Inhibitors of the Metalloproteinase Anthrax Lethal Factor

Abstract

Bacillus anthracis, a rod shaped, spore forming, gram positive bacteria, is the etiological agent of anthrax. B. anthracis virulence is partly attributable to two secreted bipartite protein toxins, which act inside host cells to disrupt signaling pathways important for host defense against infection. These toxins may also directly contribute to mortality in late stage infection. The zinc-dependent metalloproteinase anthrax lethal factor (LF) is a critical component of one of these protein toxins and a prime target for inhibitor development to produce anthrax therapeutics. Here, we describe recent efforts to identify specific and potent LF inhibitors. Derivatization of peptide substrate analogs bearing zinc-binding groups has produced potent and specific LF inhibitors, and X-ray crystallography of LF-inhibitor complexes has provided insight into features required for high affinity binding. Novel inhibitor scaffolds have been identified through several approaches, including fragment-based drug discovery, virtual screening, and high-throughput screening of diverse compound libraries. Lastly, efforts to discover LF inhibitors have led to the development of new screening strategies, such as the use of full-length proteins as substrates, that may prove useful for other proteases as well. Overall, these efforts have led to a collection of chemically and mechanistically diverse molecules capable of inhibiting LF activity in vitro and in cells, as well as in animal models of anthrax infection.

INTRODUCTION

Bacillus anthracis, an encapsulated, spore-forming, gram positive bacteria, is the causative agent of anthrax [1]. The most common form of anthrax in humans is a localized cutaneous infection that responds to antibiotic treatment. However, inhalation or ingestion of spores can lead to a systemic infection that is highly fatal even when treated with antibiotics [2]. The failure of antibiotics to cure anthrax infection has been attributed to two bipartite protein toxins, lethal toxin (LeTx) and edema toxin (EdTx), that are secreted by the bacterium. These toxins persist and damage host tissues even following antibiotic clearance of the bacterium itself. Indeed, intravenous injection of either LeTx or EdTx kills rodents and other animals, suggesting that they contribute to mortality associated with anthrax [3]. In addition, vaccination with toxin components provides protective immunity against anthrax [4], further underscoring their essential role in the disease. These observations have prompted the development of “anti-toxins” that can be administered in combination with classical antibiotics.

Both LeTx and EdTx are classic two-component bacterial toxins, having a targeting subunit that mediates uptake of an enzymatic subunit into the cytosol of host cells. The two toxins share a common targeting subunit called protective antigen (PA) but have distinct enzymatic subunits. Lethal factor (LF), a zinc-dependent metalloprotease, is the enzymatic component of LeTx, while edema factor (EF), a calcium/calmodulin-dependent adenylyl cyclase, is the enzymatic component of EdTx [5]. Toxin uptake begins when PA binds one of two homologous receptors, ANTXR1 (TEM8) or ANTXR2 (CMG2), on the target cell surface (Figure 1A) [6]. Cell surface-bound PA is processed by host furin-like proteases [7], removing a small N-terminal fragment. The processed form of PA spontaneously oligomerizes into a heptameric or octameric ring, creating multiple binding sites for LF and/or EF [8]. Following endocytosis of PA-LF or PA-EF complexes, acidification of the endocytic vesicle causes PA to dissociate from the receptor and partially insert into the endosomal membrane, forming a pore. Translocation of LF and EF through the pore is driven by the proton gradient across the endosomal membrane and requires at least partial unfolding of the proteins. Once translocated into the host cell, the enzymatic toxin components refold into their active conformations and act on cytosolic substrates [9].

Figure 1. Biochemistry and structure of anthrax lethal toxin.

A. The pathway for PA-mediated uptake of LF into cells. B. LF contains four domains. Domain I, the PA binding domain, is in pink, domains II and III are shown in dark blue and green, respectively, and domain IV, the catalytic domain, is in light blue. The catalytic zinc ion within the LF active site is shown in gray. C. Essential residues at the LF active site are shown in stick form. The figure was made using PyMol.

