј positions, and Lys at the P 4 position (11). MMP-26 also hydrolyzes several synthetic fluorogenic peptide substrates designed for stromelysin-1, gelatinases, collagenases, and tumor necrosis factor-␣ converting enzyme (4,11). According to these peptide substrate studies, MMP-26 may be capable of cleaving a broad range of substrates, although it has less catalytic efficiency than other MMPs. X-ray crystal structures of MMPs illustrate that overall topology and secondary structures are conserved (12)(13)(14)(15)(16)(17)(18). The S 1 Ј pocket, a hydrophobic pocket of variable depth, is a well defined substrate P 1 Ј-binding site in MMPs. Three types of S 1 Ј pockets can be distinguished from the available structures of MMPs (19 -20). One type is a shallow pocket, as found in MMP-1 (human fibroblast collagenase; 13) and MMP-7 (matrilysin; 16), where the pockets are limited by the side chains of Arg and Tyr, respectively, crossing the pockets. Many of the structurally known MMPs possess Leu at the corresponding site, and its side chain forms the top of the pocket rather than crossing the pocket. These Leu-containing MMPs may be further classified as deep and intermediate S 1 Ј pocket MMPs. A deep, tunnel-like pocket is found in MMP-3 (stromelysin-1; 12), MMP-12 (metalloelastase; 17), and MMP-14 (MT1-MMP; 21), whereas MMP-2 (gelatinase A; 22), MMP-8 (human neutrophil collagenase; 15), and MMP-9 (gelatinase B; 23) possess an intermediate-sized pocket, which is neither deep nor shallow. An enzyme with a shallow pocket prefers large, aliphatic residues in the P 1 Ј position, such as Leu and Met (24 -25). The remainder of the MMPs can accommodate larger amino acid derivatives, such as homophenylalanine, in the P 1 Ј position (26).
MMP-26, composed of 261 amino acid residues and lacking a hemopexin-like domain, represents the smallest member of the MMP family. The primary structure of MMP-26 can be divided into three regions that include a signal peptide, a propeptide domain, and a catalytic domain. MMP-26 identification, ex- pression, and substrate specificity have been explored by several groups (4 -11). However, the S 1 Ј pocket characteristics of MMP-26 are unknown because of the absence of an MMP-26 x-ray crystallographic structure. Therefore, in this study we have utilized previously characterized and newly developed mercaptosulfide MMPIs (27)(28)(29) together with protein sequence analyses and molecular modeling to understand the S 1 Ј pocket characteristics of MMP-26.

EXPERIMENTAL PROCEDURES
Materials-The fluorescent peptide substrates for MMPs used in this study were purchased from Bachem Chemical Co. The metal salts and Brij-35 were purchased from Fisher Scientific Inc. The hydroxamate MMPIs 444237, 444238, 444225, and GM6001 were purchased from Calbiochem. All other chemicals were purchased from Sigma.
The mercaptosulfide inhibitors were prepared and characterized as previously described (27)(28)(29). cis-1-Acetylthio-2-tert-butoxycarbonylthiocyclopentane and cis-3-acetylthio-4-tert-butoxycarbonylthio-N-tertbutoxycarbonylpyrrolidine were synthesized (29) and S-alkylated with (2S)-2-bromo-4-methylpentanoic acid or (2S)-2-bromo-4-phenylbutanoic acid; the latter bromoacids were derived from L-leucine and L-homophenylalanine, respectively (27). Subsequent coupling with L-PheNHMe or L-leucine-p-methoxyanilide (27) afforded the S-Boc and N-Boc protected inhibitors as mixtures of two diastereomers. The N-Boc group was selectively removed and replaced by the other acyl groups (29). The diastereomers were separated by flash chromatography on silica gel or by reverse-phase preparative high performance liquid chromatography on a C18 column. The slower-eluting S-Boc protected diastereomer exhibited the more potent MMP inhibition in each case. Its stereochemistry was assigned by 1 H NMR NOE analysis (MAG-182), x-ray crystallography (YHJ-294-2) (29), or by analogy. Finally, the S-Boc protecting groups were removed by brief treatment with 2 N HCl in acetic acid and the mercaptosulfide inhibitors were isolated by lyophilization of the reaction mixture.   Boldface amino acid residues form the S 1 Ј pocket. Italicized sequences are metal binding consensus sequences. Underlined residues may determine S 1 Ј pocket characteristics. To align MMP-2 and MMP-9, the 183-residue insert of fibronectin type II-like modules were deleted before the alignment. The residue numbering system is based on the sequence of MMP-26 (4).
