2-Cyanoisonicotinamide Conjugation: A Facile Approach to Generate Potent Peptide Inhibitors of the Zika Virus Protease

Abstract

The rapid generation and modification of macrocyclic peptides in medicinal chemistry is an ever-growing area that can present various synthetic challenges. The reaction between N-terminal cysteine and 2-cyanoisonicotinamide is a new biocompatible click reaction that allows rapid access to macrocyclic peptides. Importantly, 2-cyanoisonicotinamide can be attached to different linkers directly during solid-phase peptide synthesis. The synthesis involves only commercially available precursors, allowing for a fully automated process. We demonstrate the approach for four cyclic peptide ligands of the Zika virus protease NS2B-NS3. Although all peptides display the substrate recognition motif, the activity strongly depends on the linker length, with the shortest cyclization linker corresponding to highest activity (K_i = 0.64 μM). The most active cyclic peptide displays affinity 78 times higher than that of its linear analogue. We solved a crystal structure of the proteolytically cleaved ligand and synthesized it by applying the presented chemistry to peptide ligation.

Combining the best attributes of small molecules and antibodies, macrocyclic peptides are promising next-generation ligands for drug discovery.¹ Constraining a peptide ligand by cyclization can preorganize its binding interactions and thus reduce the entropic penalty and improve crucial pharmacokinetic parameters like bioavailability and metabolic stability.² Despite these great advantages for drug discovery, macrocyclic peptides remain underexplored as lead compounds in medicinal chemistry, which may relate to a lack of facile and broadly applicable synthetic methods to rapidly access macrocyclic peptide derivatives. Fundamental changes in synthetic methodology have always impacted the character of lead compounds pursued by medicinal chemists.³ Thus, advances in peptide chemistry shall further enhance the consideration of constrained peptides in drug discovery campaigns.

While Fmoc solid-phase peptide synthesis has revolutionized access to linear peptides by affordable automated synthesis on demand,⁴ selective modifications such as cyclization remain more exclusive.⁵⁻⁷ Recently, we developed a biocompatible method to rapidly cyclize peptides using 2-cyanopyridine and 1,2-aminothiol functional groups, which has been proven to be fully compatible with all canonical amino acids, including nonterminal cysteine residues.^8,9 However, the synthesis of the core amino acid 3-(2-cyano-4-pyridyl)-alanine (Cpa) requires advanced synthetic chemistry set-ups, involving the handling of potentially explosive peroxides, toxic trimethylsilyl cyanide, and carcinogenic dimethylcarbamoyl chloride.⁹ Here, we report the selective reaction between 2-cyanoisonicotinamide (CINA) and N-terminal cysteine as an alternative strategy for rapid peptide macrocyclization fully amenable to automation. In contrast to previous methods, CINA can be installed directly on the peptide chain during automated solid-phase synthesis from commercially available precursors, overcoming the necessity for advanced synthetic laboratory setups (Scheme 1). In addition, this strategy also allows for the installation of different cyclization linkers without additional effort (Scheme 1).

Scheme 1. Synthesis of Linear Peptides 1a–4a and Constrained Cyclic Analogues 1b–4b.

Isolated yields and LC-MS chromatograms (214 nm; retention time in minutes) are reported for the cyclization of 1a–4a to 1b–4b. (a) Standard SPPS in the order Fmoc-AA(ivDde)-OH, Fmoc-Lys(Boc)-OH, Fmoc-Arg(Pbf)-OH, Fmoc-Lys(Boc)-OH, Fmoc-Gly-OH, Boc-Cys(Trt)-OH. (b) Piperidine, DMF. (c) Hydrazine hydrate, DMF. (d) 2-Cyanoisonicotinic acid (CINA), HCTU, DIPEA, DMF. (e) TFA/TIPS/DODT/H₂O (94:2:2:2); DODT, 2,2′-(ethylenedioxy)diethanethiol. (f) PBS pH 7.5, 1 mM TCEP, 30 min.

