Design and Structure–Activity Relationship of a Potent Furin Inhibitor Derived from Influenza Hemagglutinin

Abstract

Furin plays an important role in various pathological states, especially in bacterial and viral infections. A detailed understanding of the structural requirements for inhibitors targeting this enzyme is crucial to develop new therapeutic strategies in infectious diseases, including an urgent unmet need for SARS-CoV-2 infection. Previously, we have identified a potent furin inhibitor, peptide Ac-RARRRKKRT-NH₂ (CF1), based on the highly pathogenic avian influenza hemagglutinin. The goal of this study was to determine how its N-terminal part (the P8–P5 positions) affects its activity profile. To do so, the positional-scanning libraries of individual peptides modified at the selected positions with natural amino acids were generated. Subsequently, the best substitutions were combined together and/or replaced by unnatural residues to expand our investigations. The results reveal that the affinity of CF1 can be improved (2–2.5-fold) by substituting its P5 position with the small hydrophobic residues (Ile or Val) or a basic Lys.

The majority of peptides and proteins (e.g., hormones, growth factors, and enzymes) essential for proper cell functioning are synthesized as inactive precursors.¹ The proteolytic activation is a necessary step to transform them into mature forms. Secretory proteins that contain a characteristic multibasic motif R/K-X_n-R/K-R↓ (where ↓ is the cleavage site and X is any amino acid except for Cys, n = 0, 2, 4, or 6) are processed by mammalian subtilisin-like proprotein convertases (PCs).² This family of enzymes consists of nine members, namely furin, PC1/3, PC2, PC4, PACE4, PC5/6, PC7, SKI-1, and PCSK9. The seven first PCs share a unique domain arrangement with high degree of homology within the catalytic core (54–70%).³ The last two members, SKI-1 and PCSK9 have different substrate specificities and structure.^1,2

Furin is a 794-amino acid type I- transmembrane protein, which is ubiquitously expressed.² It circulates through the endosomal system, between the trans-Golgi network and the cell surface. This enzyme is essential for embryogenesis as it was shown that embryos of knockout mice with silenced fur gene die in 11 days after birth.⁴ However, liver-specific conditional furin knockout did not display any deviations in the processing of furin-associated substrates, indicating redundancy among members of the PC family.⁵ This observation suggests that inhibitors targeting furin might be used safely as local treatments for pathologies such as cancer and viral infections.⁵

Besides playing a critical role in numerous physiological processes, furin is also implicated in many pathological conditions including tumorigenesis,⁶ atherosclerosis,⁷ neurodegenerative disorders,⁸ and pathogenic infections.⁹ It is also required for the activation of viral envelope glycoproteins and bacterial protoxins.¹⁰ It was suggested that inhibition of this enzyme might lead to a new treatment option for these conditions.¹⁰ Therefore, a variety of furin inhibitors have been developed including peptides,¹¹ peptidomimetics,^10,12 proteins,¹³ and small molecules.¹⁴ In regard to peptide-based inhibitors, they have been discovered through screening of various synthetic libraries¹¹ or designed based on the sequence of natural substrates.¹⁵ Among them analogs with D-arginine residues displayed the most promising activity profile, including the ability to delay anthrax toxin-induced toxemia (hexa-D-arginine)¹⁶ or to reduce the corneal damage caused by Pseudomonas aeruginosa (nona-D-arginine).¹⁷

Furthermore, the modification of an extended furin cleavage site in the immature hemagglutinin (HA0) of avian influenza virus A led to the discovery of the sequence, TPRARRRKKRT, with potent inhibitory effect (inhibitory constant (K_i) of 23 nM).¹⁵ We have improved its activity profile by removing two N-terminal residues and by stabilizing both ends with an acetyl and an amide group, respectively (optimized sequence: Ac-RARRRKKRT-NH₂, compound CF1).¹⁵ The obtained peptide was able to block the processing of viral glycoproteins and anthrax toxin by a furin-dependent mechanism without significant cytotoxicity.¹⁵