Kinetic Assays and Inhibition of Endometase-The substrate Mca-PLGLDpaAR-NH 2 was used to measure inhibition constants (11,33). Enzymatic assays were performed at 25°C in 50 mM HEPES buffer at pH 7.5 in the presence of 10 mM CaCl 2 , 0.2 M NaCl, and 0.01 or 0.05% Brij-35 with substrate concentrations of 1 M. The release of product was monitored by measuring fluorescence (excitation and emission wavelengths of 328 and 393 nm, respectively) with a PerkinElmer luminescence spectrophotometer LS 50B connected to a temperature controlled water bath. All stock solutions of inhibitors were in methanol. For inhibition assays, 10 l of inhibitor stock solution, 176 l of assay buffer, and 10 l of enzyme stock solution were mixed and incubated for 30 to 60 min prior to initiation of the assay, which was accomplished by adding and mixing 4 l of the substrate stock solution. Enzyme concentrations ranged from 0.2 to 7 nM during the assay.
Determination of Mercaptosulfide Inhibitor Concentration-The active inhibitor concentrations were estimated by titrating the mercapto group with 5,5Ј-dithiobis(2-nitrobenzoic acid) (Ellman's reagent) as described previously (36, 37). Briefly, the reaction of 5,5Ј-dithiobis(2nitrobenzoic acid) with the mercapto group produces 2-nitro-5-thiobenzoic acid. The concentration of 2-nitro-5-thiobenzoic acid is then measured by monitoring the absorbance at 412 nm. Cysteine was used to generate the standard curve with a molar extinction coefficient of 14,000 Ϯ 500 M Ϫ1 cm Ϫ1 , which is close to the value in the literature (37).

Inhibition of MMPs with Mercaptosulfide
MMPIs-An inhibitor set consisting of eight mercaptosulfide inhibitors was chosen to evaluate the S 1 Ј pocket of MMP-26 (Fig. 1). These inhibitors contain P 1 Ј and P 2 Ј residues and have a mercapto and a sulfide group as a possible bidentate metal-binding moiety. The inhibitors contain a Leu side chain (MAG-181 and -182 and YHJ-294-1 and -2) or a Homophe side chain (YHJ-72, -73, -74, and -75) at the P 1 Ј site. These inhibitors were tested against MMPs with known pocket characteristics (MMP-1-3, -7-9, -12, and -14). The inhibition potency of this class of inhibitors for the MMPs is significantly enhanced with a ␤-H configuration at the five-membered ring containing the mercapto and sulfide groups. The inhibitors with a Leu side chain are more potent against the shallow pocket MMPs, MMP-1/human fibroblast collagenase, and MMP-7/matrilysin than those with a Homophe side chain. Inhibitors with a Homophe side chain (YHJ-72, -73, 74, and -75) were more potent against the known deep-pocket MMPs such as MMP-3, -12, and -14 than those FIG. 4. A modeled structure of MMP-26 complexed with YHJ-294-2. A, the overall protein structure is shown as a molecular surface and the residues coordinating the catalytic Zn(II) (magenta sphere) are represented as brown sticks (His-208, His-212, and His-218). The inhibitor YHJ-294-2 is represented as a tube with atoms colored as follows: green, carbon; red, oxygen; blue, nitrogen; and yellow, sulfur. B, a close-up view of the S 1 Ј pocket reveals that Leu-204, His-208, and Tyr-230 may be involved in formation of the pocket walls represented as pink molecular surfaces. The depth of the pocket may be limited by His-233 (light blue molecular surface). The three His residues coordinating the Zn(II) (blue sphere) are represented by tubes colored as described in A. The homology-modeled structure of MMP-26 was generated with the Swiss Model program (38 -40) using the x-ray crystallographic structure of a cd-MMP-12-inhibitor complex as a template (Protein Data Bank code 1JK3) (17). The resulting MMP-26 structure was docked with YHJ-294-2 and energy-minimized as described under "Experimental Procedures." with Leu side chain. The inhibitors with the Leu side chain at the P 1 Ј site (MAG-182 and YHJ-294-2) inhibit MMP-7 (40 and 26 nM, respectively) and MMP-12 (130 and 93 nM, respectively) without significant differences in K i app values. However, the presence of Homophe at the P 1 Ј site dramatically distinguishes MMP-12 from MMP-7. YHJ-73 efficiently inhibits MMP-12 (13 nM), however, the potency is decreased to 1 M against MMP-7. This trend is also displayed by YHJ-75, which has a high nM K i app value against MMP-7 (300 nM) but retains potency against MMP-12 (5.6 nM). This dramatic change of potency because of changes in the P 1 Ј site of the inhibitors is consistently observed with the remaining shallow-and deep-pocket MMPs.
MMPs with an intermediate pocket can also accommodate the Homophe at the P 1 Ј residue. However, the difference in inhibitor potency observed with Leu or Homophe at the P 1 Ј residue is not as remarkable as that in the shallow-and deeppocket MMPs. Inhibitors containing Leu at the P 1 Ј site (MAG-182 and YHJ-294-2) are only slightly more potent against MMP-2 and MMP-9 than inhibitors with Homophe (YHJ-73 and -75). These Homophe inhibitors are still potent against MMP-8 with K i app values in the low nanomolar range. In general, these results indicate that mercaptosulfide inhibitors are suitable for characterizing the S 1 Ј pocket of MMPs.
