Exploring small molecules with pan-genotypic inhibitory activities against hepatitis C virus NS3/4A serine protease

Abstract

Among the many Hepatitis C virus (HCV) genotypes and subtypes, genotypes 1b and 3a are most prevalent in United States and Asia, respectively. A total of 132 commercially available analogs of a previous lead compound were initially investigated against wild-type HCV genotype 1b NS3/4A protease. Ten compounds showed inhibitory activities (IC₅₀ values) below 10 μM with comparable direct binding affinities (K_D values) determined by surface plasmon resonance (SPR). To identify pan-genotypic inhibitors, these ten selected compounds were tested against four additional genotypes (1a, 2a, 3a, and 4) and three drug-resistant mutants (A156S, R155K, and V36M). Four new analogs have been identified with better activities against all five tested genotypes than the prior lead compound. Further, the original lead compound did not show activity against genotype 3a NS3/4A, whereas four newly identified compounds exhibited IC₅₀ values below 33 μM against genotype 3a NS3/4A. Encouragingly, the best new compound F1813–0710 possessed promising activity toward genotype 3a, which is a huge improvement over the previous lead compound that had no effect on genotype 3a. This intriguing observation was further analyzed by molecular docking and molecular dynamics (MD) simulations to understand their different binding interactions, which should benefit future pan-genotypic inhibitor design and drug discovery.

Graphical Abstract

Hepatitis C virus (HCV) infection is the main cause of both acute and chronic hepatitis.¹ Acute HCV infection is usually asymptomatic and only very rarely associated with life-threatening disease.² However, chronic HCV infection, which develops from the acute HCV infections and continues over the course of many years, can cause significant complications such as cirrhosis, liver cancer and complete liver failure.^3,4 HCV is categorized into eight main genotypes based on genetic variations.^5–7 Within each genotype there are a series of minor variations, further classified as subtypes, which are numbered a, b, c, d, etc. in the order of their discovery.⁷ Among them, HCV genotypes 1, 2 and 3 have a broad global distribution. Genotype 1b is the most prevalent in the United States and genotype 3a is the most common one in south Asia.^{8, 9} It has also been reported that genotypes 1b or 3 may increase the risk of cancer in people who have already developed cirrhosis.^10–12 Therefore, these two genotypes are of particular importance in drug discovery against HCV.

HCV is a positive-sense, single-stranded RNA virus belonging to the Flaviviridae family. Its genome is translated to produce a large polyprotein that undergoes proteolytic cleavages to form ten individual viral proteins, each of which has specific functions in the viral life cycle. Ten viral proteins include three structural proteins (the core protein C, envelope glycoproteins E1 and E2), a small integral membrane protein, p7, and six important nonstructural (NS) proteins (NS2, NS3, NS4A, NS4B, NS5A, and NS5B).¹³ Among these, there are two viral proteases, the NS2 cysteine protease and the NS3 serine protease. HCV NS3 requires NS4A as a cofactor for functioning properly, and NS3/4A cleaves the junctions of NS3/NS4A, NS4A/NS4B, NS4B/NS5A, and NS5A/NS5B. NS3 is a multifunctional protein with serine protease activity at the N-terminus (aa 1–180) with a catalytic triad (S139, H57 and D81) and a nucleoside-triphosphatase dependent RNA helicase activity at the C-terminus (aa 181–631), and both activities are required for virus replication.¹⁴ Therefore, NS3/4A is essential for viral replication and the formation of infectious viral particles,^{15, 16} and is considered to be a valid drug target for anti-HCV therapy.

We have been focusing on developing small molecule inhibitors with pan-genotypic effect on multiple HCV genotypes as well as some known drug resistant mutants.^{17, 18} A sulfonamide compound, F2322–0885, with inhibitory activity toward genotypes 1a, 1b, 2a and 4 of HCV NS3/4A protease, as well as some genotype 1b NS3/4A mutants was identified as a lead in our previous study.¹⁷ Research has shown that genotype 3a, the second common genotype of HCV, is associated with genotype specific mechanisms of steatosis in addition to accelerated development of fibrosis and higher rates of hepatocellular carcinoma.^{19, 20} There are numerous structures available for genotype 1b NS3 protease, both in the absence and presence of various inhibitors.^21–23 However, there is no available structure determined for genotype 3a NS3. Soumana et al. solved a chimeric HCV NS3/4A protease crystal complex structure of genotype 1a3a and partially predicted the 3D features of genotype 3a with the identical residues in the active site,²⁴ providing interesting structural features at atomic-level for target-ligand interactions to explain the inhibitory activity toward genotype 3a NS3/4A.

