Potent Mechanism-based Inhibitors for Matrix Metalloproteinases*

Matrix metalloproteinases (MMPs) are zinc-dependent endopeptidases that play important roles in physiological and pathological conditions. Both gelatinases (MMP-2 and -9) and membrane-type 1 MMP (MMP-14) are important targets for inhibition, since their roles in various diseases, including cancer, have been well established. We describe herein a set of mechanism-based inhibitors that show high selectivity to gelatinases and MMP-14 (inhibitor 3) and to only MMP-2 (inhibitors 5 and 7). These molecules bind to the active sites of these enzymes, initiating a slow binding profile for the onset of inhibition, which leads to covalent enzyme modification. The full kinetic analysis for the inhibitors is reported. These are nanomolar inhibitors (Ki) for the formation of the noncovalent enzyme-inhibitor complexes. The onset of slow binding inhibition is rapid (kon of 102 to 104 M-1 s-1 and the reversal of the process is slow (koff of 10-3 to 10-4 s-1). However, with the onset of covalent chemistry with the best of these inhibitors (e.g. inhibitor 3), very little recovery of activity (<10%) was seen over 48 h of dialysis. We previously reported that broad spectrum MMP inhibitors like GM6001 enhance MT1-MMP-dependent activation of pro-MMP-2 in the presence of tissue inhibitor of metalloproteinases-2. Herein, we show that inhibitor 3, in contrast to GM6001, had no effect on pro-MMP-2 activation by MT1-MMP. Furthermore, inhibitor 3 reduced tumor cell migration and invasion in vitro. These results show that these new inhibitors are promising candidates for selective inhibition of MMPs in animal models of relevant human diseases.

progression and gelatinase expression and activity has been well established in many studies (2). As tumors manifest high levels of gelatinase activity, inhibitors specific for the gelatinases are highly sought.
In the past 8 years, there have been numerous approaches aimed at targeting MMP activities in tumors, and several clinical trials were carried out to test the efficacy of various inhibitors. Unfortunately, the results of these trials were disappointing due to the lack of an objective clinical response and undesired side effects. Many reasons have been postulated for these effects, but at the core of the problem remains the issue of inhibitor selectivity (3,4). Indeed, virtually all MMP inhibitors tested so far have been broad-spectrum inhibitors, designed around chelation of the active site zinc ion (5), and their spectrum of inhibition includes, in addition to MMPs, other metalloenzymes. Because targeting gelatinases remains of great promise in cancer therapy (6), efforts aimed at developing better and selective gelatinase inhibitors continue.
A mere handful of selective inhibitors for MMPs have been reported in the literature (for a review, see Ref. 7). We previously described the design and properties of inhibitor 1 (Fig. 1), which is a selective mechanism-based inhibitor for gelatinases. This compound binds to the active sites of MMP-2 and MMP-9, with the thiirane moiety coordinating with the zinc ion. This coordination to the active site metal ion activates the thiirane ring for opening by the nucleophilic attack of the active site glutamate in these enzymes (Fig. 2). A unique property of this inhibitor is that on binding to the active site zinc ion, a pattern of slow binding for inhibition sets in, leading to a rapid process for the onset of inhibition with an attendant slow process for recovery from slow binding at the noncovalent stage of inhibition. This noncovalent inhibited species leads to covalent inhibition by modification of the glutamate.
Whereas inhibitor 1, the prototype of this type of novel mechanismbased inhibitor for gelatinases, is showing promise in mouse models for diseases involving gelatinases (8,9), the poor solubility of this inhibitor in aqueous medium is a limitation of the molecule, which we have attempted to remedy. Furthermore, we have been interested in exploiting the concept behind the inhibitor design in targeting other MMPs. A computational model of the inhibitor bound in the active site of MMP-2 within the constraints of the data from x-ray absorption spectroscopy has been generated (10) (Fig. 2B). This model for inhibition of inhibitor 1 led the way in exploration of the next generation of this type of MMP inhibitor. The possibility for specific electrostatic interactions near the terminal phenyl group in inhibitor 1 bound to the active site of MMP-2 was anticipated for judiciously designed chemical functionalities into the molecular template of compound 1. We have introduced three functional groups, the methylsulfonamide (compounds 2 and 3), the nitro (compounds 4 and 5), and the acetamide (compounds 6 and 7), at the terminal phenyl ring system to exploit these electrostatic interactions. It was expected that these molecules would improve the solubility in aque-ous solutions while exhibiting high selectivity in inhibition toward gelatinases and the membrane-anchored MMP, MT1-MMP (MMP- 14), which all share a deep S1Ј binding site. As will be described herein, these expectations have been borne out, making these inhibitors valuable tools in studies of the functions of MMPs in disease processes. Furthermore, we have also prepared the oxirane variants of these molecules (compounds 2, 4, and 6). The fact that the oxirane variants are either poor inhibitors or demonstrate no observable inhibitory properties toward MMPs underscores the importance of the thiirane group for this inhibitor class.

