Discovery of Novel Neuraminidase Inhibitors by Structure-Based Virtual Screening, Structural Optimization, and Bioassay

Abstract

Neuraminidase (NA) is a significant therapeutic target for treating influenza. In this study, a new lead NA inhibitor AN-329/10738021 was discovered by structure-based virtual screening, molecular dynamics simulations, and bioassay validation. Optimization of lead AN-329/10738021, which holds a novel scaffold of N′-benzylidene benzohydrazone, leads to discovery of some novel NA inhibitors Y-1–Y-11. Compound Y-1 exerts the best inhibition activity (IC₅₀ = 0.21 μM) against NA, which is better than oseltamivir carboxylate (OSC) (IC₅₀ = 3.04 μM) and lead AN-329/10738021 (IC₅₀ = 1.92 μM). Molecular docking analysis indicates that the good potency of Y-1 may be ascribed to the elongation of the benzylidene moiety of the molecule to the 430-cavity. The results of this study may offer useful reference for development of novel NA inhibitors.

Influenza virus belongs to the Orthomyxoviridae family. It can be classified as four types (A, B, C, and D) according to the antigenicity of nucleoprotein. As the main pathogen of human influenza, influenza A is often prone to cause influenza outbreak or pandemic. Influenza A virus is an enveloped negative-strand RNA virus. Its genome, which encodes up to 11 proteins, comprises eight viral RNA (vRNA) segments (PB2, PB1, PA, HA, NP, NA, M, and NS).^1,2 The base of single-stranded RNA is unstable and more likely to mutate than double-stranded. What’s more, gene exchange, deletion, and addition may occur between different eight RNA segments, which may lead to gene recombination. Therefore, the development of highly efficient anti-influenza drugs, which are insusceptible to virus mutation, is particularly important.

Neuraminidase (NA), a surface glycoprotein in influenza virus, plays a crucial rule in viral replication and infection and is a validated target for design of anti-influenza drugs.^3,4 Currently, the first-choice anti-influenza drugs recommended for treating influenza virus are NA inhibitors, such as oseltamivir and zanamivir.⁵ Though these antiviral drugs are efficient against the current influenza virus strains, their use can lead to the occurrence of drug-resistant virus.⁶

Recently, a new pocket named the “430-cavity”, which is adjacent to the active site of the NA, is attracting great attention. The 430-cavity has large molecular volume and connects directly to the active site, which makes it a promising binding site for inhibitor design. What’s more, the 430-cavity widely exists in various subtypes of influenza virus, and NA inhibitors designed based on the 430-cavity have broad spectrum.⁷ Some research groups have designed some new NA inhibitors based on the 430-cavity and have achieved good results. For example, Ju et al.⁷ had discovered a series of novel oseltamivir derivatives targeting the 430-cavity. Zhu et al.⁸ had designed and synthesized a series of novel analogues of zanamivir as NA inhibitors. The good inhibitory activity may be ascribed to the elongation of the C-1- groups of the molecules to the 430-cavity.

In this research, the discovery of novel NA inhibitors was performed by structure-based virtual screening, molecular dynamics (MD) simulations, chemical synthesis, and bioassay. A novel lead compound was first hit by virtual screening, and then the lead compound was used as template to design some new NA inhibitors targeting the 430-cavity. Finally, the designed 11 new inhibitors were synthesized and tested for biological activity.

In this paper, 10 typical NA inhibitors (shown in Figure 1) were picked out to generate pharmacophore models from different references.⁹⁻¹³ And the pharmacophore models were constructed by the GALAHAD module of the SYBYL-X 2.1 package (Tripos Inc., St. Louis, USA). The obtained 20 pharmacophoric models were validated by performing a Güner-Henry (GH) test.¹⁴ The GH test set contains 78 active molecules and 100 inactive molecules. The best pharmacophoric model hits a total of 79 molecules in the GH test set, which includes 76 active and 3 inactive molecules. The GH test score was 0.94, a threshold value considered as an excellent model, implying that this pharmacophoric model can be used for screening.¹⁵⁻¹⁷ The detailed statistical parameters are shown in Table 1.