On the other hand, several very promising peptidomimetic compounds have been developed by Steinmetzer’s group,^12,18,19 including one of the most potent inhibitors known today, namely 4-(guanidinomethyl)-phenylacetyl-Arg-Tle-Arg-4-amidinobenzylamide (MI-1148). This compound not only displays high affinity toward furin with K_i of 5.5 ± 0.3 pM but also shows potent activity against numerous pathogenic infections.¹⁸ Its remarkable potency was elucidated with the crystal structure of the furin–MI-1148 complex, which led to the identification of strong interactions between the enzyme’s acidic catalytic cleft and the P5–P1 positions of this compound.¹⁸ Specifically, it has been shown that 4-amidinobenzylamide (4-Amba) binds to the S1 pocket in a tight binding manner through hydrogen bonds and extensive van der Waals and electrostatic interactions.^18,19

With the aim to better understand the structural determinants for furin inhibition, we decided to investigate in detail the structure of our previously developed lead compound CF1. Since the furin recognition site of various inhibitors was extensively studied before,^18,20−22 we decided to focus our investigations on the P8–P5 positions. CF1 was used as a template to prepare the peptide libraries by replacing the selected residues one by one with the natural amino acids (except for Cys) (Figure 1A, the subsites P8–P1’ are assigned according to Schechter and Berger nomenclature²³). The obtained libraries consist of 72 new peptides (the P5, P6, P7, and P8 series, Figure 1B). In addition, we designed their counterparts having two substitutions (the P8–P5 series) or modified by unnatural or/and d-amino acid residues (the P5″ and the P8″–P5″ series) at the indicated positions (Figure 1C). Moreover, a cyclic version of CF1 (compound CF96, Figure 1D) was generated, leading to a total of 23 additional compounds. Their inhibitory activity was accessed toward soluble human recombinant furin using competitive kinetic assays (for details see the Supporting Information). We also tested the stability profiles of the selected compounds. The results of these studies revealed important details regarding the structural requirements for furin inhibition and led to compounds with improved potency.

Figure 1.

Designed series of peptides. (A) Structure of CF1 compound. Summary table of the introduced modifications, compound codes, and K_i’s for (B) the P5, P6, P7, and P8 series; (C) the P8–P5, P5″, and P8″–P5” series; and (D) the cyclic peptide and azaβ³-modified analog (& indicates cyclization according to commonly used nomenclature).

The unique codes (CF2–CF97) were assigned to each of the analogs obtained in the present study (Figure 1, Table S1). Compound CF1 was reassayed in the presented experiments, and its activity profile was comparable to the previously reported value (K_i of 15 ± 3 vs 25 ± 5 nM,²⁴ respectively).

The first series of compounds (Figure 2A, CF2–CF19) contains analogs modified at the P5 position. The incorporation of small hydrophobic, nonpolar residues at this location resulted in analogs with improved affinity toward furin. A similar observation was already reported in the study comparing the cleavage preferences of the individual PCs.²⁵ The best compounds, CF8 (K_i = 6 ± 1 nM), CF9 with a basic Lys (K_i = 7 ± 1 nM), and CF17 (K_i = 7 ± 2 nM), were approximately 2-fold more active than CF1. Additionally, analogs modified with Pro (CF13, K_i = 8 ± 3 nM), Ser (CF15, K_i = 9 ± 1 nM), and Thr (CF16, K_i = 8 ± 1 nM) showed an enhanced inhibitory profile. On the other hand, compounds containing residues with acidic (CF4, K_i = 55 ± 10 nM and CF3, K_i = 130 ± 10 nM) or aromatic side chains (CF7, K_i = 22 ± 9 nM; CF19, K_i = 35 ± 10 nM; CF18, K_i = 47 ± 10 nM; CF5, K_i = 21 ± 4 nM) were from 8.6 to 1.4-fold less potent than CF1. The remaining inhibitors displayed K_i’s in the range of 11–21 nM. The obtained analogs also showed a potent inhibitory effect against PACE4 and PC5/6, but not toward PC7, as shown by using the selected compounds with various chemical character (Table S2).

Figure 2.