EXPERIMENTAL PROCEDURES
Synthesis-1 H and 13 C NMR spectra were recorded on either a Varian UnityPlus 300-MHz or a Varian INOVA 500-MHz spectrometer. Chemical shifts are reported in ppm from tetramethylsilane on the ␦ scale. Mass spectra were recorded on JEOL JMS-AX505HA and Finnigan-MAT 8430 high resolution magnetic sector mass spectrometers. For silica gel column chromatography, EMD Silica gel 60 was employed. Thin layer chromatography was performed with Whatman 0.25-mm silica gel 60-F plates. All other reagents were purchased from Aldrich, Lancaster, or Across Organics.
4-(Allylthio)phenol-To a stirred solution of 4-hydroxythiophenol (4.30 g, 34.1 mmol) in N,N-dimethylformamide (25 ml) were added K 2 CO 3 (4.71 g, 34.1 mmol) and allyl bromide (3.09 ml, 34.1 mmol) at ice water temperature, and the mixture was stirred for 15 min, prior to stirring overnight at room temperature. After the addition of 1 M aqueous HCl, the mixture was extracted with ether (3ϫ). The combined organic layer was washed with water and brine, dried over MgSO 4 , and concentrated under reduced pressure. The resultant residue was purified by silica gel column chromatography (ethyl acetate/hexane, 1:10 to 1:6) to give 9 (5.74 g, 70%) as a white semisolid. The 1 H and 13 C NMR spectra and mass spectrum were identical to the reported values (11).
1-Allylthio-4-(4-nitrophenoxy)benzene-To a stirred solution of 9 (3.46 g, 20.8 mmol) in N,N-dimethylformamide (100 ml) were added cesium carbonate (10.2 g, 31.2 mmol) and 1-fluoro-4-nitrobenzene (10) (2.94 g, 20.8 mmol) at room temperature, and the mixture was stirred at the same temperature for 2 days. After dilution with water, the mixture was extracted into hexane (3ϫ). The combined organic layer was washed with water and brine, dried over Na 2 SO 4 , and concentrated under reduced pressure to give 11 (5.32 g, 89%) as a pale yellow oil. 1 (22 ml) were added acetic acid (2.54 ml, 44.2 mmol) and zinc powder (5.80 g, 88.4 mmol) at room temperature, and the suspension was stirred for 30 min (an exothermic reaction). After dilution with ethyl acetate, the mixture was filtered through Celite. The filtrate was washed with saturated NaHCO 3 and brine, dried over Na 2 SO 4 , and concentrated under reduced pressure to give a crude 12 (577 mg) as an orange oil, which was employed in the next reaction without purification.
To a stirred solution of 12 (577 mg) in CH 2 Cl 2 (10 ml) were added pyridine (894 l, 11.1 mmol) and methanesulfonyl chloride (205 l, 2.65 mmol) at ice-water temperature. After 15 min, the mixture was warmed to room temperature, and the stirring was continued for an additional 2 h. Subsequent to the addition of saturated NaHCO 3 , the mixture was extracted with ethyl acetate (3ϫ). The combined organic layer was washed with 1 M aqueous HCl, saturated NaHCO 3 solution, and brine, dried over Na 2 SO 4 , and concentrated under reduced pressure. The resultant residue was purified by silica gel column chromatography (CH 2 Cl 2 ) to give 13 (662 mg, 89% from 11) as a pale red solid. 4-(4-Acetamidophenoxy)-1-allylthiobenzene-To a stirred solution of 12 (794 mg), which was prepared from 11 (830 mg, 2.89 mmol) in the same manner as described for compound 13, in CH 2 Cl 2 (15 ml) were added pyridine (500 l, 6.18 mmol) and acetic anhydride (292 l, 3.09 mmol) at ice-water temperature, and the mixture was stirred at the same temperature for 1 h. Subsequent to the addition of saturated NaHCO 3 , the mixture was extracted with ethyl acetate (3ϫ). The combined organic layer was washed with 1 M aqueous HCl, saturated NaHCO 3 solution and brine, dried over Na 2 SO 4 , and concentrated under reduced pressure. The resultant residue was purified by silica gel column chromatography (ethyl acetate/CH 2 Cl 2 ϭ 1/8) to give 14 (782 mg, 99% from 11) as a white solid. 