|Home | About | Journals | Submit | Contact Us | Français|
(±)-2-[(4-Phenoxyphenylsulfonyl)methyl]thiirane 1 is a potent and selective mechanism-based inhibitor of the gelatinase sub-class of the zinc-dependent matrix metalloproteinase (MMP) family. Inhibitor 1 has excellent activity in in vivo models of gelatinase-dependent disease. We demonstrate that the mechanism of inhibition is a rate-limiting gelatinase-catalyzed thiolate generation via deprotonation adjacent to the thiirane, with concomitant thiirane opening. A corollary to this mechanism is the prediction that thiol-containing structures, related to thiirane–opened 1, will possess potent MMP inhibitory activity. This prediction was validated by the synthesis of the product of this enzyme-catalyzed reaction on 1, which exhibited a remarkable Ki of 530 pM against MMP-2. Thiirane 1 acts as a caged thiol, unmasked selectively in the active sites of gelatinases. This mechanism is unprecedented in the substantial literature on inhibition of zinc-dependent hydrolases.
The central position of the epoxide ring in organic synthesis derives from the ease of its synthesis, and the ability of Brønsted or Lewis acids to control its opening by nucleophiles. Thiiranes, three-membered rings containing a sulfur atom, are typically less reactive than epoxides. Due to their latent reactivity toward nucleophiles, the aziridines, epoxides and thiiranes all have been used as irreversible enzyme inhibitors (1–3). In the pioneering work of Kim et al., epoxybutanoic acid covalently modified the carboxylate of the glutamate in the active site of the zinc protease, carboxypeptidase A (4). The catalytic zinc ion of this protease, which activates the scissile amide bond of the substrate during normal turnover, here functions as a Lewis acid for epoxide O-alkylation of this glutamate (5). The conceptual extension of this principle to the creation of efficacious matrix metalloproteinase inhibitors, zinc proteases involved in the pathophysiology of inter alia human inflammation and tumor metastasis, is an objective of this laboratory (6). We previously reported the discovery, the core SAR, and initial computational and experimental mechanistic studies with the thiirane-containing structure 1 (7–12). Thiirane 1 (also known as SB-3CT) exerts potent (nanomolar) and time-dependent inhibitory activity with high selectivity toward the gelatinase MMP sub-class (13). The selectivity of 1 for the gelatinases (MMP-2 and MMP-9) has provided experimental evidence for gelatinase activation in cell culture models of breast cancer metastasis (14–16), axon guidance (17), beta-adrenergic receptor stimulated apoptosis (18,19) and amyloid precursor protein processing (20). Compound 1 is active in an animal model of testosterone-induced neurogenesis (21) as well as in several rodent models of human disease, including laminin degradation following cerebral ischemia (22), prostate cancer metastasis to the bone (23), breast cancer metastasis to the lungs (24), blood-brain barrier integrity (25), and T-cell lymphoma metastasis to the liver (26). In contrast, the lack of efficacy of 1 in cell culture models of ovarian cancer cell metastasis implicates the activity of other MMPs during collagen degradation in this cancer (27). The favorable biological outcome in MMP-dependent disease models using 1 as a gelatinase MMP sub-class inhibitor stimulated this further mechanistic study of MMP-2 inhibition by 1.
The design of inhibitor 1 was based on the precedent of Kim’s oxirane inhibitors of carboxypeptidase A. Kim documented by X-ray crystallography that interactions of the oxirane with the zinc ion led to covalent modification of the active site glutamate common to all zinc proteases (28). This same chemistry appeared plausible for gelatinases (29). This conception led to a computer-aided design effort, resulting in the synthesis of over 100 compounds, and culminating in the discovery of 1. Compound 1 proved to be a potent inhibitor of gelatinases having time-dependent kinetics for this inhibition, a hallmark of both covalent inhibition as well as slow-binding inhibition of enzymes. In contrast, the epoxide analogue of 1 was a much weaker inhibitor, and its inhibition lacked time-dependence (7). This difference emphasized the special ability of 1 to exploit the Lewis acidity of the active site zinc. Based on the Kim precedent, we favored the glutamate alkylation mechanism, notwithstanding the observation that the enzyme activity recovered gradually. This recovery was thought to reflect hydrolytic lability of the ester linkage between the glutamate and the inhibitor. Three separate crystallographic efforts to confirm this mechanism were without success.