K_i values of peptides from the series (A) P5, (B) P6, (C) P7, and (D) P8 against furin in comparison to CF1 (black bar). Amino acid residues were assigned with the following color code: the nonpolar (gray bars), polar (white bars), and aromatic (horizontal stripes, gray bars) negatively charged (dashed pattern bars) and positively charged (dotted pattern bars). The K_i’s are expressed as the mean ± SD of at least two independent experiments.

The second series of compounds (Figure 2B, CF20–CF37) substituted at the P6 position showed no improvement in activity in comparison to the P5-modified peptides. Only two residues, namely Leu and Lys, were well tolerated in this position instead of Arg presented in CF1. The resulting analogs CF28 and CF27 displayed comparable activity profile to CF1 with K_i of 14 ± 5 nM and K_i = 15 ± 6 nM, respectively. The remaining peptides showed reduced activity from 1.3- for CF32 to 18-fold for CF21, respectively. Similarly to the P5 series, compounds modified with acidic (CF22, K_i = 140 ± 27 nM, and CF21, K_i = 270 ± 36 nM) and bulky, aromatic residues (CF36, K_i = 97 ± 1 nM) displayed the weakest affinity toward furin.

All the P7-substituted analogs (Figure 2C, CF38–CF55) demonstrated lower affinity toward furin than CF1. Their activity pattern was comparable to the P5 and the P6 series; that is, the inhibitors with acidic (CF39, K_i = 24 ± 3 nM, and CF38, K_i = 33 ± 6 nM) and bulky, aromatic (CF54, K_i = 29 ± 1 nM) residues showed the lowest affinity toward furin. However, the least potent analog from this group was CF46 modified with Met residue (K_i = 38 ± 2 nM, 2.5-fold reduced activity). Nevertheless, the range of the K_i values for the P7 series (15–38 nM) differs significantly from analogs from the P6 series (14–270 nM) or the P5 series (6–130 nM), indicating that this position is less important for furin binding.

In contrast to the previous series, the substitutions of the P8 position had little or no effect on the activity of the resulting analogs. Most of the P8-modified peptides were slightly more potent when compared to CF1 (Figure 2D, CF56–CF73); for example. the best analog was CF68 with K_i of 10 ± 1 nM. Generally, only three analogs with aromatic (CF73, K_i = 18 ± 1 nM, and CF72, K_i = 30 ± 3 nM) and acidic (CF57, K_i = 20 ± 6 nM) residues showed slightly reduced activity. However, the drop in the potency was marginal when compared to analogs from the P5, P6, and P7 series. Even the incorporation of the acidic Glu residue did not affect the inhibitory potency of the obtained peptide (CF58, K_i = 11 ± 2 nM) as it was seen in the case of the P5, P6, and P7 series (resulting in a 3.6-, 9.3-, and 1.6-fold decrease in the inhibitory potency, respectively), suggesting that the P8 position is outside of the furin binding cleft. These observations are consistent with our previous studies regarding the other PC member, namely PACE4, where the P8-library screening showed that this position has minimal impact on the inhibitory activity of the leading compound.²⁶

The comparison of the obtained libraries revealed the general trend for structural determinants necessary for furin inhibition. We also confirmed that developed compounds, despite the presence of the potential furin cleavage sites, are inhibitors not weak substrates (Figure S1). For each series, the most potent inhibitors were obtained by the incorporation of residues with hydrophobic or positively charged side chains. For all of the screened positions (except for the P8-group), the significant drop in the activity was observed once the negatively charged amino acids were incorporated. In addition, the P5 and the P6 positions were considerably more sensitive to introduced changes than the P7 position, indicating that the residues close to the recognition motif (i.e., the P4–P1 positions) are more conserved. These observations are consistent with the reported crystal structures of furin in complex with various inhibitors revealing the presence of the high negative-charge density within its active site (up to S7 subsites) and thus explaining the strong preferences for basic compounds.^3,27 Our studies have shown that small hydrophobic residues at the P5 position, such as Ile, Val, or Pro, are also well-accepted. These results are in agreement with the previously reported data showing that inhibitors with short linear acyl groups at this position (up to 5 carbons) exhibit potent inhibitory effect against furin, whereas their counterparts with longer chains are significantly less active.²⁸ It is worthwhile to note that the potency of this series of compounds was considerably improved (approximately by 10-fold) once the acyl residue was additionally modified with the basic group,²⁸ emphasizing furin’s overall preference for positively charged residues.