We demonstrate the approach for cyclic peptide inhibitors of the Zika virus protease NS2B-NS3 (ZiPro). The Zika virus is a flavivirus closely related to other health-threatening pathogens of the same genus like dengue and West Nile viruses.¹⁰ During the unprecedented outbreak in Latin America in 2016, the WHO declared Zika a public health emergency of international concern and called for intensified research and development efforts. Most Zika infections are asymptomatic. Major health concerns during the 2015/2016 epidemic arose from congenital malformations such as microcephaly and an increased risk of fetal loss associated with Zika infections during pregnancy.^11,12 A direct link between Zika infections and increased risk of the autoimmune Guillain-Barré syndrome is also concerning.¹³ No Zika virus vaccines or specific antivirals have been approved yet.¹⁴

Flaviviruses comprise a single-stranded RNA genome that is translated into a single polyprotein in the host cell. Proteolytic cleavage of this polyprotein by host cell proteases and the viral protease NS2B-NS3 into individual structural and nonstructural proteins is essential for viral replication.¹⁵ Thus, the serine protease NS2B-NS3 from flaviviruses has been suggested as a promising drug target for infections with Zika, dengue, West Nile, and other flaviviruses.^16,17 The protease’s high degree of conservation among flaviviruses may facilitate the discovery of pan-flaviviral drug candidates. The full protease complex comprises NS3 bearing the catalytic triad (serine, histidine, aspartate) and the small cofactor NS2B. The C-terminal domain of NS2B needs to wrap around the active site for sufficient substrate recognition and catalytic activity.¹⁸ This crucial interaction has informed the design of various NS2B-NS3 constructs for drug screenings, mainly differing in the way NS2B and NS3 are connected. In case of Zika, a commonly used covalently linked protease construct might favor the dissociation of the crucial C-terminal NS2B domain from NS3.^19,20 Therefore, a more natural unlinked construct was developed (referred to as bZiPro), which has been used in this study.²¹

Previously explored inhibitors of ZiPro were either small compounds or substrate-derived peptides.²² Drug-like small molecules suffered from low affinity, whereas simple substrate analogues displayed high affinity but limited activity in cellular assays. ZiPro recognizes charged basic amino acids (arginine, lysine) in P₁ and P₂, which has further challenged the discovery of drug-like peptidomimetics.²² Cyclic peptides are one promising avenue to improve pharmacokinetic parameters as well as the affinity of linear substrate analogues. Recent studies highlighted that cyclic peptides can indeed generate high-affinity ligands of the active and allosteric sites of ZiPro.^9,23,24 However, these cyclic analogues require multistep synthetic procedures, limiting extensive structure–activity relationships. In this study, we show that high-affinity cyclic peptide inhibitors of ZiPro can be generated by fully automated solid-phase peptide synthesis.

Our design of constrained ZiPro inhibitors focused on the nonprime site substrate recognition sequence GKRK of ZiPro spanning residues P₄ to P₁ (Table 1).^8,9,22 We installed a cysteine at the N-terminus and the CINA motif at the C-terminus of the peptide (Scheme 1). To study the effect of macrocyclic ring size on the enzyme inhibition, we designed four analogues 1b–4b with different CINA linkers, containing either Dap, Dab, Orn, or Lys (Scheme 1). We chose commercially available derivatives of Dap, Dab, Orn, and Lys bearing N-terminal Fmoc and side-chain ivDde protection groups. Employing the more stable ivDde protection group instead of conventional Dde allows for orthogonal side-chain deprotection using 3% hydrazine with minimal intramolecular migration during solid-phase peptide synthesis.^25,26 The linear precursor peptides 1a–4a were obtained by sequential standard solid-phase peptide synthesis, followed by selective removal of ivDde and subsequent coupling to 2-cyanoisonicotinic acid (CINA) on the solid support (Scheme 1). Inspired by our recent efforts,^8,9,27 we developed a facile and biocompatible cyclization protocol to generate constrained peptides 1b–4b from linear precursors 1a–4a (Scheme 1) directly in aqueous buffer within 30 min. We observed that the macrocyclic ring size did not affect the efficiency of cyclization and compounds 1b–4b were obtained in excellent isolated yields between 53 and 74% (estimated conversions determined by HPLC are 76–92%, Table S1).

Table 1. Inhibitors of the Zika Virus Protease NS2B-NS3.

^aPeptides cyclized or ligated from their linear precursors 1a–5a and 5b.

^bCINA = 2-cyanoisonicotinamide.

^cZika virus NS2B-NS3 protease (bZiPro) inhibition. Substrate: Bz-Nle-Lys-Lys-Arg-AMC. Inhibition constants (K_i) were calculated from measurements at three different substrate concentrations.

