Inhibition of Furin/Proprotein Convertase-catalyzed Surface and Intracellular Processing by Small Molecules*

Furin is a ubiquitously expressed proprotein convertase (PC) that plays a vital role in numerous disease processes including cancer metastasis, bacterial toxin activation (e.g. anthrax and Pseudomonas), and viral propagation (e.g. avian influenza and human immunodeficiency virus). To identify small molecule inhibitors of furin and related processing enzymes, we performed high-throughput screens of chemical diversity libraries utilizing both enzyme-based and cell-based assays. The screens identified partially overlapping sets of compounds that were further characterized for affinity, mechanism, and efficacy in additional cellular processing assays. Dicoumarols were identified as a class of compounds that inhibited furin non-competitively and reversibly with Ki values in the micromolar range. These compounds inhibited furin/furin-like activity both at the cell surface (protecting against anthrax toxin) and in the secretory pathway (blocking processing of the metastasis factor membrane-type 1 matrix metalloproteinase/MT1-MMP) at concentrations close to Ki values. Compounds tested exhibited distinct patterns of inhibition of other furin-family PCs (rat PACE4, human PC5/6 and human PC7), showing that dicoumarol derivatives might be developed as either generic or selective inhibitors of the PCs. The extensive clinical use, high bioavailability and relatively low toxicity of dicoumarols suggests that the dicoumarol structure will be a good starting point for development of drug-like inhibitors of furin and other PCs that can act both intracellularly and at the cell surface.

The multiple roles for furin in human pathophysiology have made it a target of interest for development of therapeutic agents. Numerous protein-and peptide-based furin inhibitors have been devised (23). For the most part, these are not druglike and their use as pharmaceutical agents is hampered by large size, instability, toxicity, and/or low cell permeability. Recently, 2,5-dideoxystreptamine derivatives have shown promise (24), although these molecules have yet to be examined for inhibition of intracellular processing. Important pathophysiological roles exist for furin at the cell surface, such as in the processing of anthrax protective antigen. However, maturation of other bac-terial toxins, viral envelope glycoproteins, and metalloprotease precursors such as membrane-type 1 matrix metalloproteinase (MT1-MMP), a matrix metalloprotease whose activity contributes directly to degradation of extracellular matrix components and is important for angiogenesis, tumor invasion, and metastasis (25), require processing by furin in the trans Golgi network and endosomal compartments (2,26).
Here we report identification of drug-like small molecule inhibitors through simultaneous high-throughput screening (HTS) of chemical diversity libraries with both enzyme-based and cell-based assays for furin and furin-like activities. A preliminary report of the cell-based assay has been published elsewhere (27). Combining the results of the enzymatic screen with the cellular screen allowed identification of small molecule lead compounds with the desired properties of high affinity, high cell permeability, and low toxicity. Dicoumarols, which have an extensive pharmacological history (28), were identified in this study as a family of compounds that inhibited furin reversibly and non-competitively, also inhibited rat PACE4 (rPACE4), human PC5/6 (hPC5/6), and hPC7 and blocked both extracellular maturation of anthrax protective antigen (PA) and intracellular processing of MT1-MMP and other substrates.
Enzymatic and Cellular HTS/ Dose-response Assay-Enzyme solution (15 l), containing ssfurin (11 nM), 20 mM Na/Mes, pH 7, 1 mM CaCl 2 , and 0.01% Triton X-100 buffer, or buffer alone (15 l) was delivered using the Multidrop 384 (Thermo Labsystems, Waltham, MA) into 384-well plates. Then ϳ300 nl of each compound (5-15 M) was delivered into individual wells (except control wells) using the Biomek FX pin tool (Biomek FX, Beckman, Fullerton, CA). After a 15-min incubation, 5 l of boc-RVRR-MCA (2 M) were added to all wells except the controls. The substrate in buffer without enzyme or the standard reaction containing 20 M dec-RVKR-CMK served as positive controls. After incubation (45 min), residual enzyme activities were monitored by the fluorescence intensity of 7-amino-4-methyl coumarin (AMC) ( ex 380 nm, em 470 nm) using a PHERAstar high-throughput microplate reader (BMGLabtech, Chicago, IL). All plates were barcoded for identification and linked to compounds from stock plates. Assay quality was determined by calculating mean, standard deviation, coefficient of variation, and ZЈ values for each plate. Plates for which ZЈ Ͼ 0.5 were accepted (34). Plates with ZЈ Ͻ 0.5 were repeated. Compounds with Ͼ60% ssfurin inhibition were chosen as positive hits. Compounds exhibiting Ͼ50% inhibition of other targets in the data base of the Center for Chemical Genomics were eliminated. Dose-response analysis was conducted to validate hits. The enzyme-based assay monitored AMC fluorescence. Approximately 1500 compounds fluorescent in the AMC range were ignored as they gave apparent enzyme activities greater than 100%. Examples of enzyme-based and cell-based assays for a sample compound plate are shown in Fig. 1. Note that DC3 and DC7 appear as clear hits for the enzymatic assay, enabling them to be easily identified in the cell-based assay.