Multiple approaches to counteract LeTx and EdTx have been explored in an effort to develop anthrax therapeutics. Several of these strategies involve preventing toxin uptake into cells, including blocking the interaction between PA and either LF/EF or TEM8/CMG2, using dominant negative PA mutants, and administering a soluble CMG2 decoy receptor [10–18]. The only anti-toxin currently to be approved by the FDA is raxibacumab, a fully human monoclonal antibody against PA, which has shown efficacy in both pre- and post-exposure animal models when combined with antibiotics [19]. As small molecule inhibitors have generally favorable pharmacology and lower cost in comparison to biologics, direct inhibitors of LF and EF have also been explored. Because LF has historically been regarded as more important than EF in directly contributing to mortality in late stage systemic anthrax, and because proteases have proven to be a “druggable” class of proteins, substantial effort has been directed at the discovery and development of small molecule LF inhibitors, beginning with the first report of peptide-based LF inhibitors in 2002 [20]. Here, we review efforts to identify and refine small molecule LF inhibitors, with an emphasis on the last 6 years.

LF STRUCTURE, BIOCHEMISTRY, AND BIOLOGY

LF is a 90 kDa metalloproteinase that consists of four domains (Figure 1B). The N-terminal domain I mediates binding to multimerized PA and is thus also dubbed the PA binding domain of LF. Domain I has sequence and structural homology to the corresponding region of EF required for toxin uptake, and fusion of this domain to heterologous proteins allows their uptake into cells in a PA-dependent manner [21–22]. Curiously, domain I has the same overall fold as the catalytic domain of LF, domain IV, but lacks metal binding residues and has no catalytic activity. Domain II shares its fold with bacterial ADP-ribosylating toxins but similarly appears to lack enzymatic activity due to substitutions in residues required for catalysis. Domain III is a small helical bundle inserted into domain II. Together domains II and III comprise the central portion of the protein and may serve to regulate substrate access to the active site in domain IV [23–24]. Domain IV, as mentioned, is the catalytic domain of LF. It contains the signature active site HEXXH motif, including two metal-binding His residues (His686 and His690) and the catalytic Glu residue (Glu687) arranged on an α-helix, common to many metalloproteinase families (Figure 1C). The third metal-binding residue, Glu735, is located on a separate helix. LF has an additional essential active site residue, Tyr728, which is required for metal binding and catalysis and may contribute to transition state stabilization [23, 25–26]. LF was long considered to be a unique protease lacking close structural relatives. Recently however the X-ray crystal structure of the K. pneumoniae transcriptional co-activator MtfA (also called YeeI) revealed it to have an overall fold similar to the LF catalytic domain, including all zinc-binding and essential catalytic residues [23,27]. Interestingly, in the MtfA structure, an α helix absent from LF occludes the active site, contributing an additional His residue that coordinates the catalytic zinc ion, and efforts to convincingly demonstrate protease activity of either the K. pneumonia or E. coli protein have been unsuccessful [27, 28]. These observations suggest that either MtfA is not a protease despite structural similarity to LF, or that the published MtfA structure depicts an inactive precursor or zymogen conformation. MtfA orthologs exist in hundreds of gram negative bacterial species, including E. coli. Based on the presence of core catalytic residues, the Mop virulence factor from V. cholerae was also suggested to have the same fold, but this has yet to be confirmed by structural analysis.

LF has a highly restricted substrate repertoire, and has been reported to cleave only eight protein substrates. The major LF substrates in human hosts are mitogen-activated protein kinase kinases (MKKs) [29–30]. MKKs lie in the middle of three-component mitogen-activated protein kinase (MAPK) cascades that are critical components of signaling pathways that mediate responses to diverse cellular stimuli. LF cleaves six of the seven mammalian MKKs at sites near their N-termini, outside of their catalytic domains. LF cleavage inactivates MKKs by disruption of a critical docking site required for interaction with their MAPK substrates [23,31]. Cleavage of MKKs is thought to underlie observed effects of LF on numerous cell types. For example, inactivation of the p38 MAPK pathway by MKK3 and MKK6 cleavage is likely to cripple the ability of phagocytic cells to kill B. anthracis spores, thus promoting bacterial dissemination early in infection [31]. LeTx induced mortality at late stages of infection occurs through targeting of cardiomyocytes and vascular smooth muscle cells, possibly through blocking the ERK MAPK pathway via cleavage of MKK1 and MKK2 [3].

