Optimization of Furin Inhibitors To Protect against the Activation of

Article pubs.acs.org/jmc

Optimization of Furin Inhibitors To Protect against the Activation of Influenza Hemagglutinin H5 and Shiga Toxin Hugo Gagnon,† Sophie Beauchemin,‡ Anna Kwiatkowska,† Frédéric Couture,† François D’Anjou,† Christine Levesque,† Frédérik Dufour,† Adamy Roberge Desbiens,§ Rolland Vaillancourt,§ Sylvain Bernard,‡ Roxane Desjardins,† François Malouin,*,§ Yves L. Dory,*,‡ and Robert Day*,† †

Institut de Pharmacologie de Sherbrooke (IPS) and Département de Chirurgie/Urologie, Faculté de Médecine et des Sciences de la Santé (FMSS), Université de Sherbrooke, 3001, 12e Avenue Nord, Sherbrooke, Québec J1H 5N4, Canada ‡ Département de Chimie, Faculté des Sciences, Institut de Pharmacologie de Sherbrooke (IPS), Université de Sherbrooke, 3001, 12e Avenue Nord, Sherbrooke, Québec J1H 5N4, Canada § Département de Biologie, Faculté des Sciences, Centre d’Étude et de Valorisation de la Diversité Microbienne (CEVDM), Université de Sherbrooke, Sherbrooke, Quebec, J1K 2R1, Canada S Supporting Information *

ABSTRACT: Proprotein convertases (PCs) are crucial in the processing and entry of viral or bacterial protein precursors and confer increased infectivity of pathogens bearing a PC activation site, which results in increased symptom severity and lethality. Previously, we developed a nanomolar peptide inhibitor of PCs to prevent PC activation of infectious agents. Herein, we describe a peptidomimetic approach that increases the stability of this inhibitor for use in vivo to prevent systemic infections and cellular damage, such as that caused by influenza H5N1 and Shiga toxin. The addition of azaβ3-amino acids to both termini of the peptide successfully prevented influenza hemagglutinin 5 fusogenicity and Shiga toxin Vero toxicity in cell-based assays. The results from a cell-based model using stable shRNA-induced proprotein convertase knockdown indicate that only furin is the major proprotein convertase required for HA5 cleavage.

■

recognized by PCs (furin, PC5/6, and PC7).7,16−18,20,22−27 Upon cleavage of an HA-containing multibasic cleavage site present in the TGN, HA allows viral membrane fusion with the host cell membrane, which results in the release of viral genetic material and viral replication.5,21,28−30 Similarly, the Shiga toxins produced by Shigella dysenteriae and Shiga-like toxinproducing Escherichia coli (STEC) are examples of PC-activated bacterial toxins.31,32 Shiga toxins contain an RXXR sequence in a loop stabilized by a disulfide bond that is very sensitive to cleavage by trypsin and PCs.8,33,34 Furin was proposed as the PC responsible for this cleavage, which occurs at low pH in the trans-Golgi network or in the endosomes, resulting in a rapid intoxication of cells.35 These viral and bacterial targets are particularly attractive for use in preventing future pandemics and the associated economic effects of these pathogens. The influenza virus rearrangements responsible for recent pandemic episodes were unpredictable,1 and the acquisition of this multibasic cleavage site by human influenza could lead to virulent influenza pandemics.36−38 Foodborne STEC, which is often associated with undercooked ground beef and contami-

INTRODUCTION The observed effectiveness of many anti-infective agents often rapidly decreases because of increased pathogen resistance.1−3 Targeting host cell proteins has been proposed for various pathogens as an alternative approach to developing new antiinfective agents to prevent such occurrences.4−6 The proprotein convertases (PCs), which belong to the family of serine proteases, have been suggested as potential therapeutic targets against various pathogens.7,8 Seven of nine PCs (i.e., furin, PACE4, PC1/3, PC2, PC4, PC5/6, and PC7) recognize a consensus cleavage site, R-X-R/K-R↓, in precursor proteins.9 These widely expressed proteases10−13 primarily activate precursor proteins and produce smaller bioactive products.14,15 Many reports have implicated PCs in the maturation of viral (e.g., highly pathogenic avian influenza (HPAI) hemagglutinin (HA)16−18) or bacterial toxins (e.g., Shiga toxin, Stx19). The processing of pathogen precursor proteins by PCs results in increased pathogen infectivity, symptom severity, and lethality.17,19−21 Approximately 60% of human HPAI cases caused by H5N1 result in lethal infections, with spread of the infection to organs typically unreached by the virus itself.22 These highly infectious characteristics are associated with the gain of a multibasic (RRRKKR) cleavage site in HA that is © 2013 American Chemical Society