We assessed the inhibition potential of compounds 1b–4b using the unlinked construct bZiPro and the fluorescent substrate Bz-Nle-KKR-AMC (Nle, norleucine).²⁸ We monitored dose–response curves at three different substrate concentrations and calculated inhibition constants (K_i) from three IC₅₀ values following the Cheng–Prusoff relationship (Figures S15–S18).²⁹ Unsurprisingly, all cyclic peptides 1b–4b bearing the substrate recognition motif GKRK are competitive inhibitors of ZiPro (Table 1). However, we observed a remarkable structure–activity relationship with regard to the chosen cyclization linker. For instance, compound 1b with a Dap-based linker inhibited more than 20 times stronger than analogue 2b with a Dab-based linker. Remarkably, both compounds differ only in one methylene group, clearly highlighting the importance of the cyclization linker for preorganization, which has been neglected in previous studies.⁹ Derivative 1b with the shortest linker displays significant affinity with a K_i of 0.64 μM. A Dixon plot for compound 1b confirms a competitive inhibition mode (Figure S20).

It has been suggested that most proteases recognize extended β-strands in their active sites.^30,31 Our small SAR may support previous observations that macrocyclization of substrate-derived inhibitors can preorganize the peptide in the extended conformation and thus reduce the entropic penalty.³² In addition, our study suggests that in order to maximize affinity, the right linker geometry is crucial. In our SAR, the most constrained peptide with the shortest linker gave the best results. It should be noted that Dap used in 1b is an isostere of serine, which is recognized by ZiPro in P₁′,²² potentially contributing to the outstanding affinity of 1b.

Encouraged by these results, we set out to solve the crystal structure of 1b in complex with bZiPro to better understand the SAR and the pronounced specificity for the Dap-based linker. Instead of 1b, we observed the proteolytically digested linear derivative 5 in complex with bZiPro (Figure 1). When compound 1b was built and refined (Figure S23), the resulting electron density maps indicate that the cyclic peptide had been cleaved during the crystallization process yielding the complex structure of bZipro and compound 5. Derivative 5 represents the selective hydrolysis product between P₁ Lys and P₁′ Dap(CINA) of 1b, yet again indicating that the Dap-based residue might have the capability to act as a serine isostere in P₁′.

Figure 1.

Macrocyclic peptide 1b interacting with the Zika virus protease NS2B-NS3 (ZiPro). (a) Scheme of proteolytic digest of 1b to 5 by bZiPro. The cleavage site is highlighted in blue. (b) LC-MS chromatogram (254 nm) and corresponding mass spectra after 8 h of incubation of 1b in the presence of 0.1% bZiPro in 10 mM Tris-HCl pH 8.5, indicating partial proteolysis to 5. (c) Crystal structure of bZiPro in complex with 5 at a resolution of 1.9 Å (PDB code: 7DOC). NS2B (magenta) and NS3 (yellow) are shown as cartoon representations, and compound 5 (green) is represented as sticks. The bZiPro residues involved in interactions with 5 are labeled and shown as sticks.

To confirm that the stronger inhibition of 1b is driven by the cyclic peptide and not by the proteolysis product 5, we analyzed the proteolytic stability of 1b in the presence of 0.1% bZiPro. After 8 h of incubation, we observe only minor cleavage product 5 (Figure 1b), clearly indicating that compound 1b is the active species in the ZiPro inhibition assay relating to a K_i value of 0.64 μM. The observation of 5 in the crystal structure is likely due to the high protease concentration and prolonged incubation times under cocrystallization conditions, favoring proteolytic digest. This effect has previously been observed with a related cyclic peptide inhibitor of ZiPro.⁹ The very slow proteolysis compared to the linear substrate analogue Bz-Nle-KKR-AMC used in the assay also highlights that peptide cyclization can not only increase affinity but also guard against proteolytic digest.

In the crystal structure, four bZiPro molecules were observed in a single asymmetric unit. One was occupied by 5, while another one was bound to residues K14, K15, G16, and E17 from the neighboring NS3 molecule in reverse orientation, occupying the S₄ to S₁ pockets of bZiPro. Comparison of the complex with previously reported bZiPro–ligand cocrystal structures containing substrate-derived ligands reveal similar P₁ and P₂ interactions and P₁–P₃ backbone conformations (Figure S22). Similar to a structure published previously,⁹ the P₃ lysine side chain of 5 is flipped, while the rest of the bulky main chain exits bZiPro at a nearby hydrophobic groove.