1   added mCPBA (4.2 g, 17.05 mmol) at ice-water temperature, and the mixture was stirred at room temperature for 9 days. With ice cooling, the reaction was quenched with saturated Na 2 S 2 O 3 and saturated NaHCO 3 solutions, and the mixture was extracted with ethyl acetate (3ϫ). The combined organic layer was washed with saturated Na 2 S 2 O 3 solution, saturated NaHCO 3 solution, water, and brine, dried over Na 2 SO 4 , and concentrated under reduced pressure. The resultant residue was purified by silica gel column chromatography (ethyl acetate/ hexane ϭ 3/2) to give 2 (386 mg, 62%) as a white solid. 1  [4 -(4 -Nitrophenoxy)phenylsulfonyl]methyloxirane-This material was prepared in the same manner as described for 2, with the exception that 11 was used in place of 13. The crude material was purified by silica gel column chromatography (ethyl acetate/hexane, 2/3) to give 4 (56%) as a pale yellow solid. 1  [4-(4-Acetamidophenoxy)phenylsulfonyl]methyloxirane-This material was prepared in the same manner as described for 2, with the exception that 14 was used in place of 13. The crude material was purified by silica gel column chromatography (ethyl acetate/hexane ϭ 3/1) to give 6 (34%) as a white semi-solid. 1  {4- phenoxy]phenylsulfonyl}methylthiirane-To a stirred solution of 2 (82 mg, 0.21 mmol) in MeOH-tetrahydrofuran (3:1, 2 ml) was added thiourea (41 mg, 0.53 mmol) at room temperature, and the mixture was stirred overnight at the same temperature. After concentration under reduced pressure, the residue was dissolved into ethyl acetate. The ethyl acetate solution was washed with water and brine, dried over Na 2 SO 4 , and concentrated under reduced pressure. The residue was purified by silica gel column chromatography (ethyl acetate/hexane, 1/1) to give 3 (67 mg, 77%) as a white solid. 1  [4-(4-Nitrophenoxy)phenylsulfonyl]methylthiirane-This material was prepared in the same manner as described for 3, with the exception that 4 was used in place of 2. The crude material was purified by silica gel column chromatography (ethyl acetate/hexane, 1/3) to give 5 (79%) as a pale yellow solid. 1  [4-(4-Acetamidophenoxy)phenylsulfonyl]methylthiirane-This material was prepared in the same manner as described for 3, with the exception that 6 was used in place of 2. The crude material was purified by silica gel column chromatography (ethyl acetate/hexane, 3/2 to 2/1) to give 7 (76%) as a white solid. 1  Assessment of Inhibitor Solubility-Aliquots (10 l) of the solutions of the thiirane compounds 1, 3, 5, and 7 in Me 2 SO (e.g. 10 mM, 12 mM, 14 mM and higher concentrations) were added to 990 l of buffer R (50 mM HEPES (pH 7.5), 0.15 M NaCl, 5 mM CaCl 2 , 0.01% Brij-35, 1% Me 2 SO) at 37°C. Each mixture was inspected for clarity (or turbidity) to calculate the approximate upper limit of solubility.
Enzymatic Activity Assays-Enzymatic activity was monitored with synthetic peptide, fluorescence-quenched substrates from Peptides International, Inc. (Louisville, KY). The activities of MMP-2, MMP-9, MMP-7, and MMP-14 were monitored with MOCAcPLGLA 2 pr(Dnp)AR-NH 2 at excitation and emission wavelengths of 328 and 393 nm, respectively, in buffer R. MOCAcRPKPVE norvalyl (Nva)WRK(Dnp)NH 2 was the fluorogenic substrate used to measure MMP-3 at 325 and 393 nm in buffer R. MMP-1 was assayed with (Dnp)P cyclohexylanalyl (Cha)GC(Me)-HAKN ⑀ -methylanthranoyl (NMa)NH 2 at 340 and 440 nm, in a buffer consisting of 50 mM Tris (pH 7.6), 200 mM NaCl, 5 mM CaCl 2 , 20 mM ZnSO 4 , 0.05% Brij-35. Less than 10% substrate hydrolysis was monitored (12). Fluorescence was measured using a Photon Technology International spectrofluorometer, equipped with RadioMaster TM and FeliX TM hardware and software, respectively. The excitation and emission band passes were 1 and 3 nm, respectively. An integration time of 4 s was used for data acquisition. The assays were carried out at 25°C, and the cuvette holder was kept at the same temperature. Quartz or disposable acrylic micro-or semimicrocuvettes from Sarstedt (Newton, NC) and Perfector Scientific (Atascadero, CA), respectively, were used.