However, in the course of further synthesis we noted that the thiirane ring of 1 was not stable in the presence of base. A retrospective look at the chemistry of zinc proteases alerted us to the observation by Sugimoto and Kaiser that Glu270 of carboxypeptidase A promoted enolization of a ketone (30). As stable thiol coordination of the MMP zinc is used to suppress protease activity in MMP pro-enzymes (that is, the inactive zymogen), the possibility that proper presentation of a thiol to the MMP catalytic zinc would coincide with potent inhibition of the enzyme was credible. Moreover, the thiol and thioether functional groups are key zinc-interacting functional groups in several well-studied MMP inhibitor classes (31–33). Hence, we were compelled to consider two limiting mechanisms for gelatinase inhibition by 1. Each uses the Lewis acidity of the active site zinc ion to activate the thiirane. The first mechanism is covalent O-alkylation of the cognate glutamate to the Glu270 of carboxypeptidase A, which is also present in the MMP active site (Mechanism A of Scheme 1). The second mechanism is base-catalysis— presumably by Glu404 or Glu404-activated water—removal of a proton from the relatively acidic carbon that is adjacent to the sulfone. This deprotonation initiates opening of the thiirane to an allylthiolate (Mechanism B of Scheme 1). In this mechanism, the thiirane ring acts as a caged thiol, unmasked only within the gelatinase active site. The experiments used to differentiate the two mechanisms are described.
Following incubation of MMP-2 with a 1000-fold excess of 1 (reaction of 10 nM MMP-2 with 10 µM 1 for 3 h in pH 7.5 buffer at ambient temperature), the catalytic activity of MMP-2 was 0% of the control (MMP-2 activity in the absence of 1). Two methods were used to remove the excess inhibitor from this inhibited MMP-2 enzyme. The first method was gel filtration. The inhibited enzyme was separated from excess inhibitor using a centrifugal gel filtration column, recovering 97% of the activitiy of the control (enzyme incubated without inhibitor and filtered separately). The activity recovered from the control was >95% of the total activity at the start of the incubation. This result shows that MMP-2 regains activity following separation from 1. The kinetic behavior of the recovered MMP-2 with respect to the assay substrate was identical to the enzyme from the no inhibitor control. This observation suggests that 1 left no residual modification of the enzyme. Mechanism A of Scheme 1 postulates covalent modification of the catalytically essential active site glutamate. If this mechanism were operative, full recovery of MMP-2 activity would be an improbable observation following separation of the inhibitor from the MMP-2 enzyme.
MMPs are synthesized in vivo as zymogens (34,35). The molecular basis for suppression of catalytic activity in the zymogen is the presence of a pro domain, containing a cysteine residue that engages the active site zinc ion as a thiolate ligand (36). Enzyme activity is suppressed by this stable zinc coordination (37) and by the body of the pro-domain, blocking substrate access to the active site (38,39). Transformation of the zymogen to the mature MMP occurs by proteolytic removal of the pro-domain. Efficient in vitro initiation of the proteolytic removal of the prodomain is accomplished with the organomercurial reagent APMA (40). This reagent traps the prodomain cysteine thiolate that coordinates to the active site zinc, thereby displacing the prodomain from the active site. The possibility that APMA would exert a similar effect on MMP-2 inhibited by 1—which is also characterized by direct zinc-thiolate coordination (8,38)—was evaluated as a second method. Inactivation of a 1 nM solution of MMP-2 using excess 1 (1.5 µM) for 4 h gave complete loss of enzyme activity. Addition of a solution of APMA (to 1 mM final concentration) restored 90% of the total enzyme activity that was initially present. Again, this result is improbable for a covalent mechanism of inhibition. In contrast, if the mechanism for loss of MMP-2 catalytic activity by 1 is formation of a stable zinc thiolate coordination complex, then restoration of activity by APMA (analogous to its role in MMP zymogen activation) is expected.