In this study, we used fluorescence-based enzymatic assays to compare inhibitory activities of 132 commercially available analogs of our prior lead compound against genotype 1b NS3/4A. Among these analogs, twelves showed inhibitory activity (IC₅₀ values) below 100 μM with ten of them even less than 10 μM (Fig. 1A and 1B). Two compounds, F1813–0710 and F0325–0125, exhibited IC₅₀ values slightly less than 2 μM, four (F0325–0086, F0725–0019, F0325–0092, and F0816–0111) showed between 2 – 5 μM, the IC₅₀ values of the other four (F2322–0877, F2322–0904, F1813–0711, and F0325–0093) varied between 5 and 10 μM, and the remaining two compounds (F1822–0567 and F2730–0247) showed higher IC₅₀ values between 10 – 100 μM. To further validate that these analog compounds are targeting NS3/4A, direct binding analysis was followed using surface plasmon resonance (SPR), and their K_D values were determined to all be comparable to their IC₅₀ values except the two with higher IC₅₀ values (Fig. 1B). The dose-response curve of F1813–0710 is shown as an example (Fig. 1C), and the structures of all twelve analogs and the initial lead are provided in Fig. 1D with three different highlights to visualize differences in analogs. F2322–0877 and F2322–0904 have the same thiazolemethyl linker as the lead compound F2322–0885 between the sulfonamide and amide moiety, whereas the other ten analogs have an ortho-phenyl as the linker. Both methylphenyl and phenyl are tolerated for connecting the sulfonamide moiety. Substituted phenyls and benzothiazoles are favored connecting the amide moiety. A preliminary Structure-Activity Relationship (SAR) based on 132 analog compounds is shown in Fig. S1 in the supplemental material. Interestingly, two compounds, F1813–0710 and F0325–0125, have two chlorines on the benzothiazole group and phenyl moiety, respectively, both showing the best activities against genotype 1b NS3/4A. The difference between these two compounds are the benzothiazole scaffold, with fused benzothiazole ring in F1813–0710 and the separate thiazole and phenyl rings linked by a single bond in F0325–0125. This may indicate that the benzothiazole ring could be replaced by separated phenyl and thiazole rings in future analog exploration, which will potentially enrich the analog scaffold search.

Fig. 1. Compounds tested against the HCV genotype 1b NS3/4A protease.

(A) A schematic of the overall compound test and validation process. (B) Bar graph plot of IC₅₀ values determined by fluorescence-based enzymatic assay and the dissociation equilibrium constants (K_D) by SPR of 13 sulfonamides toward the HCV genotype 1b. Bars that reached to the top represent K_D values of over 200 μM (very weak binding or no binding). (C) IC₅₀ fitting curve of F1813–0710 using the three parameter Hill equation. (D) Structures of prior lead and twelve analogs with inhibitory activities below 100 μM.

With the ten analogs of F2322–0885 showing reasonable inhibitory activity (IC₅₀ values less than 10 μM) against HCV genotype 1b NS3/4A protease, we subsequently tested their activities against other genotypes and drug-resistant mutants of genotype 1b in order to identify pan-genotypic inhibitors also possessing activities against three drug-resistant mutants. A total of five genotypes of HCV NS3/4A protease, including four other genotypes (1a, 2a, 3a, and 4), along with genotype 1b and its three mutants (A156S, R155K, and V36M), were used in this comparison. Four compounds (F1813–0710, F1813–0711, F0325–0125, F0816–0111) were identified with better activities than the prior lead F2322–0885 against all five tested genotypes (Fig. 2). The original lead did not show activity against genotype 3a NS3/4A, whereas four newly identified compounds exhibited inhibitory IC₅₀ values ranging from 14 μM to 33 μM. Of these four analogs, three (F1813–0711, F0325–0125, F0816–0111) did not retain activity against the drug resistant mutant R155K NS3/4A genotype 1b and became worse than the original lead (Fig 2C – 2E). On the other hand, F1813–0710 exhibits better inhibitory activities than F2322–0885 against all five tested genotypes (Fig. 2), especially for genotype 3a, and still maintains its activities against all three tested mutants. This huge activity difference between F2322–0885 and F1813–0710 motivated us to explore the different binding modes for the two compounds against genotype 3a using molecular modeling methods. Since no HCV genotype 3a NS3/4A protease structure has been solved to date, we used the only available structure, a crystal structure of a chimeric HCV genotype 1a3a NS3/4A protease (PDB code: 5EQR²⁴). This unique structure contains the same active site residues as the genotype 3a mutated from genotype 1a, and the rest of the residues are still maintained as the genotype 1a; hence this chimeric HCV genotype 1a3a NS3/4A protease structure was used for our computational studies. Since the active site residues of this genotype 1a3a NS3/4A are identical to those of genotype 3a, we only needed to mutate the residues that are different to match them to genotype 3a based on the sequence alignment between genotypes 1a3a and 3a (Fig. 3A ). After manually mutating the crystal structure of genotype 1a3a to the corresponding residues of genotype 3a shown in cyan color, the 3D structure of genotype 3a was built and is shown in Fig. 3B. Then molecular docking was performed using the constructed 3D structure of genotype 3a with F1813–0710 and F2322–0885, which exhibited distinct differences in their activities toward genotype 3a NS3/4A.