For dose-response assays, compounds were transferred from stock plates (3-20 mM in dimethyl sulfoxide) into plates containing ssfurin and furin buffer and serial dilution was performed to give final concentrations of 33 M to 33 pM in 30 l (final dimethyl sulfoxide concentration Ͻ2%). After 15 min incubation, boc-RVRR-MCA (2 M) was added. After 45 min incubation, residual enzyme activities were monitored by AMC fluorescence intensity. Dose-response analysis was performed using PRISM software (GraphPad, San Diego, CA). Cellular The data shown represent enzymatic and cell-based furin inhibition assays on one of 100 384well plates containing library compounds. A, the enzyme-based assay identified DC3 and DC7 as hits that inhibited furin activity Ͼ60% (red line). Green squares represent AMC fluorescence released by furin cleavage in the presence of library compounds. Red squares represent positive controls, which correspond to signals from reactions inhibited by 20 M dec-RVKR-CMK or containing substrate boc-RVRR-MCA without enzyme added. Blue squares represent negative controls, in which furin was incubated with the substrate in the absence of any inhibitor. About 2-5% of compounds on any given plate exhibited fluorescence that interfered with measurement of AMC fluorescence and resulted in apparent activity Ͼ100%; these are represented by green squares in the negative range. B, the cell-based assay identified these DC2 and DC7 and five additional potential hits as compounds that inhibited GRAP processing Ͼ55% (red line). Positive (red squares) and negative (blue squares) controls in the cell-based assay correspond to processing inhibited by 20 M dec-RVKR-CMK and to complete processing of GRAP, respectively (27). Apparent inhibition greater than 100% is due to cytotoxicity of compounds or, in the case of DC7, inhibition exceeding that seen with the positive control.
HTS and dose-response assays were performed as previously described (27). Briefly, these assays depended on inhibition of cleavage of a fusion protein, GRAPfurin, expressed in Chinese hamster ovary (CHO) cells. GRAPfurin consists of alkaline phosphatase connected to the cytosolic and transmembrane domains of ␤-secretase (BACE) through a 10-amino acid furin recognition sequence derived from Stromolysin 3 (27). Furin cleavage results in release of alkaline phosphatase, which can be assayed in the conditioned medium.
