Identification and In Silico Binding Study of a Highly Potent DENV NS2B-NS3 Covalent Inhibitor

Abstract

Dengue virus (DENV), an arthropod-borne flavivirus, has developed rapidly in the past few decades and becoming the most widespread arbovirus in the world. The vital role of NS2B-NS3 in virus replication and maturation of viral proteins makes it the most promising target for anti-DENV drug discovery. In the current work, a potent NS2B-NS3 covalent inhibitor 23 (IC₅₀ = 6.0 nM, k_inac/K_i = 1581 M^–1 s^–1) was discovered through the chemical modification of a published covalent inhibitor 1 (IC₅₀ = 500 nM, k_inac/K_i = 156.1 M^–1 s^–1), followed by in vitro assay. Further comprehensive structure–activity relationship analysis through covalent docking and molecular dynamics simulation provides informative understanding of the binding modes of covalent inhibitors targeting NS2B-NS3.

Dengue virus (DENV) belongs to the family of Flaviviridae, which is a mosquito-borne, single positive-stranded RNA virus.¹ Dengue fever is an infectious disease caused by DENV, which spreads widely in Southeast Asia, Central, and South America, where Aedes mosquitoes are active.²⁻⁴ In severe cases, infection with DENV can also cause hemorrhagic fever and dengue shock syndrome.^5,6 Current estimates indicate that about 400 million people are infected by DENV, and about 4000 deaths occur globally each year.^5,7 In fact, the infection with DENV is becoming a major threat to the workforce and the personal health of citizens all over the world.⁸ In recent years, intensive efforts have been conducted to discover novel antivirals. For example, a vaccine, namely, Dengvaxia, has been licensed, though its safety and efficacy against different DENVs remain unclear.⁹ Moreover, to the best of our knowledge, there is no other vaccine or effective small molecular drug available in the market for the prevention or treatment of dengue fever. Therefore, the development of antiviral therapy is a crucial task for DENV treatment now.

DENV RNA contains ∼10 800 nucleotides, which encodes a viral polyprotein.⁷ The NS2B-NS3 protease cleaves the viral polyprotein and plays an important role in viral replication, possessing the potential to be a promising drug target.^7,10 DENV NS2B-NS3 protease is a serine proteinase, which is responsible for cleaving viral polyprotein precursors to several smaller and functional proteins including three structural proteins and seven nonstructural proteins.¹¹ NS3 consists of an N-terminal serine protease domain and a helicase, with NS2B required to construct an active enzyme.^12,13 The crystal structure of the NS2B-NS3 complex¹⁴ reveals the protease active site composed of His51, Asp75, and Ser135 from NS3, and the central hydrophobic domain of NS2B.

Most of the NS2B-NS3 protease inhibitors bind either in the flat active site or the adjacent allosteric pocket.¹⁵⁻²⁰ Compared to the competitive inhibitors that are difficult to form high-affinity binding interactions within the active site, covalent inhibitors attract more attention.^21,22 Compound 1 was reported to be a moderate inhibitor of NS2B-NS3 protease toward the catalytic triad of protein with a promising IC₅₀ value of 500 nM.²³ On the basis of the structure of compound 1, we ultimately identified novel and potent NS2B-NS3 protease inhibitors through the chemical modification of the two benzene rings of compound 1, enhancing inhibitory activities to a new level.

NS2B-NS3 covalent inhibitors bind within the active site, forming a covalent bond with catalytic residue Ser135, which hydrolyzed the inhibitor into two fragments. The benzoyl moiety links to the side chain of Ser135 through a covalent bond, and the rest of the compound is released into the medium. The core pyrazole ester motif of the compound plays an important role in compound-protein covalent interaction, which was kept as a covalent warhead to interact with the protease catalytic triad. Meanwhile, it is demonstrated that the intact form of the compound is critical for interaction;²³ therefore, chemical modification was performed on the benzene ring that may improve the inhibitory activity.

In order to achieve better shape complementation within the binding site, various substitutions with different sizes were introduced to occupy the universal flat site. Moreover, different substitutions with diverse electronic properties were introduced into the phenyl part of compounds to address the electrical properties of core covalent warheads by inductive effect and conjugated effect. Finally, 27 novel compounds were designed and synthesized for biological evaluation.