Figure 1.

Structures of representative NA inhibitors.

Table 1. Detailed Statistical Parameters of the Best Pharmacophore Model Based on the GH Test.

parameters	value
total molecules in test set (D)	178
total number of actives in test set (A)	78
total hits (Ht)	79
active hits (Ha)	76
%Y, yield of actives [(Ha/Ht) × 100]	96.2
% A, ratio of actives [(Ha/A) × 100]	97.4
enrichment factor (EF)a	2.20
GH test scoreb	0.94

^aEF = (Ha D)/(Ht A).

^bGH test score = [Ha (3A + Ht)/(4Ht A)][1 – (Ht – Ha)/(D – A)].

The characteristics of the best pharmacophoric model are shown in Figure 2, including two positive nitrogens (NP_2, NP_3), two donor atoms (DA_1, DA_4), two hydrophobic centers (HY-7, HY-8), two acceptor atoms (AA-5, AA_6), and one negative center (NC_9). As seen from Figure 2A, the known inhibitor I (Figure 1) was superimposed on the pharmacophoric features. The carboxyl of inhibitor I can be readily superimposed on the negative center (NC_9) and H-bond acceptor (AA_5). The cyclohexene ring was superimposed on the hydrophobic center (HY_7). The aliphatic chain moiety of 3-(1-ethylpropoxy) makes it one hydrophobic center (HY_8). The carbonyl group of acetylamino was superimposed on the H-bond acceptor (AA_6). The two imino nitrogens of guanidyl could be readily identified by features of NP_2 and NP_3, respectively.

Figure 2.

Alignment of known inhibitors I (A) and lead compound AN-329/10738021 (B) on derived pharmacophoric features.

The workflow of virtual screening in this study is displayed in Figure 3. The large SPECS database, consisting of 212,713 small molecules, was selected for virtual screening in this work. The best pharmacophoric model was used as a 3D query to screen the SPECS database using the UNITY search engine in SYBYL-X 2.1. 1024 molecules were hit. The Lipinski drug five principles were used as filters to screen the drug properties of the compound, and finally Database 2 (774 compounds) was obtained.

Figure 3.

Workflow of the virtual screening process.

Subsequently, molecular docking¹⁸ had been carried out for Database 2 to study the binding modes between inhibitors and NA crystal structures (PDB ID: 2HU0). As a result, the top 100 compounds with higher docking scores and binding modes close to that of the original extracted ligand with 2HU0 were screened out to compose Database 3. By careful analysis of the spatial matching and interactions between the 100 compounds and NA protein, 10 compounds with higher total score values were selected to form Database 4 for further study. The structures of the 10 compounds are shown in Figure 4 (SPECS-1–SPECS-5) and Figure S1 (SPECS-6–SPECS-10).

Figure 4.

Chemical structures and inhibitory activity of the five NA inhibitors with their SPECS IDs.

To discover novel lead compounds, MD simulations were performed for 10 compounds in Database 4 by Amber 12.¹⁹ The stability of the NA–inhibitor complex was determined based on the root-mean-square deviation (RMSD)²⁰ values of the acceptor backbone atoms (Figure S2 and Figure S3). The results show that five compounds can well embed in the active site or 430-cavity of the neuraminidase and have smaller binding energies compared with the other five compounds. The predicted binding free energies (ΔG_bind) of the 10 compounds via Molecular Mechanics/Generalized Born Surface Area (MM/GBSA)²¹ and Molecular Mechanism/Poison–Boltzmann Surface Area (MM/PBSA)²² methods are listed in Table S1 and Table S2. It is known that there is generally a good correlation between binding free energy and compound’s activity. The more negative the ΔG_bind of a compound is, the more active the compound is.^21,22 By calculation of the binding free energies of the 10 compounds in Database 4, five compounds (SPECS-1–SPECS-5) with smaller binding free energies were selected to make up Database 5.