Fig. 2. Pan-genotypic inhibitory activity (IC₅₀) of the lead and four analogs.

Bar graphs of compound F2322–0885 (A), F1813–0710 (B), F1813–0711 (C), F0325–0125 (D), and F0816–0111 (E) against HCV NS3/4A protease from five genotypes 1b (GT1b), 1a (GT1a), 2a (GT2a), 3a (GT3a), 4 (GT4) and three drug-resistant mutants (A156S, R155K, and V36M) of genotype 1b.

Fig. 3. Predicted binding poses for F1813–0710 and F2322–0885 in HCV genotype 3a NS3/4A protease from molecular docking.

(A) Sequence alignment of HCV chimeric genotype 1a3a and genotype 3a NS3 proteases. Catalytic triad residues of the proteases are shown in red letters, and different residues between genotypes 1a3a and 3a are colored in cyan. (B) The 3D structures of the HCV genotype 3a NS3 protease (manually mutated from the genotypes 1a3a 5EQR²⁴) with catalytic triad shown in stick representation (light green) and the mutated residues (cyan ribbon). The active site is shown in tan surface representation. Predicted interactions of F1813–0710 (C) and F2322–0885 (D) with genotype 3a NS3 protease from molecular docking. The 2D structure of F1813–0710 and F2322–0885 are shown below each predicted model. The catalytic residues are shown in green sticks, while other interactive residues of NS3 protease are shown in tan sticks. The figures showing interactions were produced by UCSF Chimera.²⁵

Despite the two different linkers between the sulfonamide and amide groups in F1813–0710 and F2322–0885 [ortho-phenyl linker for F1813–0710 (Fig. 3C) – the thiazolemethyl linker for F2322–0885 (Fig. 3D)], the sulfonamide moieties in these two compounds formed similar H-bonds interactions with the backbone NH of G137 and the sidechain OH of S139. In addition, similar π-stacking interactions were observed for the imidazole ring of H57 with the benzothiazole of F1813–0710 and the phenyl ring of F2322–0885, respectively. F1813–0710 shows one more H-bond between the NH (position 10) of the amide and the backbone C=O of R155. Since docking essentially only considers the flexibility of small compounds binding to a rigid receptor in the gas phase, molecular dynamics (MD) simulations for these two docking complexes were performed to further explore the interactions between small molecules and the HCV genotype 3a NS3/4A protease in water as a solvent.

In the 6-ns MD simulations, the RMSD for the two small molecules could explore the stability of the ligands in the active site of HCV genotype 3a NS3/4A protease. The results in Fig. 4 illustrate that F1813–0710 was more stable than F2322–0885 in the HCV genotype 3a NS3/4A protease. F2322–0885 exhibited more flexibility in the binding site of the protease with an RMSD of 1.7–3 Å fluctuation during the MD simulations, indicating less stable binding. On the other hand, F1813–0710 does not show significant changes from the starting conformation, with fluctuation around 0.6–1.0 Å. This result indicates that F1813–0710 had much more stable binding with the HCV genotype 3a NS3/4A protease with tighter interactions throughout the MD simulations.

Fig. 4.

RMSD for F1813–0710 and F2322–0885 with respect to the initial conformation in their docked HCV genotype 3a NS3/4A protease complexes during MD simulations.