The possibility that a base-catalyzed deprotonation mechanism (Scheme 1, Pathway B) might better account for the ability of 1 to effect inhibition of MMP-2 catalytic activity prompted further experiment. In particular, we were curious as to its intrinsic reactivity. Accordingly, a solution of 1 in toluene in the presence of 1.1 equiv of 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU, pKa = 12) and iodomethane (10 equivalents) showed rapid and complete (within 30 min) transformation of 1 to the vinylsulfide 2, isolated as a 2:1 Z,E-diasteromeric mixture (Scheme 2). A solution of 1 in MeOH in the presence of a DBU buffer (1.1 equivalents of DBU, 1.1 equivalents of DBU·HCl) and iodomethane (10 equivalents) also showed complete loss of 1 within 3 h at rt. In contrast, the identity of the major product (80%) of this reaction was the S-methylated vinylsulfone 3, with the S-methylated vinylsulfides 2 (here also as the 2:1 Z,E-mixture) as minor products (20% total).
The solvolytic behavior of 1 parallels previous mechanistic study with structurally very closely related thiiranes and epoxides (41). In this study, Piras and Stirling demonstrated that (sulfonyl)methyl-substituted thiiranes (and epoxides) open by base-mediated deprotonation at the carbon adjacent to the sulfone, giving vinylsulfone products. Kinetic study indicated proton abstraction, alkene formation, and thiirane opening to coincide within a single transition state. Deuterium substitution adjacent to the sulfone gave a primary kinetic isotope effect of 3.2 for thiirane opening (41). Following vinylsulfone formation, further proton exchange equilibrates the vinylsulfone with its thermodynamically more stable vinylsulfide isomer 2 (42–44). The difference in product outcome between the conditions of DBU in toluene, and DBU/DBU·HCl in MeOH, may be understood in terms of the preservation of the basicity of DBU under the former reaction conditions. The kinetic vinylsulfone product is favored by the lower basicity of the DBU·HCl reaction conditions. Significantly, yet milder reaction conditions using a mixture of NaOAc and Zn(OAc)2 in MeOH as a simple mimetic of the conserved glutamate carboxylate and Zn(II) ion of the MMP active site, gave a similar outcome to the DBU·HCl reaction conditions. After one month at room temperature, the reaction mixture consisted of 40% unreacted 1 and ≥55% 3, with small amounts (≤5%, all yields by NMR) of the vinylsulfides 2.
Further evidence in support of this mechanism was obtained with a derivative of 1 having reduced carbon acidity. The acidity of alkylarylsulfoxides (approximate pKa = 33) is significantly weaker than that of the corresponding sulfone (approximate pKa = 29). Accordingly, it was anticipated that the sulfoxide analog of 1 would be more stable in solution. Indeed, sulfoxide 4 (prepared as a 1:1 diastereomeric mixture by adaptation of the routes used previously in this laboratory to prepare derivatives of 1, as fully described in the Supporting Information) was stable to DBU (3 d, rt). Sulfoxide 4 was evaluated as an MMP-2 inhibitor. Poor linear competitive inhibition of substrate hydrolysis (Ki = 2.1 µM) was seen for 4 without any evidence of time-dependent inhibition (Table 1). The pronounced difference in the kinetics behavior between 4 and 1 indicates different mechanisms for the two. Sulfoxide 4 is a linear competitive inhibitor, whereas sulfone 1 is a slow-binding inhibitor (7,13).
The catalytic velocity of substrate hydrolysis by MMP-2 in the presence of 1 progressively diminishes, as a result of the time-dependence inherent to slow-binding inhibition by 1. Classically, slow-binding behavior involves non-covalent chemistry, wherein the slow-binding event correlates to a conformational shift to a more stable enzyme-inhibitor complex. Deuterium incorporation into the inhibitor allows inquiry as to whether a kinetic isotope effect is expressed on any phase of the slow-binding kinetics. The progress curve for 1-d2 showed a five-fold decrease in the rate constant for the onset of the slow-binding transition (kon) and no difference in the rate constant for dissociation (koff). This kinetic isotope effect results in a five-fold increase in the apparent Ki, (a ratio of koff/kon). This result indicates that the entire isotopic effect is on the rate constant for the onset of inhibition (kon) and not on the rate constant for the offset (koff). It is logical that the koff values measured with 1 and 1-d2 are equal, as dissociation occurs after the deprotonation event in which the kinetic isotope effect is expressed. The appearance of this kinetic isotope effect implicates deprotonation at the α-carbon as the step that precipitates the onset of slow-binding process, leading to the more tightly bound state.