Our screening studies revealed that only two positions of a control peptide, namely the P5 and the P8, are suitable for substitutions in order to yield more potent inhibitors. Therefore, next we decided to combine the most potent modifications from these series to verify whether their effect on furin inhibition could be additive (Figure 3). Ile at the P5 position was selected as a fixed residue since its introduction led to the most potent inhibitor CF8 among all of the tested compounds (2.5-fold improvement when compared to CF1). In the case of the P8 position, the incorporation of amino acids with different chemical character was well-accepted and resulted in more potent inhibitors than CF1. Therefore, we have chosen several residues with diverse properties and prepared the P8–P5 series of compounds (Figure 1C, CF74–CF80). Even though, all the compounds from this group showed stronger affinity toward furin than CF1 with K_i values running from 8 to 12 nM (Figure 3), their potency was not improved when compared to CF8 (K_i = 6 ± 1 nM), indicating that each residue within the scaffold affects how the peptide interacts with the catalytic pocket and even a small change can alter its overall binding profile.

Figure 3.

K_i values of analogs from the P8–P5 series toward furin in comparison to CF1 (black bar) and CF8 (dotted gray bar). The modifications are indicated above bars to facilitate the comparison. The K_i’s are expressed as the mean ± SD of at least two independent experiments.

Next, we investigated whether the incorporation of the unnatural amino acids in the P5 position could improve the binding affinity toward furin and ultimately lead to more stable compounds. This new series of peptides named the P5″ was designed by introducing Ile structural isomers such as Tle, cLeu, and Nle and residues with shorter linear (Abu) or Cα-branched (Aib) aliphatic side chains (Figure 4A,B; CF81–CF85). Only one compound from this group, namely CF81 with Abu displayed promising activity profile as it blocked furin activity slightly stronger than CF1 with K_i of 11 ± 3 nM; however, its affinity was lower than for CF8. Although it seems that the Ile residue at the P5 position is optimal for furin binding, the incorporation of unnatural residues might be beneficial to enhance the stability profile of the resulting peptides.

Figure 4.

Compounds with unnatural and/or d-amino acid residues. (A) Structures and names of unnatural amino acids used in the present study. K_i values of peptides from (B) the P5″ and (C) the P8″–P5″ series (gray bars, with the exception of CF93, which was indicated as a white bar) in comparison to CF1 (black bar) and CF8 (gray dotted bar). The modifications are indicated above bars to facilitate the comparison. The K_i’s are expressed as the mean ± SD of at least two independent experiments.

Indeed, it is well-known that peptides composed of natural amino acids suffer from poor stability in vivo due to rapid proteolysis. Therefore, we decided to test the impact of the introduced modifications on the stability profile of the resulting analogs, through in vitro assays, thus obtaining quick and relative values for each compound, as we have done previously.²⁴ To have a better understanding of stability requirements, we have prepared an additional series of peptides modified at the P5 and P8 position with unnatural amino acids and/or d-amino acids or both (the P8″–P5″ series, Figure 1C, CF86–CF95). The substitution of Ile residue at the P5 position with its d-isomer resulted in CF93 with significantly reduced activity toward furin when compared to CF1 (16-fold, K_i of 240 ± 110 nM, Figure 4C). These data are in agreement with our previous results regarding PACE4 inhibitors,²⁹ demonstrating that d-amino acids are not well-tolerated in close proximity to the enzyme’s recognition motif (i.e., the P4–P1 positions). In contrast, the incorporation of d-isomers at the P8 position did not seem to alter much the activity profile of the obtained peptides when compared to their l-isomer counterparts, e.g. for analogs containing Ser (CF69) or d-Ser (CF92) with K_i of 11 ± 2 vs 11 ± 1 nM and Pro (CF67) or d-Pro (CF91) with K_i of 10 ± 1 vs 13 ± 1 nM (Figure 4C). In addition, the unnatural amino acids were also well-tolerated at the P8 position (CF94 and CF95 with K_i of 6 ± 1 and 13 ± 1 nM, respectively), indicating, once again, that this position can accept a variety of substitutions without impacting the inhibitory effect of the resulting analogs.