Enzyme Inhibition Studies-Slow binding enzyme inhibition was monitored continuously for 20 -60 min by adding the enzyme (0.5-1 nM) to a solution of buffer R containing the appropriate fluorogenic substrate and increasing concentrations of the inhibitor (final volume 2 ml) in acrylic cuvettes with stirring. The progress curves were nonlinear least squares-fitted to Equation 1 (13), where v o represents the initial rate, v s is the steady state rate, k is the apparent first order rate constant characterizing the formation of the steady-state enzyme-inhibitor complex, and F o is the initial fluorescence, using the program Scientist (MicroMath Scientific Software, Salt Lake City, UT). Association and dissociation rate constants (k on and k off , respectively) were obtained from the slope and intercept, respectively, of plots of the apparent first order rate constant k versus the inhibitor concentration according to Equation 2, describing a one-step association mechanism (Scheme 1), where S is the fluorogenic peptide substrate used, and the EI* is the product of slow binding inhibition. The expression for k on includes the requisite conformational change necessary for the formation of EI*. The K m values used for the reaction of MMP-2, MMP-9, and MMP-14 with the fluorogenic substrate MOCAcPLGLA 2 pr(Dnp)AR-NH 2 were 2.46 Ϯ 0.34, 3.06 Ϯ 0.74 (14), and 6.9 Ϯ 0.6 M (15), respectively. The inhibition constant, K i , was given by k off /k on . Alternatively, K i values were obtained by plot- For analysis of simple linear competitive inhibition, reaction mixtures containing the enzyme (ϳ1 nM) and increasing concentrations of the inhibitor, in buffer R (final volume 1 ml), were incubated for ϳ16 h at 25°C in acrylic semimicrocuvettes. The remaining enzymatic activity was measured with the appropriate synthetic peptide fluorogenic substrate for 5-10 min. The initial velocities for the reaction of the enzyme with the substrate were determined by linear regression analysis of the fluorescence versus time traces using FeliX TM . These initial rates were fitted to Equation 4 (16), where v i and v o represent the initial velocity in the presence and absence of inhibitor, respectively, using the program Scientist.
Equilibrium Dialysis-Mixtures of enzyme (10 nM) in the presence and absence of inhibitor (1 mM) were incubated at room temperature   OCTOBER 7, 2005 • VOLUME 280 • NUMBER 40 for ϳ3 h. The remaining enzyme activity was measured with the appropriate fluorogenic substrate, as described above. Part of the reaction mixture (ϳ150 l) was dialyzed in either dialysis tubing (Invitrogen) or in a 0.1-0.5-ml capacity Slide-A-Lyzer dialysis cassette (Pierce), against buffer R (3 ϫ 1 liter) containing no Me 2 SO, at room temperature, for Ͼ4-h periods prior to change of buffer to allow for equilibration, over a 48-h period. The remainder of the inhibition mixture was left on a rotator, at room temperature, over the same period of time. Both the dialyzed and nondialyzed solutions were tested for MMP activity using the proper fluorogenic substrate. Enzyme activity was expressed as a percentage relative to that in the absence of inhibitor.
Recombinant Proteins, Enzymes, and Inhibitors-Human recombinant pro-MMP-2 and pro-MMP-9, TIMP-1, and TIMP-2 were expressed in HeLa S3 cells infected with the corresponding recombinant vaccinia viruses and purified to homogeneity from the media as previously described (14). Pro-MMP-2 and pro-MMP-9 were activated by incubation with 1 mM p-aminophenylmercuric acetate (APMA) for ϳ2 h at 37°C as previously described (14). APMA was dialyzed out in collagenase buffer (50 mM Tris-HCl (pH 7.5), 5 mM CaCl 2 , 150 mM NaCl, and 0.02% Brij 35). Human recombinant active MMP-1 and MMP-7 were from R&D (Minneapolis, MN) and Chemicon International (Temecula, CA), respectively, and the recombinant catalytic domains of human MMP-3 and MMP-14 were from EMD Biosciences (La Jolla, CA). Active enzyme concentration was determined by active site titration with solutions of either TIMP-1 or TIMP-2 with known concentration. The hydroxamate-based MMP inhibitor BB-94 was synthesized in the Mobashery laboratory, and GM6001 was purchased from Chemicon. Stock solutions of BB-94, GM6001, and compounds 2-7 were prepared in Me 2 SO in the millimolar concentration range.
Pro-MMP-2 Activation on Cells-Confluent BS-C-1 cells in 12-well plates were co-infected with vTF7-3 and vT7-MT1 viruses for 45 min in infection medium (DMEM supplemented with 2.5% fetal bovine serum and antibiotics), at 37°C, as described (18). The infection medium was removed, and the cells were incubated overnight with serum-free DMEM supplemented with L-glutamine and antibiotics containing increasing concentrations (0 -5 M) of the synthetic MMP inhibitors. The cells were washed twice with phosphate-buffered saline and incubated for 6 h with serum-free DMEM containing pro-MMP-2 (10 nM). The cells were rinsed twice with cold phosphate-buffered saline and lysed with cold lysis buffer (25 mM Tris-HCl (pH 7.5), 1% IGEPAL CA-630, 100 mM NaCl) containing protease inhibitors (one pellet of Complete Mini, EDTA-free protease inhibitor mixture from Roche Applied Science/10 ml of buffer). The lysates were then subjected to gelatin zymography to monitor pro-MMP-2 activation and to immunoblot analysis to detect MT1-MMP expression and processing.