Characteristics of the S 1 Ј Pocket of MMP-26 as Probed by Mercaptosulfide MMPIs-Inhibition constants for the inhibitors in Fig. 1 were measured with MMP-26 (Table I) Characterization of MMP-26 S 1 Ј Pocket Using Commercial Hydroxamate MMPIs-The S 1 Ј site of MMP-26 was further investigated with commercially available inhibitors (Fig. 2). MMP-7/matrilysin was selected as a representative member of the shallow S 1 Ј pocket MMPs and MMP-12/metalloelastase as one of the deep S 1 Ј pocket MMPs for comparison purposes. The K i app values of the inhibitors with MMP-7, MMP-12, and MMP-26 are summarized in Table II. GM6001 is a broadspectrum and potent inhibitor of MMPs (K i app ϭ 0.4 nM for MMP-1, 0.5 nM for MMP-2, 27 nM for MMP-3, 0.1 nM for MMP-8, and 0.2 nM for MMP-9) (46). It is also the most potent synthetic MMP-26 inhibitor tested, with a K i app value of 0.36 nM. It contains a Leu residue at the P 1 Ј site, and inhibits MMP-7 (3.7 nM) and MMP-12 (3.6 nM) with similar K i app values as observed in the mercaptosulfide inhibitors with a Leu side chain at the P 1 Ј site. The potent inhibitor 444237 of deep S 1 Ј pocket MMPs and its less potent stereoisomer 444238 were designed for human MMP-8 (IC 50 ϭ 4 nM and 1 M, respectively; 45). Inhibitor 444225 was designed to be a potent deep S 1 Ј pocket inhibitor of MMP-3 (K i ϭ 130 nM; 47). The 4-methoxybenzenesulfonyl group of these inhibitors binds at the deep S 1 Ј pocket according to the crystallographic structure (45) and the structure-activity relationship of several derivatives (47). They inhibit MMP-7 and MMP-12 with at least 150-fold lower K i app values for MMP-12 than MMP-7. These deep S 1 Ј pocket inhibitors effectively inhibited MMP-26 with at least 90-fold lower K i app values than those of MMP-7, but were more potent against MMP-12. These results are consistent with MMP-26 having an intermediate S 1 Ј pocket.  (19). Thus, homology modeling and protein sequence alignment may be useful tools to predict key residues involved in forming the S 1 Ј pocket of MMP-26. Protein sequence alignment in Fig. 3 reveals a plausible explanation for residues participating in the formation of the S 1 Ј pocket of MMP-26. According to the alignment, Leu-204, His-208, and Tyr-230 may be key residues in forming the S 1 Ј pocket of MMP-26. To evaluate the prediction from the alignment, a homology modeled structure of the MMP-26 catalytic domain was constructed using the Swiss Model program (38 -40) and the crystal structure of the MMP-12-inhibitor complex (Protein Data Bank number 1JK3) (17) as a template. The mercaptosulfide inhibitors were docked into the modeled MMP-26 structure using MacroModel version 7.2. The docked structures were further energy minimized as described under "Experimental Procedures." The overall MMP-26 structure complexed with YHJ-294-2 is shown in Fig. 4A. Consistent with other MMP family members (19), the non-primed (left) side of the MMP-26 active site is relatively flat. The primed (right) side extends deeper into the surface and the well defined S 1 Ј pocket is clearly visible. The pocket that is formed by Leu-204, His-208, and Tyr-230 is illustrated in Fig. 4B. Interestingly, the depth of the pocket may be limited by His-233, consistent with the intermediate size prediction. DISCUSSION The inhibition characteristics of MMP-26 with mercaptosulfide inhibitors (Table I) and hydroxamate inhibitors (Table II) indicate that MMP-26 does not have a shallow S 1 Ј pocket. According to the protein sequence alignment in Fig. 3  A structural comparison of MMP-26 with MMP-8 further supports the similarity between the S 1 Ј pockets of these two enzymes. The overlapping structures of MMP-8 (Protein Data Bank number 1BZS) (45) and MMP-26 at the S 1 Ј pocket are displayed in Fig. 5. In MMP-8, it is known that the depth of the S 1 Ј pocket is restricted by the Arg-233 side chain projecting toward the catalytic Zn(II) (14). In MMP-26, His-233 is present in place of Arg-233, which may restrict the depth of the pocket in a similar fashion, rendering the S 1 Ј pocket to an intermediate size.
Based on the findings provided in this study and x-ray crystallographic structures of MMPs, the residue at the position MMPs can be divided into three groups based on the characteristics of their S 1 Ј pockets: shallow-, intermediate-, and deep-pocket MMPs (Fig. 7). Enzyme inhibition kinetic studies using MMPIs in combination with protein sequence analysis and homology modeling reveal that MMP-26 has an intermediate S 1 Ј pocket. Our data may provide important mechanistic and structural information to design MMP-26-specific inhibitors. As the need for innovations and new strategies for MMP inhibition in cancer and inflammation is increasing (49,50), this study may shed light on the molecular mechanisms by which highly selective and specific inhibitors targeting an individual MMP or subgroups of MMPs may be rationally designed and developed.