Then, the five compounds in Database 5 were bought from SPECS for further biological evaluation. OSC was selected as the reference compound. The results show that five compounds all exert good potency, with IC₅₀ values of less than 20 μM (Figure 4). Compound AN-329/10738021 has the best inhibitory activity (IC₅₀ = 1.92 μM), which is better than OSC (IC₅₀ = 3.04 μM). Moreover, among the five compounds, three compounds (AN-329/10738021, AH-487/40686965, and AH-487/406867204) with the lowest IC₅₀ values all have the same skeleton structure of N′-benzylidene benzohydrazone. Accordingly, compound AN-329/10738021 could be used as a lead compound to perform further modification to obtain new inhibitors with good inhibition activity. As shown in Figure 2B, lead compound AN-329/10738021 was superimposed onto the pharmacophoric features. Compared with the features in Figure 2A, the 3,5-dihydroxy substituted benzohydrazide makes the C-1 and C-2 position one acceptor center (AA_5) and negative center (NC_9). Two positive nitrogen atoms of the −C=N–NH– group could be readily identified by features of NP-2 and NP-3. The 5-nitrobenzylidene can be readily superimposed on the hydrophobic (HY_8) center. Consequently, the lead compound was valuable for further study for its good inhibitory activity.

According to the features derived from the pharmacophoric model and molecular docking results, some new NA inhibitors (Y-1–Y-11) were designed (Table 2 and Scheme 1). Lead compound AN-329/10738021 was used as template. The design idea is summarized in Figure 5. The 430-cavity is mainly composed of Trp403, Lys432, Ile427, and Thr439, etc., which could be regarded as a hydrophobic pocket. Therefore, some hydrophobic groups, such as nitro groups (Y-1–Y-4) and methoxy groups (Y-5, Y-6), were introduced to ring B and expected to interact with the 430-cavity. It was revealed that the active site of NA could be empirically divided into five regions, termed subsites S1–S5, that are essential for the interactions with different NA inhibitors (Figure 6).²³ S1 site consists of three positively charged arginine residues (Arg118, Arg292, and Arg371), which can form very strong electrostatic interactions with anionic substituents from the inhibitor and provide hydrogen-bonding environment. S2 site is comprised of Glu119 and Glu227, which can form hydrogen bonds with basic groups. S3 site is a small hydrophobic region formed by the side chains of Trp 178 and Ile 222 adjacent to a polar region provided by the side chain of Arg 152, which can interact with various hydrophobic groups; S4 site is a hydrophobic region comprised of side chains of Ile 222, Ala 246, and the hydrophobic face of Arg 224, which also can interact with some hydrophobic groups. S5 is a mixed polarity region formed from the carboxylate of Glu 276 and the methyl of Ala 246, which may produce interactions with various groups by the hydrophilic carboxyl of Glu276 and the hydrophobic methyl of Ala246. The flexible 430-cavity is next to the subsite S1,⁸ which can provide a large cavity adjacent to the active site. This provides a good chance to design highly novel NA inhibitors that target both the 430-cavity and the known active subsites. Herein we designed and synthesized a series of lead compound analogs (Y-1–-Y11) with different substitutions at aryl B rings to explore the 430-cavity.

Table 2. Activity of the Compounds Y-1–Y-11 in the Bioassay against NA.

Compd	IC50 (μM)	Compd	IC50 (μM)
Y-1	0.21 ± 0.04	Y-7	13.77 ± 2.89
Y-2	3.71 ± 0.53	Y-8	9.67 ± 1.33
Y-3	5.47 ± 1.32	Y-9	10.56 ± 1.89
Y-4	3.97 ± 0.55	Y-10	22.74 ± 3.55
Y-5	14.83 ± 3.76	Y-11	36.60 ± 4.17
Y-6	11.64 ± 2.41	OSC	3.04 ± 0.47
Leada	1.92 ± 0.41

^aLead is AN-329/10738021.