Inhibitor Affinities, Mechanisms, and Enzyme Selectivitiesssfurin was active-site titrated as described (30). Initial assays were performed in furin assay buffer (20 mM Na/MES, pH 7.0, 1 mM CaCl 2 , 0.1% (v/v) Triton-X-100) (29) by preincubating enzyme (2.0 nM) with inhibitors at a range of concentrations for 30 min at room temperature, then adding substrate boc-RVRR-MCA (2 to 4 M) and measuring its rate of hydrolysis (residual enzyme activity), which was linear for 15-30 min. AMC fluorescence was measured using an f max 96-well fluorimeter (Molecular Devices). IC 50 values were derived using a Dixon plot (35). Analysis of these data using Eadie-Hoftsee plots suggested that the dicoumarols and compound B3 were non-competitive, whereas compounds B1, B2, and B5-9 were competitive inhibitors (data not shown). In the case of dicoumarols, initial inhibition assays were done by preincubating enzyme with inhibitors because inhibitors appeared to exhibit slow binding. However, this was an artifact due to inhibitor aggregation, which occurred reversibly after dilution into aqueous solution from concentrated stock solutions in dimethyl sulfoxide. When inhibitors were diluted into assay buffer and preincubated for Ն60 min prior to addition of enzyme and substrate, no slow-binding characteristics were seen. Dicoumarols also exhibited complex formation with the AMC substrate boc-RVRR-MCA when the substrate was present at higher concentrations (at or above the K m ), making accurate determination of K i values difficult. A similar phenomenon was described previously (36). The dicoumarols were also fluorescent, which required correction. To circumvent these problems, we performed more accurate inhibition assays using the chromogenic substrate Ac-RVRR-pNA, which permitted use of higher substrate concentrations at high inhibitor concentrations. The K m of ssfurin for this substrate was determined to be 12.3 M. p-Nitroaniline release was monitored at 410 nm using a TECAN Safire 2 , microplate reader in 96-well format. Dicoumarol compounds were diluted into assay buffer and incubated for 2 h. Enzyme and substrate were then added together, and inhibition curves (linear) were measured for 30 min. Substrate concentrations were varied from 4 to 80 M and inhibitors were assayed at concentrations that ranged, roughly, from 0.1 K i to 3 K i . Assays were performed in triplicate and repeated at least 3 times. To determine inhibition mechanisms and K i values, residual enzyme activity was plotted versus inhibitor concentration, generating a series of curves that differed by substrate concentration. The plots were fitted using untransformed equations for competitive, non-competitive, or uncompetitive inhibition using KaleidaGraph (version 4.0). This method was used to determine K i values and the inhibitory mechanism for all dicoumarol inhibitors and compounds B2 and B3. K i values determined in this way were nearly identical to IC 50 values obtained previously from Dixon plots.
Intracellular Processing Assays-To assay processing of the intracellular substrate CPA95 using transient transfection, cells were plated in 6-well dishes, incubated overnight, then transfected with the CPA95 expression plasmid. CPA95 is a derivative of rat carboxypeptidase A1 in which the trypsin cleavage site at Arg-95 (FQAR 95 Q) has been altered to create a consensus recognition sequence for furin (RQKR 95 Q) (37).
At 36 h post-transfection, medium was replaced with Opti-MEM (Invitrogen) containing compound or vehicle. After 2 h, medium was discarded and replaced with fresh Opti-MEM containing compound or vehicle. After overnight incubation, conditioned medium and cells were harvested and analyzed by Western blotting.
Anthrax Toxin Killing Assays-Anthrax toxin killing assays were performed as described (33) except as noted below. J774A.1 murine macrophage cells were washed once with modified Ringer's buffer (RB * ) (38) and overlaid with RB * or RB * containing compounds to be tested at various concentrations. Anthrax toxin (AT) (recombinant PA 83 (12 nM) and recombinant lethal factor (1.2 nM)) was then added, cells were incubated for 2.5 h and live versus dead cells were determined using the ToxCount cell viability assay (Active Motif, Carlsbad, CA). Cells were photographed using a Zeiss Axiovert fluorescence microscope and numbers of green (live) and red (dead) cells (600 total) were counted to calculate percent protection.
Western Blot Analysis of Intracellular Processing-Western blotting was performed as described (39). Protein concentration was determined using the detergent-compatible protein assay kit from Bio-Rad. Carboxypeptidase A was detected using rabbit polyclonal antibody AB1213 (Chemicon, Temecula, CA). solMT1-MMP expression was detected using rabbit anti-MT1-MMP antibody (Chemicon). HA-MT1-MMP expression was detected using a mouse monoclonal antibody to HA (Covanence, Princeton, NJ). Blots were incubated with appropriate horseradish peroxidase-conjugated secondary antibody followed by detection with chemiluminescent horseradish peroxidase substrate (Pierce) or ECL Plus (GE Healthcare). Anthrax PA processing was quantified using a Typhoon Trio (GE Healthcare).