LeTx has long been known to cause rapid cytolysis of mouse primary macrophages and macrophage cell lines, and this phenomenon is the basis for a biological assay commonly used to evaluate LF inhibitors [33]. Interestingly, macrophage cell killing can occur through both MKK-dependent and MKK-independent mechanisms, depending on the mouse strain background and the presence or absence of inflammatory stimuli. Macrophages from “susceptible” strains of mice undergo rapid MKK-independent pyroptosis, a non-apoptotic form of programmed cell death characterized by activation of inflammatory pathways [34]. Macrophage pyroptosis depends on the presence of specific alleles of the inflammasome component NLRP1B that are susceptible to cleavage and activation by LF. By contrast, macrophages from “resistant” strains of mice have NLRP1B alleles that are cleaved, but not activated, by LF [35–36]. However, LF inactivation of the p38 MAPK pathway causes macrophage apoptosis in the presence of inflammatory mediators such as bacterial peptidoglycan [37]. While macrophage pyroptosis is not thought to occur in human anthrax, activation-dependent macrophage apoptosis may contribute to disabling the host immune system in human infection.

PEPTIDE BASED LF INHIBITORS

A common approach to the design of metalloproteinase inhibitors is to conjugate metal chelating groups to peptide substrates, providing high affinity binding to the protease active site [38]. For most metalloproteinases, peptides having an N-terminal metal chelating group followed by residues preferred by the protease at positions C-terminal to the cleavage site (the so-called “primed” side residues) provide the most potent inhibitors. Unusually, the first LF inhibitors reported were long C-terminal peptide hydroxamates that incorporated “non-primed” side residues (N-terminal to the cleavage site) found in MKK substrates, a typical example being (d-Arg)₉-Val-Leu-Arg-CONHOH [18, 30]. The stretch of nine Arg residues provided high affinity for LF, and also promoted uptake of the inhibitor into cells [20].

With an aim to create a more potent inhibitor, Li et al. recently set out to determine the substrate sequence preference of LF in order to incorporate preferred amino acids into these positions [39]. Using a mass spectrometry-based platform to analyze cleavage rates of components of a peptide mixture, they found LF to prefer the residues WLM-YPL in positions P3 through P3’ (where the cleavage site for a protease is defined as Pn…-P3-P2-P1-P1’-P2’-P3’-…Pn, with cleavage occurring between the P1 and P1’ residues). These sequence preferences largely correspond to previous peptide library analysis of LF that indicated strong preferences for Tyr at the P1’ position and Pro at the P2’ position [40]. With this information, the authors sought to optimize the previously reported peptide hydroxamate LF inhibitor (d-Arg)₉-Val-Leu-Arg-CONHOH by substituting favored residues. The resulting inhibitor, (d-Arg)₉-Trp-Leu-Met-CONHOH had a 5-fold increase in potency (K_i = 0.3 nM) compared to the original. Their study suggests that sequence optimization based on substrate preference can improve the potency of peptide based inhibitors [31].

While these long peptide hydroxamates are highly potent LF inhibitors in vitro, their activity in inhibiting macrophage killing by LeTx is relatively weak, requiring µM concentrations. There is evidence to suggest that what efficacy is observed in cultured cells may be at least partly attributable to weak inhibition of furin by the polyArg sequence [41]. Li et al. found that the hydroxamate group is susceptible to hydrolysis by prolonged incubation with LF, converting it to a weaker LF inhibitor, potentially explaining the low efficacy in cells [41]. Replacement of the hydroxamic acid group with the hydrolysis-resistant N, O-dimethyl hydroxamic acid (DHMA) group modestly improved activity in cell culture. However, the DMHA peptide was less potent than the corresponding hydroxamate as an LF inhibitor, presumably because methylation reduces zinc binding affinity. As it is not clear whether the hydroxamate group is labile in the context of the small molecule inhibitors described below, it remains to be determined if DMHA will serve as a generally useful replacement for hydroxamic acid in future LF inhibitors.