Received: April 29, 2013 Published: December 10, 2013 29

dx.doi.org/10.1021/jm400633d | J. Med. Chem. 2014, 57, 29−41

Journal of Medicinal Chemistry

Article

Figure 1. Ki, stability, and toxicity of the Ac-RARRRKKRT-NH2 peptide inhibitor. (A) Peptide stability in DU145 cells cultured in the presence or absence of FBS or in plates containing medium + FBS alone (no cells). (B) Ex vivo plasma stability measured by incubating the peptide with mouse plasma at 37 °C. All stability studies were analyzed using HPLC. Half-life and R2 values were determined using an exponential fit of the HPLC data and were calculated using GraphPad Prism. (C) Peptide degradation patterns in cell culture medium analyzed using MALDI-TOF. The peaks corresponding to peptide cleavage are framed. (D) Peptide cytotoxicity at the indicated concentration on a monolayer of HEK293FT cells after 24 h of incubation measured using the MTT assay (n = 4). The relative cytotoxicity was normalized against nontreated cells. (E) Acute cytotoxicity of the peptide at the indicated concentration measured using a lactate dehydrogenase (LDH) assay after 4 h of incubation with subconfluent HEK293FT cells. The relative cytotoxicity was compared with a 100% toxicity control (n = 4). Data represent the mean ± SEM of three independent experiments.

and Stx Vero cell toxicity. An examination of the specificity of PCs in HA cleavage provides insight into which PCs should be targeted for the design of broad-spectrum antiviral/antitoxin PC inhibitors. Our findings demonstrate that shRNA-mediated silencing is a suitable tool to study PC cleavage specificity in cell-based studies.

nated vegetables, is responsible for thousands of hospitalizations worldwide39 and costs $405 million (in 2003) in the U.S. each year.40 Because of the important role of PCs in these pathogens, recent efforts have been undertaken to develop inhibitors to prevent the harmful effects of viral infections and bacterial toxin activation.7,41−43 A peptide derived from the TPQRERRRKKR cleavage site of H5N1, TPRARRRKKRT, was found to be a specific nanomolar inhibitor of furin that is able to protect mice from anthrax inhalation.8 This peptide was further optimized by C-terminal amidation and N-terminal acetylation to yield an improved nanomolar inhibitor, Ac-RARRRKKRT-NH2, which effectively prevents the processing of viral glycoproteins and anthrax toxin without significant cellular toxicity.8 For this inhibitor to be useful in the prevention of systemic infections such as H5N1 or the vascular and renal damage caused by Stx, further structural optimization to achieve an acceptable therapeutic index and suitable bioavailability is required. Here, we describe peptidomimetic modifications of our lead PC inhibitor (Ac-RARRRKKRT-NH2) that result in increased stability with low toxicity. An azaβ3-amino acid scan revealed details of the structural requirements of the peptide. The modified peptide efficiently prevents HA5-mediated cell fusion

■

RESULTS Lead Peptide Properties. To optimize our lead PC peptide inhibitor (Ac-RARRRKKRT-NH2) for in vivo evaluation, three major aspects were considered: the inhibition potency (Ki), stability, and safety within the therapeutic range. This peptide, which is derived from the HA5 multibasic cleavage site, displays selective nanomolar inhibition toward PCs in the order of furin > PACE4 > PC5/6 > PC7.7,8 The stability of this inhibitor in the presence of cultured cells or in mouse plasma had not been previously determined. Thus, we evaluated the stability of this peptide in two assays. First, Figure 1A shows the stability of the peptide in cell culture conditions (±FBS) and in culture medium with FBS alone (in the absence of cells). Second, Figure 1B represents the peptide stability in an ex vivo plasma stability assay. Both assays indicated that the 30