Gelatin Zymography and Immunoblot Analysis-Gelatin zymography was performed using 10% Tris/glycine SDS-polyacrylamide gels, containing 0.1% gelatin, as previously described (19). The samples for immunoblot analysis were subjected to reducing SDS-PAGE followed by transfer to nitrocellulose membranes. MT1-MMP was probed with rabbit polyclonal antibody 815 to MT1-MMP from Chemicon.

Kinetic parameters for MMP inhibition by compounds 3, 5, and 7
The enzymes (0.5-1 nm) were added to a solution of the proper synthetic fluorogenic substrate and increasing concentrations of the inhibitor in buffer R. Substrate hydrolysis was monitored for up to 60 min. The kinetic parameters were evaluated as described under "Experimental Procedures." Enzyme Migration and Invasion Assays-For migration assays, HT1080 cells were cultured in 6-well plates in complete medium until they reached confluence. Prior to the migration assay, the cells were treated with serum-free medium containing mitomycin C (25 g/ml,), in the presence and absence of concanavalin A (25 g/ml, 30 min) to induce pro-MMP-2 activation (20). Scratch wounds were then carefully made in the confluent monolayer using a disposable plastic pipette tip. After gentle rinsing twice with phosphate-buffered saline to remove detached cells, serum-free media containing increasing concentrations of inhibitor 3 were added, and the cells were incubated at 37°C for various times. Photographs were taken using an Olympus model DF 12-2 camera connected to a Nikon TMS-F microscope at ϫ10 magnification, at the indicated time points. The extent of wound closure in the presence or absence of inhibitor was quantified by measuring the width of the wound with a ruler using an amplified PowerPoint figure.
Tumor cell invasion was carried out in 8-m pore Transwell inserts (BD Biosciences) coated with 50 g of Matrigel/filter. HT1080 cells suspended in serum-free DMEM containing 0.1% bovine serum albumin and various doses of inhibitor 3 (0.1-10 M) or 1% Me 2 SO (vehicle) were seeded (2 ϫ 10 5 cells/insert) on the Matrigel-coated inserts. The lower compartment was filled with DMEM containing 5% fetal bovine serum. After an 18-h incubation at 37°C in a humidified atmosphere with 5% CO 2 , the upper surface of the membrane in each insert was wiped off with a cotton swab to remove all of the noninvading cells. The cells that migrated to the lower side of the Matrigel-coated filter were fixed and stained with Diff-Quik (Dade Behring Inc., Newark, DE) and counted under a microscope in three different fields. Each treatment was assayed in quadruplicate.
Chemosensitivity Assay-Cell viability after exposure of the cells to the inhibitors was assessed by 2-(4-iodophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium, monosodium salt (WST-1, Roche Applied Science) staining. Briefly, HT1080 cells (2 ϫ 10 4 cells/well) were seeded in 96-well culture plates in complete medium. After overnight culture, the medium was replaced with serum-and phenol-free media containing 0.1% bovine serum albumin and supplemented with or without inhibitor 3 (0 -10 M final concentrations). Control medium was supplemented with 1% Me 2 SO. After an 18-h incubation, WST-1 (10 l/well) was added, and the difference in absorbance at 450 and 655 nm (reference filter) was measured using a Bio-Rad Benchmark microplate reader. Data were collected using the Microplate Manager software. The absorbance of blank wells containing control medium but no cells (typically Ͻ5%) was subtracted. Each treatment was assayed in quadruplicate.