METAL-CHELATING SMALL MOLECULE LF INHIBITORS

While substrate peptide-based inhibitors are often the starting point in metalloproteinase inhibitor development, generally modifications are made through medicinal chemistry efforts to make the inhibitors less peptide-like in order to improve potency and pharmacological properties. Early on, short peptide-based hydroxamates incorporating primed side residues originally designed to inhibit matrix metalloproteinases (MMPs) were found to be modest LF inhibitors as well [40]. The first efforts to modify these compounds to optimize LF inhibition were carried out by a group at Merck, producing the hydroxamate compound 1 (K_i = 24 nM, Figure 2), which showed efficacy in a rabbit inhalational anthrax model [42]. Crystallography of LF in complex with 1 (cited as ligand 915 in the PDB entry 1YQY, and therefore subsequently referred to as L915) verified that the hydroxamate group chelated the active site zinc ion with the anticipated geometry, and showed that the fluorophenyl group occupied the hydrophobic S1’ specificity pocket (which binds to the P1’ residue in substrates), a key feature of potent competitive small molecule LF inhibitors.

Figure 2. LF inhibitors with metal chelating groups.

In a series of papers, a group from PanThera Biopharma investigated two lead series of small molecule hydroxamates based on the Merck compound L915 [43–45]. Initially the group explored replacement of the sulfonamide group by alternative two-atom or one-atom linkers (typified by compounds 2 and 4 in Figure 2, respectively). Potent analogs were identified that incorporated a long sidechain at the C2 position. Several analogs in each series were identified that displayed sub-nM inhibition of LF in vitro and could protect LeTx-challenged rats from death [46]. The compounds displayed excellent (>50,000-fold) selectivity for LF over mammalian MMPs. From structure-activity relationship (SAR) studies, potent inhibition and activity in vivo relied on an α-substituted benzylamine on the C2 side chain. The X-ray crystal structure of 2 in complex with LF was solved, revealing a canonical hydroxamate-zinc ion interaction and protrusion of the proximal fluorophenyl group into the S1’ pocket [44]. In addition, the critical benzylamine group bound to the S3 pocket, forming a water-mediated hydrogen bond to the metal-binding Glu735, explaining its contribution to efficacy of the compound. Additional modifications to the compounds produced 3 and 4 (Figure 2), which maintained potent inhibition of LF and improved pharmacological properties. These compounds were subsequently tested in a mouse model of infection with B. anthracis spores. It was observed that the compounds alone provide some protection from death. When administered in combination with subprotective doses of antibiotics, these compounds provide complete protection of animals subjected to spore challenge. Additionally, inhibitor 4 completely protected mice from spore induced death when combined with a monoclonal antibody against EF. These animal studies validate the potential value of potent LF inhibitors as a component of a combination therapy when combined with otherwise ineffective monotherapies. Overall, the development of these molecules is highly significant as they are the first compounds to match the inhibitory potency of long peptide hydroxamates, yet have the favorable properties of small molecules in vivo.

In 2012, Calugi et al. reported a novel series hydroxamic acid inhibitors related to Merck’s L915 compound, but based on a d-proline scaffold [47]. Similar to the Merck inhibitor, these compounds incorporated phenyl sulfonamide substituents designed to interact with the S1’ specificity pocket. They synthesized and evaluated a series of derivatives incorporating various S1’-interacting groups as well as hydrophobic substituents to the 3- or 4- position of the proline pyrrolidine ring. The most potent compound (5, Figure 2) inhibited LF with an IC₅₀ value of 1.4 µM.

FRAGMENT BASED LF INHIBITOR DISCOVERY

Fragment-based drug discovery approaches involve screening of low molecular weight compounds with the intention of discovering weak inhibitors. Fragments can then either be elaborated to generate larger molecules, or multiple fragments can be linked together, with the aim of producing tight binding inhibitors. For metalloproteinases, fragments containing metal chelating groups are a logical starting point. Indeed, in one of the first demonstrations of fragment-based drug discovery, tight binding inhibitors of MMP-3 were identified by linking a simple hydroxamate with biphenyl compounds that bind to the S1’ pocket [48].