Article

plasma stability from a half-life of 3.7 h to 5.6 h, 4.8 h, and 8.6 h for peptides 2, 3, and 4, respectively (Table 2). We hypothesized that additional peptidase activity could occur at P8 but at a slower rate, which would account for the degradation of the P1-modified peptide. The azaβ 3 -R modification at P8 (peptide 5) alone did not significantly alter the plasma half-life, but the addition of an azaβ3-R at P8 and P1 (peptide 6) resulted in increased stability from a half-life of 5.6 h for peptide 5 to 7.7 h for peptide 6. Because peptide 6 was one of the most stable modified peptides, with stability comparable to that of peptide 3, its degradation pattern after 6 h of incubation in cell culture was analyzed using mass spectrometry (Figure 3A). Peptide 6 demonstrated a clear reduction in the C-terminal degradation pattern, as indicated by the decreased intensity of the masses corresponding to ΔRT and ΔKKRT. The modification of P1 with ΔAgm (peptide 3) and Amba (peptide 4) also resulted in more potent inhibitors, with a decreased Ki from 25 nM to 6 nM and 1.9 nM, respectively. The azaβ3-R modification at P8 (peptide 5) alone decreased the Ki from 25 nM to 10 nM, but this change in Ki was not observed for peptides 3 (azaβ3-R at P1) and 6 (concomitant addition of azaβ3-R at P1 and P8), which demonstrated Ki values of 21 nM and 33 nM, respectively. The increased stability and potency of these peptides are interesting qualities from a therapeutic perspective because the amount of peptide required to achieve an effective dose in vivo would be reduced. The resulting modifications at P1, P8, or both positions resulted in a peptide with low toxicity. MTT assays indicated no significant differences in toxicity to HEK293FT cells among all of the modified peptides at concentrations of up to 150 μM, except for peptide 5 (P8 azaβ3-R), which demonstrated a nontoxic concentration of 100 μM (Table 2). The evaluation of acute cytotoxicity via LDH release indicated that all peptides exhibited less than 5% toxicity after 4 h of incubation at 100 μM (Table 2). The in vivo acute toxicity assays indicated that peptide 3 (P1 Amba) demonstrated a notable decrease of tolerability to a single dose by lowering the tolerable concentration from >10 mg/kg to 7−10 mg/kg. Although only peptides 2 and 6 (P1 and P1−P8 azaβ3-R, respectively) resulted in signs of toxicity at approximately 10 mg/kg, this toxicity was not lethal. Because of this toxicity, higher doses were not evaluated (Table 2). The ability of FITC-bound peptides to enter the cell and bind to the membrane was assayed using flow cytometry (Figure 3B−D). Peptide 8 and peptide 9 (P1 and P1−P8 azaβ3R, respectively, corresponding to peptide 2 and 6 bearing FITC) demonstrated similar membrane interactions and cell entry in comparison to the unmodified peptide. However, both peptide 10 and peptide 11 (P1 ΔAgm and Amba, respectively) resulted in a 2.7- and 4.7-fold increased cell penetration, respectively. These results correlate with the increased hydrophobicity of each compound (the calculated log P (http://www.chemicalize.org) of ΔAgm is −1.28 and that of Amba is −0.26) in comparison to Arg alone (a calculated log P of −3.16). Interestingly, the increased hydrophobicity also correlates with a decreased cell surface membrane signal, which was removed via trypsin washes (which digest the peptide and cell surface proteins). Although 17% of the total FITC signal originated from the membrane interaction on the cell surface, this signal was reduced 2-fold for peptide 10 (P1 ΔAgm), and no signal was detected from trypsin washes for peptide 11 (P1 Amba). These results strongly suggest that both peptides can