RESULTS AND DISCUSSION
Design and Synthesis of MMP Inhibitors-The computational model for binding of inhibitor 1 to the active site of MMP-2 is shown in Fig. 2B. The site for substitution at the para position of the terminal phenyl ring of the inhibitor is indicated by an arrow. Fig. 3 shows the synthetic route for oxiranes 2, 4, and 6, and thiiranes 3, 5, and 7. Chemoselective allylation of commercially available 4-hydroxythiophenol (8) provided phenol 9 (11) in 70% yield, which was subsequently coupled with 1-fluoro-4-nitrobenzene (10) in the presence of cesium carbonate to afford the diphenyl ether 11 in 89% yield. The nitro group of 11 was reduced over elemental zinc, and the resulting amine 12 was treated with methanesulfonyl chloride or acetic anhydride to give the corresponding amides 13 and 14, respectively, in high yields. Oxidation of 13, 11, and 14 to their corresponding oxiranes 2, 4, and 6 required an excess of mCPBA (10 -12 eq.) and long reaction times (9 -10 days), due to the low reactivity of the olefin moieties; the isolated yields of the oxiranes were moderate (34 -62%). Finally, the conversion of compounds 2, 4, and 6 to the corresponding thiiranes 3, 5, and 7 was accomplished by the treatment with thiourea in good yields. Compounds 3 and 7 had improved solubility in aqueous solution compared with the prototypic inhibitor 1. Solubility was investigated in 50 mM HEPES (pH 7.5), 0.15 M NaCl, 5 mM CaCl 2 , 0.01% Brij 35, 1% Me 2 SO, which was the buffer that we used for all of the kinetic experiments (see below). The maximal solubility for the samples was as follows: 1, 80 M; 3, 340 M; 5, 60 M; 7, 540 M.
Enzyme Inhibition Kinetics-The mechanism of action of this class of MMP inhibitors stipulates that the thiirane sulfur would coordinate with the active site zinc ion (Fig. 2). Consistent with this expectation, the three synthetic thiiranes of this study (compounds 3, 5, and 7) are excellent inhibitors of gelatinases (and of MT1-MMP in the case of inhibitor 3) (TABLE ONE), whereas the corresponding oxiranes (compounds 2, 4, and 6) are either poor inhibitors or not inhibitory at all toward all of the tested MMPs (TABLE TWO).
In slow binding inhibition, upon binding of the inhibitor to the enzyme, the complex undergoes a requisite conformational change that is not readily predisposed for the reversal of the inhibition (21)(22)(23). The slow binding inhibitor shows a unique profile for the onset of inhibition that is discerned by nonlinear progress curves. We see slow binding behavior for inhibitor 3 (with MMP-2, MMP-9, and MMP-14), for inhibitor 5 (only with MMP-2), and for inhibitor 7 (only with MMP-2) (Fig. 4).
The second order rate constants for the onset of slow binding inhibition (k on ) are rapid (10 2 to 10 4 M Ϫ1 s Ϫ1 ), and the first order rate constants for the reversal of the process from the noncovalent enzyme-inhibitor complexes (k off ) are slow (10 Ϫ4 to 10 Ϫ3 s Ϫ1 ; e.g. the t1 ⁄2 values for reversal for inhibitor 3 with MMP-2 and MMP-9 are 34 and 13 min, respectively). The dissociation constants for the nonco-

Kinetic parameters for MMP inhibition by compounds 2, 4, and 6
The enzymes (0.5-1 nm) were incubated with increasing concentrations of inhibitor in buffer R. The remaining activity was measured with the appropriate synthetic fluorogenic substrate. The kinetic parameters for rapid, competitive inhibition were evaluated as described under "Experimental Procedures.   itors 1, 3, 5, and 7 as mechanism-based inhibitors are different from one another, despite the similar structural template for the class. Briefly, inhibitor 1 can inhibit both MMP-2 and -9 (the gelatinases), inhibitor 3 inhibits the gelatinases plus MMP-14, and most interestingly, inhibitors 5 and 7 are mechanism-based inhibitors only for MMP-2. Furthermore, these are nanomolar inhibitors for their targeted enzymes and exhibit comparable values for the k on and k off parameters for the slow binding components of their kinetics. Covalent Versus Noncovalent Inhibition of MMPs-The thiirane class of MMP inhibitors was designed to be covalent enzyme inhibitors. Upon formation of the noncovalent enzyme-inhibitor complex, the ubiquitous active site glutamates of MMPs (Glu 404 for MMP-2, for example) were expected to be covalently modified by the inhibitor with the requisite thiirane ring opening (Fig. 2). The kinetics of inhibition indicate two components, a noncovalent stage (slow binding) and a subsequent stage that may be attributed to the covalent modification of the active site glutamate, as will be outlined.

Mechanism-based Inhibition of MMPs
The covalent component of inhibition results in modification of the glutamate as an ester on its side chain carboxylate. The earlier x-ray absorption spectroscopy analysis with inhibitor 1 (10) had provided evidence for the covalent bond formation, in that upon the onset of inhibition, the method revealed the formation of a thiolate from the thiirane of the inhibitor (ring opening), coordinated to the active site zinc ion.