Two groups have taken a fragment-based approach to identify new LF inhibitor scaffolds. The Cohen group initially screened a series of zinc-binding groups against LF to identify potential alternatives to the hydroxamate group [49], which has been observed to have poor oral bioavailability and limited zinc ion selectivity [40]. From this screen, simple hydroxypyrothiones inhibited LF about 50-fold more potently than a simple hydroxamate. Elaboration by conjugation of a biphenyl group aimed at targeting the S1’ pocket produced compound 6 (Figure 3), which modestly inhibited LF (IC₅₀ = 14 µM). Systematic evaluation of a large number of analogs revealed that substitution of a thioamide for the amide linker produced more potent analogs, including 7 (IC₅₀ = 5 µM, Figure 3). Molecular modeling suggested bidentate coordination of the zinc ion by the hydroxyl oxygen and thione sulfur atoms, as well as substrate-like hydrogen bonding interaction between the thioamide sulfur atom and backbone amides in a β-strand within the active site cleft. Screening a focused library of N-substituted hydroxypyridinethiones produced several compounds (including 8, Figure 3) of similar potency without further culture, these studies demonstrated the value of fragment based screening in identifying new zinc-binding groups, and produced compounds that are potentially valuable as a starting point for further optimization to create potent small molecule LF inhibitors.

Figure 3. LF inhibitors identified by fragment-based drug discovery and by virtual screening.

Taking a similar approach, Pellecchia and co-workers screened a library of ~500 fragments that contained metalchelating groups, including about 200 hydroxamates, to find inhibitors of LF and other metalloproteinases [50]. Fourteen fragments inhibiting LF >50% at 100 µM were found. Merging two fragments into a single molecule produced compound 9 with improved activity as an LF inhibitor (IC₅₀ = 3 µM). This compound also blocked MKK1 cleavage in cultured cells treated with LeTx, and protected macrophages from cytotoxicity.

IDENTIFICATION OF LF INHIBITORS BY VIRTUAL SCREENING

In order to identify new LF inhibitor scaffolds and zinc-binding groups, several laboratories have undertaken in silico screening efforts. Amin and co-workers virtually screened 35 million compounds by “topomeric” shape-based screening, in which compounds were scored by 3D shape similarity to the Merck inhibitor L915 [51]. The 22,133 top scoring hits were then computationally docked onto the LF crystal structure, yielding about 300 compounds scoring better than L915 and predicted to have drug-like properties. Of 39 commercially available compounds tested for inhibition of LF in vitro, three structurally similar dibenzylamines (10 – 12, Figure 3) had IC₅₀ values in the 50 – 75 µM range. In silico docking suggested an overall binding mode similar to that predicted for 6, with the phenolic oxygen coordinating the active site zinc ion and the distal benzyl group engaging the S1’ pocket.

Work by the Pellecchia group sought to develop a pharmacophore model for S1’ pocket binding inhibitors [52]. Crystallography of multiple peptide and inhibitor complexes revealed conformational flexibility of a loop delimiting the S1’ pocket of LF, in particular allowing the S1’ pocket to accommodate large substituents. Reasoning that this unique feature of LF might be exploited to produce more selective inhibitors, they performed a virtual screen of 200 commercially available biphenyl sulfonamides, which led them to discover 13 (Figure 3) as a modest in vitro inhibitor of LF (IC₅₀ = 14 µM). In vitro testing of an additional series of symmetrical biphenyl sulfonamides produced low µM inhibitors, including 14 (Figure 3). While modest in potency, these compounds displayed greater selectivity for LF over human MMPs compared with the Merck compound L915. Docking studies led to a new LF pharmacophore model incorporating a zinc binding group linked by a hydrophilic spacer to a hydrophobic component capable of fitting into the S1’ pocket and interacting with an adjacent region.