Intracellular IC 50 Determination and MT1-MMP Processing Quantification-For IC 50 determination, Western blots were scanned and processed using a GE Healthcare Storm 860 phosphorimager and ImageQuant software (GE Healthcare) or films scanned and quantified by NIH ImageJ software. For MT1-MMP processing quantification, the films were scanned and quantified using NIH ImageJ software. The intensity of the processed band was divided by the intensity of the total (processed ϩ unprocessed) bands for the control lane to obtain the percentage processed. The percentage processed in the untreated control represents 0% inhibition. Percentage processed in the presence of compounds were normalized to untreated controls. IC 50 values were expressed as mean percent inhibition and plotted against compound concentration. IC 50 values were defined as the concentration of compound required to reduce CPA95 processing by 50% relative to control. Assays were repeated in triplicate (error Ϯ S.E.).

HTS of Small Molecule Libraries Using Enzymatic and Cellular
Assays-With the goal of identifying compounds that inhibited furin and were bioavailable within the Golgi compartment, we utilized two independent screens, one that utilized the purified enzyme and a second that utilized a live cell assay for furin and furin-like activities. The enzymatic assay measured cleavage of a fluorogenic peptide substrate by purified secreted, ssfurin (29,30). The cellular assay utilized CHO cells expressing a trans Golgi network-localized furin substrate, GRAP, whose cleavage by furin or other PCs results in secretion of soluble alkaline phosphatase (27). Details of the screens are found under "Experimental Procedures." Both assays were used to screen commercial small molecule diversity libraries from ChemDiv and Chembridge (ϳ30,000 compounds), with positive hits defined by inhibition of Ͼ60%. Results of the screens were analyzed further to eliminate false-positives, toxic compounds, and nonspecific inhibitors. Inhibitory activities were confirmed by dose-response assays, resulting in selection of 12 compounds that were characterized in detail. This set was supplemented with 2 compounds (B5 and B9) identified by screening against the yeast furin homologue, Kex2, and with structural analogs (DC1, DC2, DC4, DC6, and B10) of compounds identified in the screen.
Five of the compounds derived from the HTS (DC3, DC5, and DC7-9) were derivatives of dicoumarol (DC4) (Fig. 2A). Dicoumarol itself and several other dicoumarol derivatives not identified in the screen (DC1, DC2, and DC6) were obtained commercially and also found to inhibit ssfurin.
Affinities and Mechanisms of Inhibition-Initially, affinities of all of the inhibitors shown in Fig. 2 were characterized by determining IC 50 values assuming a simple binding mechanism with boc-RVRR-AMC as a substrate (see "Experimental Procedures"). Preliminary analysis of these data by Eadie-Hoftstee plots suggested that all of the dicoumarols and B3 inhibited by a non-competitive mechanism and that the basic compounds (B1, B2, B5, and B9) were competitive inhibitors (data not shown). The dicoumarol compounds DC1-9 and compounds B2 and B3 were then characterized more carefully to determine accurate K i values using Ac-RVRR-pNA as a substrate to avoid problems with inhibitor fluorescence and inhibitor-substrate aggregation (see "Experimental Procedures"). These data were fit to untransformed equations for competitive, non-competitive, and uncompetitive inhibition. Fig. 3, A-D, shows examples of inhibition data fit assuming a non-competitive mechanism for DC1, DC4, and DC7 (panels A-C) and a competitive mechanism for B2 (panel D). Fig. 3, E and F, show plots of K i deter-mined assuming competitive, non-competitive, or uncompetitive inhibition for DC1 and B2, respectively. As seen in Fig. 3E, only the non-competitive inhibition model gave a consistent K i value for DC1 (and for the other dicoumarols and B3, data not shown). As seen in Fig. 3F, only the competitive inhibition model gave a consistent K i value for B2. We conclude that the dicoumarol compounds inhibit by a purely non-competitive mechanism. B2 is clearly competitive. The other basic compounds appear also to be competitive, but were not characterized in greater detail.
Inhibition of ssfurin by dicoumarol compounds was reversible (Fig. 4). When ssfurin was preincubated with DC4 at 10 ϫ K i in the absence of substrate and then diluted 10-fold into buffer containing substrate, activity was slightly in excess of that expected (50%) for full reversibility (Fig. 4, curve 2). No such recovery of activity was observed when enzyme was diluted into a reaction containing DC4 at 10 ϫ K i (Fig. 4, curve 3). Incubation of ssfurin with DC3 and DC9 for up to 2 h also showed no evidence of irreversible inactivation (data not shown).