Whereas a slow binding step need not necessarily be a prerequisite for covalent chemistry, both the mechanism-based process leading to covalent enzyme modification and the slow binding behavior produce time dependence for the loss of activity seen with these inhibitors (Fig. 4). Our experience with inhibitor 1 had shown that slow binding led to covalent chemistry, with a longevity for the final inhibited species substantially exceeding the duration that would have been anticipated from 4 times the t1 ⁄ 2 for recovery of activity from the slow binding component of inhibition (in other words, four half-lives leading to an anticipated 94% recovered of activity due to the noncovalent component). This is the case with inhibitors 3, 5, and 7 as well. The slowest t1 ⁄ 2 calculated for recovery of activity from the noncovalent slow binding species for the best inhibition (compound 3 with MMP-2) is 34 min. Yet, a mere 1% of activity recovery was seen for MMP-2 inhibited by inhibitor 3 after 48 h of dialysis. Four half-lives for recovery from inhibition (94% anticipated recovered activity) with this inhibitor and MMP-2 should be achieved in just under 2.5 h (136 min) were it merely the slow binding event that accounted for MMP-2 inhibition. This is clearly not the case, and the inhib-

Mechanism-based Inhibition of MMPs
ited enzyme is more stable than the k off (from which t1 ⁄ 2 is evaluated) indicates. The results of dialyses for inhibitors 3, 5, and 7 are given in Fig. 5.
Having documented above that mere slow binding behavior cannot be responsible for the complete inhibition that we see, we need to explain why any recovery of activity should be seen if we deal with covalent chemistry. The answer is that the stability of covalent bonds is relative. Esters are among the least stable covalent bonds in aqueous solution (24). This bond would undergo hydrolysis, resulting in recovery of activity. The process accelerates when there is a more significant exposure of the ester bond to water, conditions that can arise when the protein is denatured.
Matrix-assisted laser desorption time-of-flight mass spectrometry analysis, performed on an Applied Biosystems Voyager-DE STR (Framingham, MA) instrument at the Harvard Microchemistry and Proteomics Analysis Facility (Cambridge, MA), was attempted on samples containing MMP-2 (10 M) in the presence and absence of inhibitor 3 to detect a shift in molecular mass consistent with a complex of active MMP-2 (ϳ62 kDa) with the inhibitor. However, after several attempts with different conditions, we failed to detect a 400-Da addition in molecular mass to the 62-kDa peak. The difficulty is that at this high end of mass detection, the signals are broadened, and the identification of the small incremental increase due to the mass of the inhibitor was not possible within the resolution of the instrument.
Effect of Gelatinase Inhibitors on Pro-MMP-2 Activation by MT1-MMP-MT1-MMP has been identified as the physiological activator of pro-MMP-2 (25). This reaction is regulated at multiple levels, and its rate is significantly enhanced by TIMP-2, which, by binding active MT1-MMP, acts as a "receptor" for pro-MMP-2 on the cell surface (25,26). The binding of pro-MMP-2 to the MT1-MMP⅐TIMP-2 complex, facilitates the first pro-MMP-2 cleavage by a neighboring TIMP-2-free MT1-MMP molecule (26). Pro-MMP-2 activation requires a second autolytic cleavage (27), leading to full activation. We have previously shown that broad spectrum synthetic MMP inhibitors (e.g. marimastat) enhance pro-MMP-2 activation by MT1-MMP in the presence of TIMP-2 (15), a process that appears to involve stabilization of mature MT1-MMP at the cell surface by the MMP inhibitor. This enhancing effect on pro-MMP-2 activation was not observed when the cells were exposed to inhibitor 1 (15), which exhibits lower affinity toward MT1-MMP, a feature of its selectivity for inhibition of gelatinases. Therefore, we proposed that nonspecific targeting of MT1-MMP by broad spectrum MMP inhibitors might, under certain conditions, elicit a counterproductive effect by enhancing the activity of the MT1-MMP/gelatinase A axis (28). Because inhibitor 3 is also selective for the gelatinases, we postulated that it might behave like inhibitor 1 in a cellular system of pro-MMP-2 activation by MT1-MMP in the presence of TIMP-2. To this end, BS-C-1 cells, which express low levels of endogenous TIMP-2, were infected to express MT1-MMP and incubated with pro-MMP-2 in the presence of either GM6001, a broad spectrum MMP inhibitor, or inhibitor 3, as described (15). Pro-MMP-2 activation was followed by gelatin zymography. As shown in Fig. 6A, exposure of the MT1-MMPexpressing cells to as little as 40 nM GM6001 induced pro-MMP-2 activation, as determined by the appearance of the active form. Higher inhibitor concentrations further enhanced pro-MMP-2 activation, under these conditions. Of note, this enhancing effect of broad spectrum MMP inhibitors such as GM6001 requires the endogenous TIMP-2, as we have previously shown (15). Consistently, GM6001 caused a dose-dependent accumulation of active MT1-MMP (57 kDa) (Fig. 6B). In contrast, when the cells were incubated with inhibitor 3 (up to 4 M), pro-MMP-2 activation was not observed. Also, the accumulation of active MT1-MMP was not observed with inhibitor 3, consist-ent with its reduced affinity for this protease when compared with MMP-2 (TABLE ONE). Although inhibitor 3 is also a mechanismbased inhibitor for MT1-MMP, its lower affinity relative to MMP-2 is likely to preclude this inhibitor influencing pro-MMP-2 activation under these conditions. It is also possible that covalent inhibition of MT1-MMP, as opposed to a reversible inhibition, alters the availability of the active site of MT1-MMP for TIMP-2 binding, a prerequisite for pro-MMP-2 activation (15). Although more studies are required, these results suggest that the concept behind inhibitor 3 is a promising framework from which to further develop more effective and selective MT1-MMP inhibitors, a key protease in tumor cell invasion. Nevertheless, these studies further demonstrate the selectivity of inhibitor 3 in a live cellular system and lend credit to the hypothesis that selectivity, rather than affinity, may be key to the successful therapeutic application of synthetic MMP inhibitors.