peptide was rapidly degraded with a half-life of 3.7 h in plasma, 8 h in cell culture with FBS, and 10 h in cell culture without FBS. Incubation in culture medium with FBS alone resulted in a degradation that was similar to that observed in plasma, with a 4 h half-life. An analysis of the degradation pattern of the peptide using mass spectrometry (Figure 1C) indicated that Cterminal peptidases were primarily responsible for the peptide degradation. After 6 h of incubation, a mass corresponding to the Ac-RARRKKR peptide (1167 m/z) was predominantly observed with second-order degradation masses corresponding to ΔRT (1112 m/z), ΔKRT (756 m/z), and ΔKKRT (635 m/ z) (Figure 1C). The toxicity of the peptide was previously assessed via the ATP release from RAW264.7 cells at concentrations of up to 100 μM.8 Toxicity was further evaluated by measuring long-term cytotoxicity via metabolic activity (MTT assay) after 24 h of incubation on a cell monolayer and by acute cytotoxicity via lactate dehydrogenase (LDH) release after incubation with cells for 4 h. HEK293FT cells were incubated with peptide concentrations of up to 150 μM without significant long-term toxicity (Figure 1D) and with peptide concentrations of up to 100 μM without significant LDH release (values below 5%) after 4 h of incubation (Figure 1E). The peptide exhibited low toxicity to other cell lines (i.e., A549 and MDCK cells) at concentrations of up to 100 μM (Supporting Information, Figure S1). This encouraging lowtoxicity profile agrees with the in vivo acute toxicity of the inhibitor. After intravenous injection of the peptide into mice, no signs of toxicity were observed at concentrations of up to 10 mg/kg body weight after a single peptide injection. Peptidomimetic Optimization of P1. In vivo studies indicated that the inhibitor stability required optimization to counter the C-terminal peptidase degradation while maintaining low toxicity and Ki values in the nanomolar range; therefore, the introduction of an Arg amino acid mimetic in P1 was a promising strategy. Three Arg mimetics were evaluated: azaβ3-Arg (azaβ3-R, peptide 2), 4-aminobut-2-en-1-ylguanidine (4-aminobutylguanidine, which is commonly known as agmatine, containing an alkene chain and abbreviated as ΔAgm, peptide 3), and 4-amidinobenzylamide (Amba, peptide 4) (Table 1 and Figure 2). The Amba and ΔAgm modifications represent decarboxylated arginine mimetics that could only be introduced at P1, which resulted in the loss of the P1′ threonine. All modifications at the P1 position increased the Table 1. List and Numbering of Peptide peptide number

peptide sequence

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Ac-RARRRKKRT-NH2 Ac-RARRRKK[azaβ3R]T-NH2 Ac-RARRRKK-ΔAgm Ac-RARRRKK-Amba Ac-[azaβ3R]ARRRKKRT-NH2 Ac-[azaβ3R]ARRRKK[azaβ3R]T-NH2 FITC-[βAla]-RARRRKKRT-NH2 FITC-[βAla]-RARRRKK[azaβ3R]T-NH2 FITC-[βAla]-[azaβ3R]ARRRKK[azaβ3R]T-NH2 FITC-[βAla]-RARRRKK-ΔAgm FITC-[βAla]-RARRRKK-Amba Ac-RA[azaβ3R]RRKKRT-NH2 Ac-RAR[azaβ3R]RKKRT-NH2 Ac-RARR[azaβ3R]KKRT-NH2 Ac-RARRR[azaβ3K]KRT-NH2 31



Article

Figure 2. Peptidomimetic modification of the lead peptide. Structure of the Ac-RARRRKKRT-NH2 peptide and the arginine mimetic used to optimize the peptide.

the peptide inhibitor. The replacement of P5 by P3 with azaβ3 analogues (peptides 13, 14, and 15) resulted in a gradual loss of inhibitory ability, with Ki values of 89 nM, 123 nM, and 591 nM, respectively. These results suggest that the disruption of the peptide backbone by azaβ3 analogues at these positions results in the improper orientation of the Arg and Lys side chains for enzyme inhibition. P2 was not evaluated because this position is known to be critical for substrate binding to furin.48,49 In Vitro Cleavage of the HA5-Spanning Peptide. On the basis of an alignment of the available sequences, the influenza HA5 multibasic cleavage site is present in highly diverse evolutionary forms. Thus, it was of interest to determine whether the evolution of this cleavage site can be differentially recognized by various PCs. Figure 4A shows the sources of three select cleavage sequences from HA5. The first sequence, which is denoted as Hong Kong, corresponds to the original HA5 sequence that was found in infected humans. This sequence influences HPAI infection in humans and is present in circulating clade 2.3.4, which has been found in humans. The second sequence, which is denoted as Vietnam, originates from avian HA5 clade 2.3.2, which has not been found in humans and which exhibits high similarity (200 >150 >150 >150 >100 >150

Optimization of Furin Inhibitors To Protect against the Activation of

Recommend Documents