Inhibition of Extracellular Processing of Anthrax Protective Antigen-Furin and furin-like proteases act at the cell surface to process the precursor form of anthrax protective antigen (PA), PA 83 , to the mature form, PA 63 , which forms the heptameric prepore that binds and facilitates translocation of anthrax lethal factor and edema factor (40). Inhibition of furin activity protects against anthrax toxicity caused by PA and lethal factor ("AT") (33, 41). As previously described (33), addition of recombinant PA (12 nM) ϩ lethal factor (1.2 nM) killed Ͼ99% of J774A.1 murine macrophages within 2.5 h of incubation (Fig.  5B, AT), whereas control cells exhibited no lysis (Fig. 4B, "Neg"). Compounds DC1, DC2, DC3, and B3 all protected macrophages from AT-mediated cell death, and exhibited distinct titration curves (Fig. 5A). DC1 and B3 exhibited optimal protection at 5 ϫ K i , showing toxicity at higher concentrations. In contrast, DC2 and DC3 exhibited optimal protection at 20 ϫ K i , with no significant toxicity. Fig. 5B shows representative images of cells treated with AT in the presence of optimal concentra-  To confirm that protection was due to inhibition of PA cleavage, we tested each ability of the compound to inhibit processing of PA 83 to PA 63 . When macrophage cultures were incubated with 12 nM PA 83 for 2.5 h and cell extracts were analyzed by immunoblotting with anti-PA antibody, ϳ40% of cell-associated PA 83 was processed to PA 63 (Fig. 5C). When cells were treated with DC1, DC2, or DC3 at concentrations equivalent to 5 ϫ K i or B3 at concentrations equivalent to 2 ϫ K i , no significant cleavage of PA 83 was observed (Fig. 5C).
Inhibition of Intracellular Processing-To examine inhibition of intracellular processing reactions catalyzed by furin or other PCs, we employed live cell assays using the engineered furin substrate CPA95 (37). CPA95 is a secreted substrate engineered to contain a furin cleavage motif at residue 95 (see "Experimental Procedures"). CPA95 is a secreted substrate engineered to contain a furin cleavage motif at residue 95. Expression of CPA95 in transfected CHO cells resulted in the appearance of both unprocessed (43 kDa) and processed (32 kDa) forms of the protein in the extracellular medium. Treatment of these cells with specific inhibitory compounds resulted in a dose-dependent decrease in the processed form and increase in the unprocessed form (Fig. 6A). The compounds tested showed a range of inhibitory capacity with DC1, B3, DC4, B10, and DC2 exhibiting IC 50 values of ϳ4, ϳ7, ϳ20, ϳ25, and ϳ55 M, respectively (Fig. 6B). The same compounds tested in Fig. 6 also inhibited intracellular processing of pro-von Willebrand factor (data not shown). (25). To determine whether compounds identified in this study could inhibit processing of MT1-MMP, CHO cells were transfected with a plasmid encoding HA epitope-tagged MT1-MMP. In the absence of furin inhibitors, transfection of MT1-MMP in cells resulted in the appearance of the precursor (ϳ63 kDa) and mature (ϳ60 kDa) forms of MT1-MMP as shown by Western blot analysis using an HA-specific antibody (Fig. 7A, bottom). Addition of DC4, DC1, DC2, B10, and B3 resulted in marked reduction in the appearance of mature MT1-MMP. Inhibition of processing, relative to control, was quantified as and expressed as a percentage (Fig. 7A,  top). Treatment of these cells with compounds inhibited MT1-MMP processing. An 8.8% reduction was observed with 40 M DC4, 34% with 10 M DC1, 63% with 50 M DC2, 42% with 40 M B10, and 86% with 15 M B3. In this experiment, the well characterized furin inhibitors decRVKR-CMK (20 M) and ␣1-PDX (42), the latter introduced by cotransfection, inhibited MT1-MMP processing by 40 and 81%, respectively (Fig. 7A,  top). To confirm the inhibitory effects of these compounds on furin/PC-mediated MT1-MMP cleavage, we transfected CHO cells with a plasmid encoding soluble MT1-MMP (solMT1-MMP; i.e. MT1-MMP lacking the transmembrane domain) (43), which is completely processed to the mature form by endogenous furin (26). As shown by Western blot analysis using an MT1-MMP antibody, processing of solMT1-MMP by furin/PCs in CHO cells resulted in appearance of the mature (ϳ50 kDa) form of solMT1-MMP in cell lysates and extracellular medium (Fig. 7B, lane 1). Addition of 10 M DC1, 50 M DC2, 40 M B10, or 15 M B3 resulted in a reduction of processed solMT1-MMP in cell lysates comparable with that seen with 20 M decRVKR-CMK (Fig. 7B), with no significant mature solMT1-MMP released to the medium.