While several pharmacophore models for LF inhibitors have been proposed based on particular series of analogs having common properties, recently two groups have attempted to generate more universal pharmacophore models that incorporate elements of distinct classes of inhibitors. A model developed computationally by Roy et al. used as a training set both reported hydroxamate analogs of L915 and a series of structurally unrelated furan derivatives reported by the Pellecchia group [53]. Their best model, which involved two hydrogen bond acceptors and two hydrophobic features, performed well in predicting the activity of a set of 98 reported inhibitors that were excluded from the training set (R² = 0.77 for predicted activity vs. reported activity). A virtual screen of 2 × 10⁶ molecules using this model followed by docking studies returned 17 predicted inhibitors, though these were not tested for LF inhibition.

The Amin group also attempted to computationally define a universal LF active site inhibitor pharmacophore based on five reported inhibitors falling into three structural classes [54]. The resulting in silico screening model correctly identified 72% of the most potent experimentally identified LF inhibitors (IC₅₀ values in the nM range) and rejected all reported weakly active compounds (IC₅₀ values >100 µM). Following this test, they used the model to virtually screen for and identify new LF inhibitors. Their model incorporates multiple key features, including a zinc binding group, hydrophobic components that can interact with hydrophobic side chains of residues in the LF active site, and hydrogenbond donor and acceptor groups that interact with additional active site residues. It will be interesting to see how the two reported “universal” pharmacophore models perform in future studies aimed discovering novel inhibitors based on in silico screening.

DISCOVERY OF LF INHIBITORS BY HIGH THROUGHPUT SCREENING

While most clinical small molecule protease inhibitors were developed starting from substrate analogs [38], high throughput screening (HTS) remains important for discovery of new scaffolds, including novel classes of inhibitors such as those binding to allosteric sites. The development of efficient fluorescence resonance energy transfer (FRET) peptide substrates for LF enabled high throughput screens that led to the identification of some of the first LF inhibitors, which included both classical competitive inhibitors as well as noncompetitive inhibitors [55]. In addition, efforts to identify LF inhibitors provided a testing ground for novel screening technologies, for example utilizing NMR or mass spectrometry to detect substrate cleavage [55].

While the use of conventional HTS strategies to discover LF inhibitors has waned in recent years, in the past five years two groups reported the identification of new leads from screening diverse compound libraries. The Pellecchia group performed a high throughput screen of 16,000 drug like molecules using a FRET peptide cleavage assay [56]. From this screen, they identified a hit (compound 15, Figure 4) with structural similarity to L915, only having a thiophene group in place of the hydroxamate as the zinc binding group. SAR studies demonstrated the requirement of the thiophene and benzothiazole rings for inhibitor potency, and explored a variety of R groups that presumably bind at the S1’ pocket. The most potent compound identified (16, Figure 4) inhibited peptide cleavage by LF in vitro with a K_i value of 1 µM, and similar potency in protecting macrophages from LeTx. The compound also offered modest post-exposure protection in vivo in a mouse inhalation anthrax model.

Figure 4. LF inhibitors identified by HTS.

Similarly, Wei et al. screened a focused library of 1200 compounds originally designed as potential β-secretase inhibitors using a FRET peptide assay [57]. They identified compound 17 (SM157, Figure 4) as a modest hit (K_i = 9.2 µM). Despite the presence of a sulfonamide group common to many active site-binding inhibitors, 17 lacked an obvious metal-binding group and displayed non-competitive inhibition of peptide substrate cleavage, suggesting that it may bind at an allosteric site.

A group from Microbiotix characterized a large number of analogs of a non-hydroxamate LF inhibitor previously identified by Panchal et al. in a FRET peptide-based HTS screen [58–59]. Starting with the original symmetric urea hit (NSC12155, 18, Figure 4), they created a series of asymmetric amide derivatives and completed an SAR study to determine which substituents provided optimal LF inhibition. The most potent analogs, compounds 19 and 20 (Figure 4) inhibited LF cleavage of a peptide substrate with IC₅₀ values of 3.0 µM and 1.5 µM, respectively, and showed selectivity for LF over other metalloproteinases including botulinum neurotoxin A, MMP-1 and MMP-9. Curiously, while the original inhibitor 18 displayed competitive inhibition and was shown by X-ray crystallography to bind the LF active site, kinetic analysis suggested that the new asymmetric amide derivatives are non-competitive inhibitors of LF. These observations suggest that the modifications fortuitously produced compounds with a different mechanism of action. Whether this change in mechanism is accompanied by a distinct binding site or the induction of conformational changes affecting LF activity is not currently clear.