In our previous work, the stability profile of CF1 was determined by high-performance liquid chromatography (HPLC),²⁴ that requires the use of a substantial amount of a compound (i.e., in μg/mL), which could inhibit proteases involved in the degradation, resulting in less accurate data. In the present work, we decided to employ a more sensitive method based on high-resolution mass spectrometry (HRMS, described in detail in the Supporting Information) and reevaluate the ex vivo plasma half-life (t_1/2) of our control peptide (CF1) using the ng/mL concentration range. As previously, the stability evaluation was performed in mouse plasma by incubating the peptide at 37 °C at different time points. The obtained t_1/2 is considerably different (∼63-fold) from the previously reported value, i.e., 3.54 min vs 3.7 h,²⁴ and in our opinion, it better represents the actual situation (Figure 5A). However, comparing the data can only be done in a relative manner as intra-assays. Two potent analogs with mixed modifications from the P8–P5 and P8″–P5″ series, namely CF75 and CF89, were selected in order to assess their stability profile. Both compounds turned out to be less stable than CF1. The t_1/2 of an analog with natural residues (CF75) was estimated to be 0.74 min, whereas its counterpart with d-amino acid and an unnatural substitution (CF89) was only slightly more stable with t_1/2 of 1.11 min (Figure 5A). After 2 min in mouse plasma, less than 15% and 25% of the intact peptide remains for both analogs (i.e., CF75 and CF89, respectively; Figure 5B).

Figure 5.

Stability profile of the selected furin inhibitors. (A) Ex vivo plasma t_1/2. (B) Percentage of the intact peptide remaining at the different time points. Each time point was done in triplicate, and data are presented as the mean and mean ± SD.

We then examined the impact of the remaining positions (P8–P5) on the stability profile of our furin inhibitors. We selected the analogs modified with Phe at various positons, namely compounds CF5, CF23, CF40, and CF59 for our investigations. For comparison, we also included a peptide having Phe at the P1′ position (CF97). In the case of this compound, Arg residue at the P1 position was replaced by its mimetic (azaβ³-Arg, Figure 2S, Supporting Information), that should act as a stability enhancer based on our previous report (1.5-fold improvement of plasma t_1/2 when compared to CF1)²⁴. Our data indicate that the presence of Phe in the N-terminal part of peptide (i.e., the P8 – P5 positions) is highly unfavorable for the stability of the resulting peptides, especially in regard to the P5 and P6 positions resulting in analogs having t_1/2 of 0.32 and 0.21 min, respectively (Figure 6). Interestingly, in our previous study, the substitution of these positions with azaβ³-Arg also led to considerably (∼3-fold) less stable compounds,²⁴ indicating that this part of the structure contributes significantly to the overall stability of our furin inhibitors. Indeed, it seems that the Arg residue in these positions makes the peptide chain less exposed and accessible for endoproteases when compared to compounds containing Phe or azaβ³-Arg. Similarly, the Arg residue at the N-terminal end seems to have a stronger protective effect against proteolytic degradation than Phe, Pro, or D-Pro (Figure 5 and 6). In regard to an analog with the Phe residue at the P1′ positions (CF97), its t_1/2 was only slightly reduced when compared to CF1 (t_1/2 of 2.59 vs 3.54, accordingly). This is most likely due to the presence of azaβ³-Arg at the P1 position.

Figure 6.

Ex vivo plasma t_1/2 of analogs modified with a Phe residue. Each time point was done in triplicate, and data are presented as the mean.

Our previous data suggested that the modification of the C-terminal end of furin inhibitors, especially by replacing Arg at the P1 position with its decarboxylated analog (i.e., 4-Amba), is the most effective strategy to prevent their proteolytic degradation.²⁴ The present work supports these observations, since the modification of the N-terminal part of CF1 even with unnatural residues (CF89) did not yield a more stable peptide. Nevertheless, the obtained data expanded our understanding of the stability requirements, indicating that the impact of the P5 and P6 positions (probably in combination with the C-terminal modification) should be further explored.