Inhibition of Intracellular Maturation of MT1-MMP-Inhibition of furin/PC-dependent maturation of MT1-MMP offers a potentially important new avenue to impede tumor invasiveness and metastasis
Inhibition of rPACE4, hPC5/6, and hPC7-A subset of compounds was tested against purified catalytic domains of rPACE4, hPC5/6, and hPC7, other members of the PC family that, like furin, function in the constitutive secretory pathway, the endocytic pathway, and/or the cell surface and exhibit broad tissue distributions. K i values are shown in Table 2. Although all compounds tested inhibited all four enzymes (ssfurin, rPACE4, hPC5/6, and hPC7), each compound exhibited a unique pattern of inhibitory potency. DC1 and DC3 exhibited similar K i values for ssfurin, rPACE4, and hPC5/6 but showed significantly lower affinity for hPC7. DC2 inhibited all four PCs with similar efficacy. B3 had its strongest inhibitory activity against hPC7, exhibited similar inhibitory activities toward ssfurin and rPACE4, and was less effective with hPC5/6. All four compounds also inhibited human ␣-thrombin to some degree.
Toxicity of Furin Inhibitors under Conditions of Tests of Inhibition of Processing-Cytotoxicity was examined by incubating J774A.1 murine macrophages with compounds for 2.5 to 8 h (Table 3). Dicoumarol derivatives DC1, DC2, DC3, and DC4 showed no significant toxicity when incubated with macrophages at 20 ϫ K i for 8 h (Table 3). B3 showed some toxicity when incubated with macrophages at 2 ϫ K i .

DISCUSSION
Proprotein processing reactions provide potential targets of drug intervention for both chronic and infectious human disease (44). Because a large number of processing reactions are catalyzed by PCs, these enzymes have emerged as important potential drug targets (23). Drugs that are not cell-permeable may be effective in the case of extracellular processing events, such as maturation of anthrax protective antigen. More generally, however, small molecules with high bioavailability and cell permeability and exhibiting low toxicity are required to inhibit processing in intracellular compartments. Furin, PACE4, PC5/6, and PC7 all function in the constitutive secretory pathway, in the endocytic pathway, and/or at the cell surface and exhibit broad tissue distributions (2). Because of their similar patterns of substrate recognition and at least partial overlap in expression, it is currently unclear whether the goal of drug development targeting pathophysiological effects of furin or furin-like processing should be directed toward inhibitors exhibiting a high degree of discrimination between PC family members or toward more generic inhibitors that can block activities of all the PCs.
Here we report the results of a high-throughput screen for small molecule inhibitors of furin/PCs that relied on simultaneous use of enzymatic and cellular assays. Use of the cell-based assay identified compounds that inhibited intracellular processing catalyzed by furin alone or by furin along with other PCs while eliminating compounds with substantial toxicity. Use of the enzymatic assay identified compounds that directly inhibited ssfurin. The combination of assays allowed us to focus on furin inhibitors that had high promise for intracellular inhibition. The combined screens identified a set of competitive inhibitors that were largely unrelated structurally and a set of non-competitive inhibitors, mostly from one structural family. Many of the competitive inhibitors were basic compounds that performed poorly due to higher toxicity or lower cell permeability when further cell-based assays were employed. We focused our efforts instead on the non-competitive inhibitors, all of which were structurally related or analogous to dicoumarol. Identifi-  cation of these compounds as furin inhibitors is significant for several reasons. First, to our knowledge, no other noncompetitive furin inhibitors have been described. Second, inhibition by several of these compounds (DC3, DC4, and DC9) was reversible, indicating that these molecules do not undergo nonspecific covalent reactions with ssfurin. Third, dicoumarols have an extensive history of use as pharmacological agents that makes them an attractive starting point for drug development.