The molecular docking results imply that some H-bond interactions between the small compounds and the receptor may promote their binding with HCV genotype 3a NS3/4A protease. To reveal these contributions in more detail, H-bonds was monitored during the 6-ns MD simulations. The percentage of H-bonds during the full simulations are listed in Table 1, which indicates that the two oxygen atoms (O19 and O20 in Fig. 3A) of the sulfonamide in F1813–0710 produce two more stable H-bonds with the sidechain OH of S139 and the backbone NH of G137 than those in F2322–0885 (for O19 and O20 in Fig. 3B), with frequencies of 49% and 35%, respectively, in F1813–0710, compared to 19% and 23% in F2322–0885. Additionally, the other H-bond for F1813–0710 with R155 is the predominant one (with 74% occupancy) during MD simulations. This result demonstrates that F1813–0710 is stable when binding to HCV genotype 3a NS3/4A protease.

Table 1.

H-bond interactions for F1813–0710 and F2322–0885 against the HCV genotype 3a NS3/4A protease during MD simulations.

HCV genotype 3a NS3/4A	F1813–0710	Occupancya (%)	F2322–0885	Occupancy (%)
R155 (backbone)	NH10	74
S139 (sidechain)	O19	49	O19	19
G137 (backbone)	O20	35	O20	23

^aOccupancy is defined as the percentage of simulation time that a specific interaction exists.

The enthalpy contributions of individual amino acids were calculated by the MM-GBSA method^{26, 27} revealing the impact of specific residues on the binding affinity of F1813–0710 and F2322–0885. A comparison between two compounds is presented in Fig. 5 with the energy contributions of the pivotal residues. Residues H57, K136, and G137 play prominent roles in the binding of F1813–0710, while the enthalpy contributions of these residues for F2322–0885 are also significant, but relatively lower than those of F1813–0710. For example, H57, K136, and G137 contributed −2.8, −0.8 and −1.2 kcal mol⁻¹ more, respectively, to the F1813–0710 interaction than to that with F2322–0085. This agrees with the experimental results of F1813–0710 exhibiting better inhibitory activity against genotype 3a NS3/4A. Although L132 and L135 contributed more to the F2322–0885 interaction, the differences are relatively small and cannot counterbalance the differences from H57, K136, and G137. In addition, R155 contributes significantly more energy to F1813–0710 than to F2322–0085, which is consistent with the H-bond analysis of R155 showing high occupancy with F1813–0710 during MD simulations. A similar trend was observed with another catalytic triad residue, S139. In summary, the pivotal residues in the active site contributed more energy to F1813–0710, explaining its favored binding with HCV genotype 3a NS3/4A.

Fig. 5.

Per residue contributions for HCV genotype 3a NS3/4A protease with F1813–0710 (black) and F2322–0885 (red).

In summary, a total of 132 commercial analogs of the prior lead F2322–0885 were subject to the inhibitory activity test against HCV genotype 1b NS3/4A protease. Ten of them showed inhibitory activity less than 10 μM against genotype 1b, four of which exhibited either enhanced or comparable activities against the other four tested genotypes 1a, 2a, 3a, 4 and three drug-resistant mutants (A156S, R155K, and V36M) of genotype 1b. Overall, F1813–0710 was the best analog with improved inhibitory activities against all five tested genotypes while maintaining inhibitory activities toward the HCV mutants. Interestingly, a very distinct inhibitory activity difference was observed for F1813–0710 and F2322–0885 toward genotype 3a. F1813–0710 was able to inhibit HCV genotype 3a NS3/4A protease at 14 μM IC₅₀ value, whereas F2322–0885 did not have any effect. Subsequent molecular docking and MD simulations were applied to investigate the different binding modes for these two compounds with genotype 3a. F1813–0710 is more stable than F2322–0885 binding with the HCV genotype 3a NS3/4A protease during MD simulations, with lower RMSD and more stable H-bond interactions. In addition, the pivotal residues in the active site contributed more energy to F1813–0710, which is consistent with the experimental results of F1813–0710 as a favored inhibitor of HCV genotype 3a NS3/4A protease.

• Ten analogs showed inhibitory activities below 10 μM against genotype (GT) 1b NS3/4A protease

• SPR confirmed the specificity against GT 1b NS3/4A protease

• Four analogs were identified as pan-genotypic inhibitors of GTs 1a, 1b, 2a, 3a, 4

• The new lead possessed much more promising activity than the prior lead toward GT 3a

• Docking and MD demonstrate the differences between the two leads toward GT 3a