Scheme 1. Synthesis of Compounds Y-1–Y-11.

Figure 5.

Design of new inhibitors based on the lead compound.

Figure 6.

Schematic diagram of S1–S5 sites division and binding mode of NA and sialic acid.²³

Compounds Y-1–Y-11 were synthesized by feasible synthetic strategies in Scheme 1. Their activities against NA were further biologically tested. The inhibitory profiles for in vitro tests are displayed in Figure S4. For comparison, lead compound AN-329/10738021 andOSC were evaluated as the reference compounds. As seen from Table 2, compound Y-1 exhibits the most potent inhibitory activity (IC₅₀ = 0.21 μM), which is better than AN-329/10738021 (IC₅₀ = 1.92 μM) and OSC (IC₅₀ = 3.04 μM). The activities of Y-2 (IC₅₀ = 3.71 μM), Y-3 (IC₅₀ = 5.47 μM), and Y-4 (IC₅₀ = 3.97 μM) are comparable to that of OSC. It can be seen that compounds with strongly electronegative groups (such as nitro group and halogen atoms) in ring B show better inhibitory activity, whereas the introduction of methoxy and amide groups (Y-5–Y-7) in ring B maybe decrease the compounds’ activity. As shown in Scheme 1, the only difference between Y-1–Y-4 and Y-8–Y-11 is the existence of an additional hydroxy group at the para-position of the hydrazone group, which decreases compounds’ activity. This trend could be rationalized with the strong hydrophilicity of the hydroxy group. Molecular docking analysis indicates that the hydroxyl group of the A ring at the para-position of hydrazone group is close to the S3 or S4 region of the NA active site (see Figure S5). Nevertheless, S3 and S4 are hydrophobic regions, which interact with hydrophobic groups well.

To better comprehend the differentiation of inhibitory activity of Y-1, AN-329/10738021, and OSC, molecular docking was carried out to study the binding modes between NA and inhibitors. Figure 7 shows the binding poses of Y-1, AN-329/10738021, and OSC with NA (PDB 2HU0). The 3D figures of Y-1, lead compound (AN-329/10738021), and OSC had been aligned in Figure 8 to ensure that protein conformation is the same for all three complexes. From Figures 7 and 8 it can be seen that the rings A of Y-1 and AN-329/10738021, as well as the whole skeleton of OSC, can all well implant into the active site of neuraminidase. However, it is worth noting that the aryl B rings in Y-1 and AN-329/10738021 could be elongated into the 430-cavity, whereas the OSC could not reach the 430-cavity.

Figure 7.

Binding modes of OSC, AN-329/10738021 and Y-1 with NA (PDB 2HU0). Hydrogen bonds are indicated by green dotted lines (for the hydrophobic surfaces on protein, brown represents the maximum hydrophobic, and blue represents the maximum hydrophilic).

Figure 8.

Proposed binding modes of Y-1, AN-329/10738021 and OSC in the active site of NA (PDB 2HU0).