Dicoumarol (DC4) has been used clinically as an anticoagulant with vitamin K antagonist activity similar to that of warfarin (45)(46)(47). A large number of dicoumarol derivatives have been synthesized and assessed for anticoagulant activity (48). The use of dicoumarol as an oral anticoagulent is indicative of high bioavailability, low toxicity, and high cell permeability. Indeed, several dicoumarols examined here exhibited high cell permeability as the IC 50 values for inhibition of intracellular processing (Fig. 6B) were quite close to the K i values measured with purified ssfurin. Although the furin inhibitors identified in this screen exhibited K i values in the micromolar range, their effective concentrations for inhibition of both cell surface and intracellular processing were also in the low micromolar range. Inhibition of intracellular processing reactions by other characterized furin inhibitors, including engineered proteins (20,49) and reactive peptides (50,51), has required inhibitor concentrations ranging from ϳ1,000 to 10,000 ϫ K i . In contrast, IC 50 values for inhibition of intracellular processing by DC1, DC2, and DC4 were ϳ4 ϫ K i , 17 ϫ K i , and 1 ϫ K i , respectively ( Figs. 2A and 6B). Moreover, IC 50 values for previously characterized furin inhibitors when assayed in cell-surface processing reactions have also been high in comparison to K i values: 210 ϫ K i for RRD-eglin (33), 120 ϫ K i for Ac-Arg-Glu-Lys-boroArg (33), 2800 ϫ K i for D9R (52), and 190 -360 ϫ K i for 2,5-dideoxystrepatmine derivatives (24). In contrast, IC 50 values for inhibition of intracellular processing by DC1, DC2, and DC3 were Ͻ5 ϫ K i , 1 ϫ K i , and ϳ2 ϫ K i , respectively ( Figs. 2A and 6B). This remarkable efficiency of inhibition may be related to the ability of these compounds to inhibit both the intracellular and surface pools of processing enzymes.
Although the therapeutic target of dicoumarols when used as anticoagulants is presumptively vitamin K epoxide reductase (53), another enzyme, NAD(P)H quinone oxidoreductase (DT diaphorase), has also been shown to be a highaffinity target (54). In addition, dicoumarol has been reported to inhibit gap junction formation in cultured cells (55) and ADP-ribosylation of CtBP3/BARS, resulting in alterations in Golgi morphology and function (56). These   effects may or may not influence the toxicity of dicoumarol but have not precluded clinical use. The PCs represent a new set of physiological targets for inhibition by dicoumarol and its derivatives. Because of the functional redundancy of the PCs, it is difficult to ascertain whether inhibition of processing in the individual cellular assays (processing of anthrax PA, CPA95, GRAPfurin, von Willebrand factor, or MT1-MMP) is due solely to inhibition of furin, inhibition of other PC family members or both. Nevertheless, our results suggest both that single compounds can inhibit multiple members of the PC family (Table 2) and that selectivity between enzymes can be achieved through modification of the dicoumarol structure ( Fig. 2A).
For several of the compounds tested, concentrations required for inhibition in cellular assays are similar to the therapeutic doses of dicoumarol, suggesting that affinities may be close to that needed for drug models. Nevertheless, because dicoumarols inhibit multiple target enzymes, optimization of compounds will be necessary. Dicoumarol represents a good platform for structural modification that might yield compounds with increased affinity for furin and other PCs. Increasing selectivity, both for individual PCs and for PCs versus other targets (e.g. vitamin K epoxide reductase and NAD(P)H quinone oxidoreductase) should further reduce off-target effects. Production of higher affinity derivatives may also assist in identifying the dicoumarol binding site within furin and yield information regarding the molecular mechanism of non-competitive inhibition.