Two dimethyl-substituted derivatives of 1 were made. The α,α-dimethyl derivative 5, which is incapable of thiirane opening by deprotonation adjacent to the sulfone, shows no detectable inhibition of MMP-2 up to its solubility limit of 40 µM. The γ,γ-dimethyl derivative 6, which is capable of thiirane opening but is sterically encumbered compared to 1, is a linear competitive inhibitor with decreased affinity compared to 1. The dramatic difference between 1 and these two derivatives emphasizes the importance of the spare structure of 1 to the mechanistically important criteria of nanomolar potency and slow-binding onset of potent inhibition.
Vinylsulfone 7 is the kinetic product of base-mediated solvolysis of 1 and is also the presumed product of MMP-catalyzed action upon 1. The mechanism-dependent release of a latent thiolate by 1 coincides with the onset of the more stable inhibited state. Compound 7 was synthesized, and upon evaluation as an inhibitor of MMP-2 showed competitive inhibition with no evidence of a slow-binding kinetic transition. Its Ki value of 530 ± 70 pM is noteworthy. As this value is below the assay concentration of the enzyme, it was calculated using Morrison’s equation for analysis of tight-binding inhibition (45,46).
Thiirane 1 is a selective inhibitor of the gelatinases by a mechanism that ultimately results in thiolate ligand presentation to the active site zinc. The plausibility of thiirane opening by general base-catalyzed deprotonation at the carbon adjacent to the sulfone thiolate, when 1 is bound within the active site of MMP-2, was assessed computationally (Fig. 2). This assessment was made firstly by examination of the computationally predicted structure of the MMP-2·1 complexes (as 1 is racemic, and as both enantiomers of 1 are competent inhibitors, both the R-1 and S-1 complexes with the MMP require evaluation). Recognition by the gelatinase active site of inhibitor stereoisomers is precedented. The crystal structure of MMP-2 (PDB code: 1CK7) was transformed to that of the active enzyme by removal of the pro-domain in silico. Using the program DOCK, the R-enantiomer of 1 was computationally inserted into the active site, so as to allow coordination of the thiirane sulfur to the active site Zn(II). The structure of the S-enantiomer of 1 bound to the active site was generated from that of the R-enantiomer by inversion of configuration. Each complex was computationally equilibrated by dynamics simulation to bring the thiirane sulfur into the coordination sphere of the zinc ion. Further dynamics simulations (2 ns duration) of each complex (of the two enantiomers) were performed individually. We focused on the proximity of the carboxylate of Glu404 to the methylene hydrogens adjacent to the sulfone. As is shown in Figure 2, both enantiomer complexes show the approach of one oxygen atom of the carboxylate to within 2.4 Å of one of these hydrogens. This proximity is necessary for a proton abstraction initiating the thiirane ring-opening step. For the R-1 enantiomer, the pro-S hydrogen is in close contact to a Glu404 oxygen, while the pro-R hydrogen makes closer contact for S-1. While the time scale for these bond-making and bond-breaking events is much longer than this dynamic simulation, the calculated transition state for the non-enzymatic solvolysis of 1 fully substantiates this proposed mechanism (47). These simulations support C–H proton abstraction within the enzyme-inhibitor complex as the critical event leading to the tight-binding inhibition.