In search of compounds with improved anti-DENV NS2B-NS3 activity, analogues of lead compound 1 with pyrazole ester moiety were obtained through a facile convergent synthetic route, as shown in Scheme 1. Cyanoacetohydrazide was added with purchased or synthesized benzenesulfonyl chlorides in a solvent of EtOH to yield intermediate 4. To prepare commercially unavailable benzenesulfonyl chlorides, the sulfo-group was introduced into the 4-position of substituted benzene before sulfonyl-chlorination. Then, the intramolecular cyclization of 4 occurred with a condition of NaOH, followed by neutralization and precipitation to obtain the key intermediate 6, which carries the important pyrazole ester moiety. On the other hand, benzoyl chloride intermediate 8 was generated by acyl chlorination from corresponding benzoic acid. Finally, compounds 1, 9–17, and 22–30 without an amino group at R₂ were prepared by esterification reaction of 6 and 8. Moreover, compounds 18–21 and 31–35 carrying an amino group at R₂ were synthesized through an additional reduction step in the presence of H₂, using Pd/C as the catalyst. The proposed synthetic route provides us convenient access to analogues, and detailed procedures are described in the Supporting Information. All newly synthesized compounds were characterized by ESI mass spectrometrty, ¹H NMR, and ¹³C NMR.

Scheme 1. General Synthesis for Compounds.

Reagents and conditions: (i) DCM, 0 °C, 4 h; (ii) SOCl₂, DMF (cat.), DCM, reflux for 6 h; (iii) cyanoacetohydrazide, EtOH, rt, 2 h; (iv) 1 M NaOH (aq.), rt, 30 min; (v) 3 M HCl (aq.), rt, 10 min; (vi) SOCl₂, DMF (cat.), DCM, reflux for 2 h; (vii) DIPEA, DCM, rt, 10 h, (viii) H₂, Pd/C, ethyl acetate, rt, 12 h.

The activities of synthetic derivatives were determined according to the fluorescence intensity from fluorescent products generated by cutting substrate benzoyl-Nle-Lys-Arg-Arg-AMC in a fluorometric assay. Aprotinin and lead compound 1 were used as positive controls. The inhibition rates (%) of compounds were determined from triplicate measurements at the concentration of 150 nM (as shown in Table 1). The IC₅₀ (half-maximal inhibitory concentration) values were further determined for the compounds with a promising inhibition rates in a similar method.

Table 1. Structures and Activities of Compounds 1, 9–21.

^aCompound 1 is the lead compound from the literature.

The inhibition rate (%) of lead compound 1 was 59.67% in this test condition, generally consistent with the reported data.²³ As disclosed from Table 1, changing the benzene ring at R₂ into an amine and without any modification at R₁ (compound 21) exhibited a slight improvement of the inhibitory activity, while compounds with a CF₃ at R₁ showed poor biological activities, suggesting that an electron-withdrawing group (EWG) is disfavored at this position. But a chlorine atom at R₁, which is also an EWG, could improve inhibitory activity toward NS2B-NS3 (compound 19), probably due to the particular interaction between halogen and protein. Compounds 10,12,15,16 and 20 containing a phenoxy group at R₁ were observed to possess relatively high inhibition rate (%) values, particularly compound 20 with an amine at R₂ showed a 100% inhibition at test concentration, which is speculated that the electronic donation from oxygen atom or the spacing effect of benzene could enhance compound potency. In order to find more potent inhibitors, 14 compounds were further synthesized, with the inhibition rate (%) values shown in Table 2. It is found replacement of chlorine in compound 19 with fluorine led to an activity loss (compound 31), but placing bromine at R₁ and keeping the same R₂ yielded an inhibitor with 100% inhibition rate (compound 32), indicating that a strong EWG may cause a loss in activity but a relatively weak one like a bromine is tolerable. A surprising 100% inhibition rate was shown on compound 22 which contains a hydrogen atom at R₁ and benzene at R₂, and addition of ethoxy group at R₁ leads to another potent compound 23 with 100% inhibition rate, whereas a tert-butyl group in the R₁ position did not show any improvement of activity (compounds 24, 28, 35).

Table 2. Structures and Activities of Compounds 22–35.

Further analysis indicates that the effect of R₁ on inhibitory activity relies on the functional group at R₂: when the substituent at R₂ is a bulky group, electronic factors of R₁ are more important than steric factors, and EWGs can cause a loss in activity; when the substituent at R₂ is a small group like amine, the inhibitory activities usually depend on the size of substituents at R₁ and a relatively large group is able to enhance inhibitory potency. However, the bulky phenoxyl does not show a promising activity in the fluorometric assay, probably due to the unfavorable spatial interaction between protease and compounds.