From 2D figures one can see that the aryl B rings of Y-1 and AN-329/10738021 were projected toward the 430-cavity to form hydrogen bonds and hydrophobic interactions. For example, for Y-1, the 3-nitro may form two hydrogen bonds with residue Arg428 of the 430-cavity. The 4-hydroxyl can form one hydrogen bond with residue Glu425. The aryl B ring can form extensive hydrophobic interactions with some key residues Ile427, Pro431, and Lys 432 of the 430-cavity. For the lead AN-329/10738021, the aryl B ring may form hydrophobic interactions with residues Ile427 and Pro431 of the 430-cavity. The 5-nitro may form three hydrogen bonds with residues Arg118 and Arg371 of the S1 site. Therefore, the good inhibition activity of Y-1 and lead AN-329/10738021 may be attributed to their elongated aryl B ring groups. The A ring moiety of Y-1 and AN-329/10738021 can also interact with some active region of the NA active site. For example, the 3-hydroxyl of Y-1 may form one hydrogen bond with the hydrophilic carboxyl of Glu276 of the S5 site. The 3-hydroxyl of AN-329/10738021 may form one hydrogen bond with residue Glu227 of the S2 region. The 5-hydroxyl may form three hydrogen bonds with residues Arg156 and Glu119 of the S2 site. Regarding the acylhydrazone moiety, for Y-1, the carbonyl oxygen atom can form three hydrogen bonds with residues Tyr406, Arg292, and Arg371 of the S1 site, and the double bond nitrogen can form one hydrogen bond with Arg371; no hydrogen bond is formed with the near residues for the lead AN-329/10738021. In addition, the acylhydrazone group can form other extensive interactions, such as van der Waals force and carbon hydrogen bond with the around residues of the active site. In comparison of Y-1 and AN-329/10738021, although the aryl B ring may both interact with the 430-cavity, the introduced group in Y-1 can enter deeper into the 430-cavity and it can form more hydrogen bonds and stronger hydrophobic interactions with the residues of the 430-cavity, which may lead to its stronger inhibitory activity compared with the lead AN-329/10738021. For OSC, the deprotonated carboxyl may form multiple hydrogen bonds with the near residues (Arg118, Arg292, Arg371, and Tyr347). At the same time, it can form strong electrostatic interactions with the three positively charged arginine residues (Arg118, Arg292, Arg371) of the S1 site. The carbonyl oxygen atom of acetylamino may form one hydrogen bond with residue Arg152 of the S3 site. The cyclohexene ring and the aliphatic chain moiety of 3-(1-ethylpropoxy) may form strong hydrophobic interactions with residues Trp178, Arg224, and Ile222 of the S3 and S4 sites. Although OSC can well interact with the NA active site, the docking results show that it could not interact with the 430-cavity, which perhaps leads to its weaker activity compared with Y-1 and the lead AN-329/10738021.

MD simulations were also carried out for all synthesized compounds. The stability of the NA-inhibitor complex was determined based on the RMSD values of the acceptor backbone atoms (Figure S6). On the basis of MD simulations, the binding free energies (ΔG_bind) between the 11 new inhibitors and NA are calculated and shown in Table 3. In this research, ΔG_bind was calculated by MM/PBSA and MM/GBSA methods. The PB model is theoretically more rigorous than the GB models, and MM/PBSA is often considered to be naturally superior to MM/GBSA in predicting binding free energies.²⁴ As can be seen from Tables 2 and 3, the smaller the ΔG_bind of inhibitor-NA is, the stronger the activity of the inhibitor is for most inhibitors. For example, Y-1 has the lowest ΔG_bind (−31.26 kcal·mol^–1) and it exerts the best NA-inhibiting activity. What’s more, Y-2 (−27.85 kcal·mol^–1), Y-3 (−21.87 kcal·mol^–1), and Y-4 (−18.41 kcal·mol^–1), which have better inhibitory activities, also have lower ΔG_bind values. The ΔG_bind values of Y-1–Y-3 are all less than −20.00 kcal·mol^–1. Y-5–Y-9 have similar inhibitory activities, and their IC₅₀ values are about 10.00 μM. And their ΔG_bind values are between −15.00 and −20.00 kcal·mol^–1 except Y-6. Y-10 and Y-11 have poor activity in inhibition of NA, especially Y-11, which has the lowest activity (36.60 μM) and the highest binding energy (−13.00 kcal·mol^–1). Hence one can see that the theoretical results are consistent with the experimental results.

Table 3. Calculated Binding Free Energies (ΔG_bind, kcal·mol^–1) of Ligand–2HU0 Complexes via MM/PBSA and MM/GBSA Methods.