The departure of the cysteine thiol ligand from the active site zinc is a key event in pro-MMP maturation (38,39). In addition, the ability of endogenous thiols to suppress the cellular activity of the mature MMP enzyme (wherein the inhibitory cysteine ligand has been removed proteolytically), by engagement of the active site zinc, is well known. It is therefore hardly surprising that appropriate incorporation of the thiol functional group into precedented MMP inhibitor motifs attains exceptional inhibitory potency, as shown by Freskos et al. (48,49) with a structural motif related to 7, by our group with dithiol structures also related to 7 (8), and by others exploiting other thiol MMP/TACE inhibitor classes (32). However, application of the thiol as the zinc-binding group within a practical MMP inhibitor is challenging, due to the promiscuity of the thiol for metal binding and due to the ease of its oxidative and metabolic transformations (32). These challenges underscore the virtues of thiirane 1 as an MMP inhibitor. Its thiol is masked within the three-membered thiirane, with liberation of its latent thiol dependent on the base-catalysis machinery unique to the gelatinases. Moreover, the chemical structure of 1 is found by experiment to uniquely complement the machinery of the MMP-2 and MMP-9 gelatinase active sites. Inhibitor 1 yields a monodentate ligand for presentation to the active site zinc. X-ray absorption spectroscopy (XAS) comparison of the zinc-sulfur bond of MMP-2 inhibited by 1 to the cysteine-liganded zinc of pro-MMP-2 shows remarkable similarities in Zn—S bond length, ligation, and coordination geometry (8,50).
Thiirane 1 integrates three complementary structural components. The phenoxyphenyl moiety is a proven structural complement to the hydrophobic tunnel within the gelatinase active sites (51). The sulfone makes a key hydrogen bond to Leu-191 (MMP-2 numbering) in the active site (51) and the sulfone activates the adjacent methylene hydrogens for the deprotonation step that fragments the thiirane (41). The MD simulations indicate spatial feasibility for the Glu404 carboxylate to serve as the general base in this reaction. The third structural aspect of 1 is the thiirane, which first coordinates to the active site zinc (and is thus activated for fragmentation), and which is transformed to a thiolate anion upon thiirane ring opening. The transformation of the ligand from thioether (thiirane) to thiolate strengthens the inhibitor coordination to the zinc, mimicking the behavior of the cysteine of the pro-domain. The tight integration of these structural components is emphasized by the dramatic loss of inhibitory competence for many closely related structures to 1 (7).
The kinetics of MMP-2 (and MMP-9) inhibition by 1 indicate time dependence for loss of activity (13), and the electronic coordination environment of the inhibited enzyme implicates a thiirane opening event (50). Although 1 generates (and 7 has) a vinyl sulfone functional group, also well applied as a latent electrophile for protease inhibition, we have no evidence supporting covalent reaction of these vinyl sulfones with MMP-2. All of the experimental observations support the mechanism for gelatinase inhibition by 1 to be the thiirane-opening pathway B of Scheme 1. The salient question that follows this conclusion is the structural basis for the selectivity that 1 exhibits for gelatinase inhibition (7). The other members of the MMP family are either not inhibited by 1 at all, or are inhibited poorly. The success of this inhibitor in selective inhibition of gelatinases both in in vivo and in vitro experiments is impressive, and is a unique aspect of the inhibitor class. As our experiments reveal, the onset of inhibition is initiated by the rate-limiting deprotonation α to the sulfone function. This event is the likely step that initiates the time-dependence for inhibition, slow-binding behavior of the inhibitor, which leads to the formation of the thiolate. Its strong coordination to the zinc ion leads to tight-binding chemistry for the enzyme-inhibitor complex. Since the active sites of MMPs are very similar to each other, it is likely that 1 would interact relatively poorly with the active site zinc ions of MMPs, but in the cases of gelatinases, the complex is able to proceed via the requisite deprotonation event to tight-binding inhibition at picomolar level of potency. In essence, the thiirane is a caged entity that leads to the formation of the thiolate only in the active site of gelatinases. This chemistry for the mechanism of action is entirely unprecedented in the substantial literature of zinc protease (including MMP) inhibition. Compound 1 and its structural congeners have considerable promise for applications to diseases that are gelatinase-dependent.
This research is supported by the National Institutes of Health.
Additional Supporting Information may be found in the on-line version of this article: enzymatic assay methods, computational methods, and synthetic procedures.