In order to find the most potent inhibitor, we determined the IC₅₀ values of some compounds possessing satisfactory inhibition rates at different compound concentrations. As shown in Table 3, Compounds 20 and 32, containing the same amine at R₂ and different R₁ substitutions, were found to have IC₅₀ values of 9.0 nM and 74 nM, respectively. Compound 22 with benzene at R₂ and unsubstituted R₁ had an IC₅₀ value of 12 nM, and its analogue 23 with the ethoxyl group at R₂ exhibited an IC₅₀ value of 6.0 nM, which marks the importance of EDGs (electron donating group) at this position.

Table 3. IC₅₀ Values of Potent Inhibitors.

Compd	1a	20	22	23	32
IC50 (nM)	500	9.0	12	6.0	74

^aCompound 1 is the lead compound.

Enzyme kinetics of inhibitors were also conducted to reveal the kinetic features of the inhibition process. The same time-dependent inhibition was observed both in compounds 1 and 23 (Figure 1A, C), allowing us to infer the same covalent inhibition mechanism of compound 23 as the reported covalent inhibitor 1.²³ Meanwhile, the second-order rate constant (k_inact/K_i) values of compounds 1 and 23 were determined to describe the efficiency of the inhibition reaction. Compound 23, with a k_inact/K_i value of 1585 M^–1 s^–1, is significantly more potent than compound 1 which has a k_inact/K_i value of 156.1 M^–1 s^–1, indicating that compound 23 is able to form an adduct with the enzyme, and inactivate the protease at a much faster rate.

Figure 1.

k_inact/K_i Values of compounds 1 and 23 on DENV NS2B-NS3 protease. The time-dependent inhibition of protease at different concentrations of compound 1 (A) and compound 23 (C). The k_obs values of the two compounds were plotted against inhibitor concentration. The k_inact/K_i value of individual compounds was determined for compound 1 (B) and compound 23 (D). (E) The k_inact/K_i values for different compounds. The values are presented as means ± SEM.

After the above process, compound 23 was identified to be the most potent inhibitor, which was further undergone cytotoxicity assay and cell protection test. In these tests, BHK-21 cell was chosen as the host cell for evaluating the antiviral efficacy of compound 23 due to the high infection rate by DENV serotype 2. The cellular toxicity and the protective effect were tested at a compound concentration of 5 μM. The CCK-8 assay was used to assess the protective effect of compound 23 on the viability of infected BHK-21 cells after DENV infection. In toxicity assay, compound 23 showed no cytotoxicity as shown in Figure 2A. Additionally, compound 23 markedly increased the cell survival rate in the cells protective effect test (Figure 2B), which is about 1.3 times the virus infection control rate, suggesting 23 as a novel structure possessing an amazing protective effect without observed cytotoxicity. These results are generally consistent with the above fluorescence assays for DENV NS2B-NS3 protease. Overall, through the design, synthesis and evaluation process described above, compound 23 is identified as a promising novel NS2B-NS3 inhibitor at nanomolar level potency.

Figure 2.

Cellular toxicity and protective effect of compound 23. (A) Comparison of the cell survival rates among normal cells (Control) and compounds treated cells shows these compounds have no cellular toxicity. (B) Treatment with compound 23 significantly increased the survival rate of DENV-transfected cells. **P < 0.01, ***P < 0.001.

To better understand the interaction modes between these compounds within NS2B-NS3 active site, covalent docking, molecular dynamics (MD) simulation, and other computational methods were utilized to gain insight into ligand–protein interaction. First, due to the necessity of validating the reliability of the molecular docking results, the lead compound 1 was docked into the active site. As shown in Figure 3A, the docked conformation of compound 1 forms similar interactions with conserved amino acids,²³ indicating that the interactions of these inhibitors can be modeled by CovDock with reactive residue setting as Ser135. Thus, through this validated covalent docking protocol, the molecules were docked into the active pocket of DENV NS2B-NS3 protease (PDB ID: 3U1I).

Figure 3.

Putative binding modes of lead compound 1 and compound 23. (A) 2D binding mode of compound 1. (B) Overall binding pose of compound 1 within NS2B-NS3 pocket. (C) 3D binding mode of compound 1. (D) 2D binding mode of compound 23. (E) Overall binding pose of compound 23 within NS2B-NS3 pocket. (F) 3D binding mode of compound 23.