To further investigate the relevance of the proteolysis product, we synthesized ligand 5 through a CINA ligation strategy (Scheme 2). Similar to the CINA cyclization approach, both precursor fragments 5a and 5b were accessed from standard solid-phase peptide synthesis. To fully account for the different terminal amide and acid groups in 5, precursor fragment 5a was assembled on Rink amide resin, while the precursor fragment 5b was assembled on 2-chlorotrityl resin. Unlike cyclization, the ligation was performed for 3 h with crude linear fragments 5a and 5b in aqueous PBS (pH 7.4) containing 1 mM tris(2-carboxyethyl) phosphine (TCEP) (Scheme 2). Ligation product 5 was obtained with an overall isolated yield of 36% (estimated conversion determined by HPLC is 45%, Table S1), demonstrating the effectiveness of bimolecular ligation between CINA and N-terminal cysteine.

Scheme 2. Synthesis of 5 Using Peptide Ligation.

Isolated yield indicated. (a) Fmoc-Dap(ivDde)-OH, HCTU, DIPEA, DMF. (b) Piperidine, DMF. (c) Boc₂O, DIPEA, DMF. (d) Hydrazine hydrate, DMF. (e) 2-Cyanoisonicotinic acid (CINA), HCTU, DIPEA, DMF. (f) TFA/TIPS/DODT/H₂O (94:2:2:2); DODT, 2,2′-(ethylenedioxy)diethanethiol. (g) Fmoc-Lys(Boc)-OH, DIPEA, DMF. (h) Standard SPPS in the order Fmoc-Arg(Pbf)-OH, Fmoc-Lys(Boc)-OH, Fmoc-Gly-OH, Boc-Cys(Trt)-OH. (i) PBS pH 7.5, 1 mM TCEP, 3 h.

We assessed proteolysis product 5 for its ZiPro inhibition and found a K_i value of 50.2 μM, which is 78 times higher than that for the cyclic analogue 1b. This observation proves that cyclic peptide 1b is the active species and not the proteolysis product 5. It also confirms that 1b is almost stable against proteolysis under standard assay conditions and that proteolysis originates mainly from prolonged incubation time and high protease concentration required for crystallization.

The data presented in this study indicate a large affinity gap between cyclic compound 1b and linear analogue 5. Apart from the flipped orientation of P₃ lysine, the cocrystal structure between 5 and ZiPro displays the most important interactions between a substrate-derived ligand and ZiPro commonly observed in other cocrystal structures. Despite this, the inhibition constant of 5 is only 50 μM, which is 78 times higher than the inhibition constant of 1b. Although a model of ligand 1b with ZiPro indicates a more common and thus favorable interaction with P₃ lysine (Figure S23), a lower entropic barrier of binding might be an important contributor to the significantly higher affinity of 1b.

In summary, we present a cost-effective synthetic strategy to rapidly access macrocyclic peptides which is fully amenable to automation. The process produces peptides in high yield, allowing for direct structure–activity relationships without the need for purification. Utilizing CINA as a nontoxic, economical, and readily available building block may allow transformation to large scale industrial production of cyclic peptide ligands. CINA-based cyclization facilitates the screening of various ligation linkers, which is crucial to identify the optimal linker geometry. Using this technology, it required only a small number of compounds to identify a high-affinity ligand of the Zika virus protease. We are therefore confident that the strategy will be equally successful for alternative drug targets.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.0c00657.

• Experimental details for Zika protease construct and preparation, synthesis and screening of macrocyclic and linear peptides, molecular structure of all peptides, LC-MS data and chromatograms of all peptides, inhibition assay, dose–response curves, Dixon plot, cocrystallization protocol, proteolytic stability assay, X-ray data collection and refinement statistics, description of interactions of 5 and bZiPro, Figures S1–S23, and Tables S1 and S2 (PDF)

Accession Codes

Atomic coordinates of the bZiPro:5 cocrystal structure have been deposited in the Protein Data Bank (PDB) under the accession code 7DOC.

Author Contributions

^# J.-P.Q and B.S. contributed equally.

The authors declare no competing financial interest.