IDENTIFICATION OF INHIBITORS OF FULLLENGTH PROTEIN SUBSTRATE CLEAVAGE BY LF

The efforts to identify LF inhibitors described above have involved either elaboration of compounds with metal-binding groups or HTS using short peptide substrates. Both of these strategies will necessarily produce inhibitors that either bind at the protease active site or that allosterically influence the active site. A potential drawback of this strategy is that additional effort is required to avoid cross-reactivity with host metalloproteinases. An alternative strategy is to intentionally develop inhibitors that target protease exosites, such as regions outside of the protease active site that interact with protein substrates distal from the site of cleavage. LF is thought to have such an exosite essential for efficient cleavage of MKKs [60]. Theoretically, exosite-targeting LF inhibitors should not cross-react with host metalloproteinases, some of which are important for host defense against infection [61]. In addition, because the LF exosite has not been structurally characterized, exosite targeting inhibitors have potential value as structural probes to map the exosite. Recently, several groups have performed inhibitor screens using full length protein substrates, rather than short peptide substrates of LF, which have the potential to identify such exosite-targeting inhibitors.

Park et al. utilized a gel-based cleavage assay to screen 480 compounds for inhibition of LF cleavage of MKK1 in vitro [62]. This assay identified four compounds that inhibited LF with IC₅₀ values in the low µM range. While three compounds were found to be toxic to cells, one compound (21, Figure 5) successfully protected cells from LeTx challenge with an IC₅₀ of 39 µM. No kinetic analysis or structural characterization was undertaken to determine the mode of inhibition of these compounds.

Figure 5. Compounds discovered as inhibitors of MKK cleavage by LF.

We and our co-workers developed an HTS approach explicitly intended to identify exosite-targeting inhibitors of LF [63]. Our assay took advantage of the close proximity of the LF cleavage site to the N termini of MKKs. Using MKK6 fluorescently labeled at its N terminus allowed us to assay cleavage by fluorescence polarization, as proteolysis liberated a low molecular weight fluorescent fragment. This assay was used to screen a diverse library of 2835 synthetic compounds and natural products small molecules for inhibition of MKK6 cleavage by LF. Hits from this assay were then subjected to a secondary screen for inhibition of peptide cleavage. Active site inhibitors that affected peptide cleavage were discounted, and remaining hits were confirmed as inhibitors in a gel-based MKK cleavage assay. A single hit, the lichen natural product stictic acid (22, Figure 5), was found to specifically inhibit cleavage of MKKs without affecting cleavage of peptides. SAR analysis indicated that the central lactone ring was essential for activity, and reduction of the aldehyde group to an alcohol reduced potency, suggesting possible covalent interaction with LF. This compound and several close analogs also inhibited cleavage of MKKs and cell death in LeTxtreated macrophages. In addition to producing LF exosite inhibitors, this work has provided a new general screening method for identifying protease exosite inhibitors.

In another example of HTS to identify inhibitors of full length protein substrate cleavage, Kim et al. developed a modified yeast two-hybrid assay where LF cleavage disrupted activation of reporter genes [64]. They used this method to screen a library of chemical compounds for inhibition of MKK1 cleavage and identified one compound, 23 (Figure 5), that protects macrophages from LeTx induced cytotoxicity. Future work will be required to determine the mode of inhibition of the compound, and whether it selectively blocks cleavage of full length MKK substrates.

CONCLUSION

The efforts of many groups in the last several years have contributed to the development and improvement of LF inhibitors. Highly potent compounds have now been generated with efficacy in animal models, suggesting that they may have favorable pharmacological properties to become viable clinical candidates. Several new classes of molecules have emerged as well, some of which will undoubtedly be optimized in the coming years to generate highly potent analogs as well. As in the early days of LF inhibitor development, in recent years LF has been used as a platform to develop new technologies, such as those aimed at identifying exosite-targeting inhibitors. The selection of chemically diverse as well as mechanistically diverse inhibitors identified by the groups discussed in this review serve as a promising foundation for further clinical development of LF inhibitors.