As disclosed from Figure 3A, Ser135 forms a covalent bond with the ester motif of compound 1, which conforms to the covalent mechanism. The benzene ring at the end of compound 1 interacts with Tyr161 through a Pi-Pi stacking interaction, and a nitrogen atom of pyrazole motif forms a hydrogen bond with Gly133. As expected, compound 23 shares very similar binding mode to compound 1, except a hydrogen bond between the amino group and Val36. (Figure 3D, F)

For a better grasp of further information and verification of docking results, 100 ns MD simulations were conducted for optimized compounds. The root-mean-square deviation (RMSD) values of the atoms on protein backbones and ligand were calculated and recorded in Figure 4A. The RMSD curve of the protein backbone of NS2B-NS3 (3UI1) in complex with compound 23 is pretty stable with fluctuation around 2 Å during the process of MD simulation (Figure 4A). Also, the RMSD curve of compound 23 reached equilibrium status along the MD simulation process, confirming the stable binding mode from the docking process. Stable intermolecular interactions are also observed during the MD simulation, like the π–π stacking interaction between Tyr161 and compound 23, hydrophobic interaction between Pro132 and compound 23, as well as hydrogen bond formation with Gly133 and Val36, confirming the ligand–protein binding modes.

Figure 4.

Molecular dynamics simulation of compound 23 with NS2B-NS3. (A) RMSD of DENV NS2B-NS3 protein backbone atoms (blue line) and ligand (compound 23) (orange line) monitored throughout the 100 ns molecular dynamics simulations. (B) Normalized stacked bar chart representation of interactions and contacts over the course of the MD trajectories for the complex of protein and compound 23. Hydrophobic contacts fall into three subtypes: π–cation; π–π; and other, nonspecific interactions.

It is also observed the methoxy group of compound 1 adopts a converse direction toward solvent area (Figure 3B) while the corresponding ethoxy group of compound 23 occupies the pocket formed by Leu51, Ala56, and Thr53 (Figure 3E), contributing to the tight interaction toward NS2B-NS3, leading to higher binding affinity.

Asteroid plot represents the neighborhood residues of selected ligands. As disclosed from Figure 5A, Ser135 and Tyr161 of NS2B-NS3 are important for the interaction with compound 1, keeping stable interacting distance with compound 1 during MD simulation (Figure 5C). Similarly, Ser135 and Tyr161 are also observed in asteroid plot of compound 23 complex (Figure 5B), forming stable interactions with compound 23 (as revealed from Figure 5D). Moreover, Leu51 is involved in the interaction with compound 23 with a distance around 5 Å, which again confirms that compound 23 particularly occupied a pocket formed by Leu51, Ala56, and Thr53.

Figure 5.

Visualization and analysis of protein–ligand contacts. (A) Asteroid plots of protein–compound 1 complex. (B) Asteroid plots of protein–compound 23 complex. (C) Distances between compound 1 and crucial residues during MD simulation. Compound 1 forms covalent bond with Ser135 with a stable distance of about 1.5 Å, and interacts with Tyr161 through π–π stacking with stable distances around 5 Å. (D) Distances between compound 23 and crucial residues during MD simulation. Compound 23 forms covalent bond with Ser135 with a stable distance of about 1.5 Å, interacts with Tyr161 through π–π stacking with stable distances around 5 Å, and interacts with Leu51 through hydrophobic contacts.

In summary, we set out to discover a series of DENV NS2B-NS3 protease inhibitors from a lead compound with moderate activity. Fluorometric assays confirm that compounds 20, 22, 23, and 32 exhibit favorable inhibition rate values at a test concentration of 150 nM, among which compound 23 is determined to be the most potent inhibitor with an IC₅₀ value of 6.0 nM and a k_inac/K_i value of 1581 M^–1 s^–1, possessing a significant protective effect on cell survival rate. Furthermore, covalent docking, MD simulation, and other silicic methods were used to probe the interactions of these compounds and protein. The results we reported here provide a basis for further studies in animal models against DENV.

Glossary

Abbreviations

DENV

dengue virus

IC₅₀

half maximal inhibitory concentration

ESI

electrospray ionization

NMR

nuclear magnetic resonance

EWG

electron withdrawing group

EDG

electron donating group

BHK-21

baby hamster kidney 21

CCK-8

cell counting kit 8

MD

molecular dynamics

RMSD

root-mean-square deviation

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.1c00653.

• Details of the chemical synthesis and characterization for compounds, experimental procedures for biological assays, details of docking and molecular dynamics, dynamic cross-correlation map, stability assay of compound 23 (PDF)

The authors declare no competing financial interest.