No.	IC50 (μM)	VDW	EEL	ΔGgas	ΔGGB	ΔGSA	ΔGsolv(GB)	ΔGbind(GB)	ΔGPB	ΔGSA	ΔGsolv(PB)	ΔGbind(PB)
Leada	1.92	–36.74	–11.41	–48.04	31.36	–4.36	26.99	–21.05	32.67	–3.67	29.21	–18.83
Y-1	0.21	–38.30	–17.98	–56.28	26.17	–4.60	21.57	–34.71	28.60	–3.58	25.02	–31.26
Y-2	3.71	–34.39	–15.54	–49.93	27.87	–4.09	23.78	–26.15	25.45	–3.37	22.07	–27.85
Y-3	5.47	–30.57	–30.60	–61.17	40.53	–4.48	36.05	–25.12	42.28	–2.98	39.30	–21.87
Y-4	3.97	–40.86	–23.47	–64.33	44.33	–5.16	39.80	–24.52	49.95	–4.03	45.91	–18.41
Y-5	14.83	–29.53	–55.04	–84.57	68.21	–4.24	63.97	–20.60	68.09	–3.01	65.08	–19.49
Y-6	11.64	–42.66	–8.16	–50.82	27.72	–5.33	22.38	–28.44	31.54	–4.03	27.52	–23.21
Y-7	13.77	–35.57	–36.48	–73.05	49.70	–4.60	45.09	–26.96	55.88	–3.23	52.64	–19.41
Y-8	9.67	–28.81	–39.42	–68.23	52.79	–4.07	48.72	–19.51	54.67	–3.27	51.39	–16.84
Y-9	10.56	–27.12	–6.09	–33.21	18.32	–3.05	15.27	–17.94	18.00	–2.66	15.35	–17.87
Y-10	22.74	–29.56	–25.01	–54.58	37.69	–3.70	34.00	–20.58	41.13	–2.90	38.23	–16.34
Y-11	36.60	–21.01	–5.31	–26.32	15.57	–2.39	13.18	–13.14	15.45	–2.13	13.31	–13.00

^aLead is stands for AN-329/10738021.

In this research, a novel lead NA inhibitor AN-329/10738021 (IC₅₀ = 1.92 μM) was discovered by virtual screening, MD simulations, and bioassay validation. Structural optimization of AN-329/10738021 leads to discovery of a series of 11 novel inhibitors (Y-1–Y-11). These inhibitors exhibit good inhibition activity in the range of potency of FDA-approved oseltamivir carboxylate. They have concise chemical structures and are relatively easy to synthesize, with good potential for further improvement by introducing novel moieties and extension of structures. Particularly, compound Y-1 exerts the best NA-inhibition activity (IC₅₀ = 0.21 μM), which is better than OSC (IC₅₀ = 3.04 μM). Molecular docking analysis indicates that the good inhibition activity of Y-1 may be ascribed to the elongation of benzylidene moiety of the molecule to the 430-cavity. The 430-cavity plays an important role in discovery of novel NA inhibitors.

Glossary

Abbreviations

NA

neuraminidase;

OSC

oseltamivir carboxylate;

MD

molecular dynamics;

GH

Güner-Henry;

RMSD

root-mean-square deviation;

MM/GBSA

Molecular Mechanics/Generalized Born Surface Area;

MM-PBSA

Molecular Mechanism/Poison–Boltzmann Surface Area

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.9b00447.

• Methods and materials, database preparation, structure-based virtual screening, Table S1, Table S2, Figure S1, Figure S2, Figure S3, Figure S4, Figure S5, Figure S6, general procedure for synthesis of novel compounds and in vitro neuraminidase inhibition assay (PDF)

Author Present Address

^† School of Chemical and Environmental Engineering, Shanghai Institute of Technology, Shanghai 201418, China.

Author Contributions

The manuscript was finished through contributions of all authors.

The authors acknowledge financial support by Natural Science Foundation of Shanghai (No. 15ZR1440400), Collaborative Innovation Fund (No. XTCX2016-14), Middle and Youth Teachers Scientific and Technological Talents Developing Fund (No. ZQ 2018-20), and Shanghai Municipal Education Commission (Plateau Discipline Construction Program).

The authors declare no competing financial interest.