Inhibitor 3 Inhibits HT1080 Cell Migration and Invasion-It is well established that tumor cell migration and invasion depend on gelatinase activity. Therefore, we wished to evaluate the effect of inhibitor 3, which is selective for the gelatinases, on the migration and invasion of HT1080 cells as described under "Experimental Procedures." Cell migration was monitored under conditions of pro-MMP-2 activation, which was achieved by concanavalin A treatment, and inhibition of cell proliferation. As shown in Fig. 7, A and B, exposure of HT1080 cells to various doses (0 -20 M) of inhibitor 3 significantly inhibited (80% at 2 M) their migration in a scratch wound assay when compared with untreated cells. Likewise, the ability of HT1080 cells to invade Matrigel-coated filters was significantly reduced by inhibitor 3, and as little as 100 nM of inhibitor caused Ͼ25% inhibition of HT1080 cell invasion (Fig. 7C). These effects of inhibitor 3 could not be ascribed to cytotoxicity, since no evidence of cell toxicity was detected when HT1080 cells were exposed to inhibitor 3 up to concentrations of 10 M, as determined using the WST-1 chemosensitivity assay (data not shown). Given the high selectivity exhibited by this compound toward MMP-2 relative to other MMPs (TABLE ONE), the slower migration in the presence of 200 nM inhibitor 3, a concentration too low to inhibit other MMPs, including MT1-MMP, suggests that the observed effect was most likely due to MMP-2 inhibition. These results further demonstrate the ability of inhibitor 3 to act as a selective gelatinase inhibitor in cellular systems and to alter MMP-dependent processes. The new characteristics of inhibitor 3 and its high selectivity make this inhibitor an excellent candidate for future in vivo testing in relevant human disease models in mice.
The thiirane class of mechanism-based inhibitors was conceived, designed, and prepared by us for the first time in our pursuit of selectivity in inhibition of MMPs of importance to several disease processes. We have revealed in the present report that inhibitor 3 targets MMP-2, -9, and -14, whereas inhibitors 5 and 7 are inhibitory only toward FIGURE 6. GM6001, but not inhibitor 3, enhances MT1-MMP-dependent pro-MMP-2 activation by BS-C-1 cells. BS-C-1 cells, co-infected to express MT1-MMP, as described under "Experimental Procedures," were incubated for 16 h with serum-free DMEM containing the indicated inhibitor concentrations. After rinsing, the cells were incubated for 6 h with serum-free DMEM supplemented with recombinant pro-MMP-2 (10 nM). The cells were lysed, and the lysates were subjected to zymographic (A) and immunoblot analysis (B) using the anti-MT1-MMP polyclonal antibody 815. L, I and A represent the latent, intermediate and active forms of MMP-2, respectively, and the 60-and 57-kDa forms represent pro-and active MT1-MMP, respectively. FIGURE 7. Inhibition of HT1080 cell motility and invasion by inhibitor 3. A and B, confluent cultures of HT1080 cells in 6-well plates were treated with mitomycin C (25 g/ml) in serum-free DMEM for 30 min. Scratch wounds were made on the monolayers, and the wounded cultures were then incubated with serum-free DMEM supplemented without or with various amounts of inhibitor 3 (0 -20 M) for up to 20 h. At each time period, the cultures were photographed (A), and the width of the scratch wound was measured as described under "Experimental Procedures." C, HT1080 cells were seeded in 8-m pore Transwell filters coated with Matrigel (50 g/filter) in the presence or absence of inhibitor 3 (0.1-10 M). The number of cells that invaded to the lower side of the filter was counted in three representative fields. Each value represents the mean Ϯ S.E. of four independent determinations. Asterisk, p Ͻ 0.05; double asterisk, p Ͻ 0.001 when tested against the control using Tukey-Kramer multiple comparisons test (p ϭ 0.004 by analysis of variance).