In addition, with the objective of developing more stable inhibitors, we have also prepared the cyclic version of CF1 using a side-chain-to-tail approach by forming a peptide bond between the side chain of Lys at the P5 position and the P1′ backbone COOH. The obtained compound CF96 displayed dramatically reduced activity toward furin with a K_i value of 2500 ± 300 nM (167-fold drop when compared to a control peptide). These data suggest that the link between the P5 and the P1′ positions creates a steric conflict preventing the resulting peptide from binding efficiently with furin catalytic pocket. The other parts of the molecule might be, however, more suitable for the cyclization as demonstrated by our previous studies regarding PACE4 inhibitors³⁰ or the recent work on macrocyclic furin inhibitors.³¹ Both reports showed that potent compounds (in low nM range) can be generated using head-to-side-chain cyclization (e.g., by linking the P5/P6 backbone NH and the P3 side chain in the case of furin inhibitors), suggesting that it is a promising direction for further exploration.

Taken together, our data provide a better understanding of structural requirements for furin inhibitors. We have shown that the P8 and P5 positions of CF1 can be easily modified to generate more potent compounds. Specifically, small hydrophobic amino acids are preferred in the P5 position, whereas almost any residue is well-accepted in the P8 position (Figure 2A and D), indicating that the N-terminal end of the peptide is outside of the furin catalytic pocket. Even though the introduced modification did not yield more stable compounds, the important data has been collected from these studies. For example, we have identified the positions (i.e., P5 and P6) that are more prone to the proteolytic degradation and showed that the residues originally present in CF1 offer a better protective effect. The results presented here can serve as a solid foundation for further studies aiming to obtain more potent and stable furin inhibitors.

There is some urgency to develop furin inhibitors as potential antiviral agents. Previous work on different virus families have shown that the acquisition of a multibasic furin cleavage site results in increased pathogenicity.³² Indeed, the recent coronavirus SARS-CoV-2 outbreak causing acute infection of the respiratory tract (COVID-19) can serve as an example of how a coronavirus can become a pandemic threat, most likely due to the presence of a furin cleavage site in the spike protein (S-protein).³³ All viruses use opportunistic means to enter cells, through the use of host receptor binding proteins and/or host proteases. However, it seems that furin involvement in proteolytic activation of viral fusion proteins supports spread of the infection. For the COVID-19 pandemic, newly developed vaccines are an ultimate solution; however, the need for effective antiviral drugs directed toward enzymes determining viral infectivity and pathogenicity (such as furin)^33,34 remains essential.

Glossary

Abbreviations

4-Amba

4-amidinobenzylamine

Abu

l-2-aminobutyric acid

Aib

2-aminoisobutyric acid

cLeu

1-amino-cyclopentane-1-carboxylic acid

HPLC

high-performance liquid chromatography

HRMS

high resolution mass spectrometry

K_i

inhibitory constant

MI-1148

4-(guanidinomethyl)-phenylacetyl-Arg-Tle-Arg-4-amidinobenzylamide

Nle

l-norleucine

t_1/2

half-life; Tle, l-tert-leucine.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.0c00386.

• Full experimental details, analytical data of all the compounds, K_i values for the selected PCs, enzyme and substrate concentrations, cleavage experiment and CF97 structure. (PDF)

Author Contributions

^⊥ M.L.G. and A.K. contributed equally. M.L.G., A.K., and T.Ł., peptides design and data analysis; M.L.G., preparation of peptide libraries; A.K., enzyme kinetic and plasma stability studies; T.Ł., preparation of the selected peptides and kinetic studies; P.N., synthesis of control peptides; K.L., optimization of the plasma stability assay and data compilation; R.D., design of the project; A.P., supervision of the project; Y.L.D., design of synthesis protocols; M.L.G. and A.K., manuscript preparation; A.K. and R.D., manuscript corrections. All authors reviewed the manuscript.

The authors declare the following competing financial interest(s): Robert Day is a stakeholder in Phenoswitch Bioscience Inc.