Antiviral Disaccharide Lead Compounds against SARS‐CoV‐2 through Computer‐Aided High‐Throughput Screen

Abstract

SARS‐CoV‐2 infects human epithelial cells through specific interaction with angiotensin‐converting enzyme 2 (ACE2). In addition, heparan sulfate proteoglycans act as the attachment factor to promote the binding of viral spike protein receptor binding domain (RBD) to ACE2 on host cells. Though the rapid development of vaccines has contributed significantly to preventing severe disease, mutated SARS‐CoV‐2 strains, especially the SARS‐CoV‐2 Omicron variant, show increased affinity of RBD binding to ACE2, leading to immune escape. Thus, there is still an unmet need for new antiviral drugs. In this study, we constructed pharmacophore models based on the spike RBD of SARS‐CoV‐2 and SARS‐CoV‐2 Omicron variant and performed virtual screen for best‐hit compounds from our disaccharide library. Screening of 96 disaccharide structures identified two disaccharides that displayed higher binding affinity to RBD in comparison to reported small molecule antiviral drugs. Further, screening PharmMapper demonstrated interactions of the disaccharides with a number of inflammatory cytokines, suggesting a potential for disaccharides with multiple‐protein targets.

Antiviral mechanism of disaccharide lead compounds: Disaccharide lead compounds bind to the RBD structure of SARS‐CoV‐2/ SARS‐CoV‐2 Omicron variant spike protein, restraining their formation of a ternary complex with heparan sulfate (HS) and angiotensin‐converting enzyme 2 (ACE2), thus inhibiting virus attachment to the cell surface.

Introduction

The ongoing pandemic of coronavirus disease 2019 (COVID‐19) caused by severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) has spread rapidly with an increasing death toll of more than 6 million and confirmed cases exceeding 583 million (by date of 9 August 2022). ^[1] Although over 12 billion vaccine doses have been administered, there is still a large population worldwide nonvaccinated. Furthermore, due to the rapid mutations of the virus, infected or 3‐doses vaccinated individuals still became infected. ^[2] Omicron variant through mutations in the spike protein makes the virus more difficult to be recognized by antibodies, achieving immune escape. On the other side, the mutations are at their ACE2 binding interface, maintaining a high level of invasive ability. ^[3] Therefore, COVID‐19 still poses a huge threat to human life and health around the world, and development of novel drug candidates is still a top priority ^[4] to combat the disease.

The strategy to develop candidate compounds is to target either structural or non‐structural proteins of coronavirus, aiming to control invasion and replication process of the virus, such as the novel Pfizer agent paxlovid which show inhibiting effect on proteases. ^[5] Spike protein is a structural protein that can interact with human angiotensin‐converting enzyme 2 (ACE2). ^[6] The binding of spike protein to ACE2 induces dissociation of S1 subunit from ACE2, prompting S2 subunit to transit from a metastable pre‐fusion to a more‐stable post‐fusion state that is essential for virus‐membrane fusion ^[6] Heparan sulfate proteoglycans (HSPGs), existing in almost all mammalian cell membranes, is a common receptor of many viruses ^[7] including SARS‐CoV‐2. Binding of heparan sulfate (HS) to the viral surface proteins promotes viral attachment to the specific receptor (ACE2) by increasing its local concentration. ^[8] Moreover, it has been shown that HS acting as a coreceptor on the surface of host cells may affect the conformation of SARS‐CoV‐2 spike protein by forming ternary complexes, promoting virus infection. ^[9] These findings suggest that SARS‐CoV‐2 spike protein‐HS‐ACE2 interactions play a key role in the viral infection. Therefore, it has become an increasingly attractive approach to seek for competitive inhibitors of HS‐virus interaction to attenuate viral infection.

In this study, we constructed a disaccharide library of HS analogues, e. g. the basic disaccharide units (N‐acetylglucosamine, GlcNAc, glucuronic acid, GlcA, and iduronic acid, IdoA). To analyze interaction between the disaccharides and spike protein, we have developed pharmacophore models focused on the receptor binding domain (RBD) of spike protein in SARS‐CoV‐2 and SARS‐CoV‐2 Omicron BA.1 (referred to as the Omicron) variant and used the models to screen the disaccharide library. ^[10] The virtual screening has identified two disaccharide structures that exhibited higher affinity towards RBD than some of the reported potential antiviral drugs. Molecular docking analysis revealed the competitive mechanism of SARS‐CoV‐2 spike protein‐HS interactions by the disaccharides. Moreover, potentials of these disaccharides binding to inflammatory cytokines are evaluated by molecular docking.

Computational Methods

Preparation of disaccharide library

The library (in .sdf format) contains 3D structure of 96 heparan sulfate (HS) analogue‐ disaccharides (Table S1). Since HS is composed of alternating α‐D‐glucosamine linked to β‐D‐glucuronic or α‐L‐iduronic acid through 1,4 glycosidic bond, the basic structures of HS ([→4)‐β‐D‐GlcpA‐(1→4)‐α‐D‐GlcNAc‐(1→] and [→4)‐β‐D‐GlcpA‐(1→4)‐α‐D‐GlcNS‐(1→]) were created by Chem3D 17.0 (Figure 1). The sulfated disaccharide structures were obtained through optimizing each sugar to minimize molecules’ energy by Discovery Studio v4.5 ^[11] (Dassault Systemes, Biovia Corp., San Diego, CA, USA (accessed on 6 July 2021)) computer program. The energy minimization was performed using CHARMm force fields until the RMS gradient of 0.1 kcal/(mol×Å) was reached.

Figure 1.

Structure of the disaccharides. The carbon atom serial number is shown in the monosaccharide plane structure.

Structure modelling of RBD of SARS‐CoV‐2 Omicron variant

The X‐ray 3D crystal structure of spike protein (PDB ID: 6M0J) was retrieved (in .pdb format) from the Protein Data Bank (https://www1.rcsb.org (accessed on 31 January 2022)). The spike RBD of SARS‐CoV‐2/ SARS‐CoV‐2 Omicron variant were cleaned by DISCOVERY STUDIO v4.5. In the process, all the water molecules were removed, and the missing atoms of the incomplete amino acid side‐chain or backbone as well as the hydrogen atoms were added to the 3D structure. Omicron variant has a total of 36 mutations in spike proteins, including 15 mutations in the receptor‐binding (RBD) domain. ^[12] In this experiment, AlphaFold2 ^[13] (ColabFold: AlphaFold2 using MMseqs2 (accessed on 3 April 2022)) was used to predict spike RBD structure of SARS‐CoV‐2 Omicron variant. The 3D coordinates of all heavy atoms for a given protein using the primary amino acid sequence are predicted by AlphaFold2 network. The entire algorithm framework of AlphaFold2 learns the representation of multiple sequence alignment (MSA) and amino acid pair (pairwise) of proteins through collaborative learning, and then incorporates evolutionary information on protein sequences, physical and geometric constraints information on protein structures into deep learning networks. The quality of the model was evaluated by parameters predicting local distance difference test (pLDDT).

Pharmacophore modelling and virtual screening

The pharmacophore models were established in a structure (spike RBD of SARS‐CoV‐2/ SARS‐CoV‐2 Omicron variant)‐based manner, by probing possible interaction sites between the RBDs and ACE2. ^[9] The pharmacophore models were built based on ranks from validation of scoring parameters (sensitivity and specificity) by the internal rule‐based scoring function. which is a function of the feature set in the pharmacophore model and the feature‐feature distances of different types of features to define the maximum hydrogen bonding distance based on a Genetic Function Approximation (GFA) model. The first step was to locate the binding pocket of RBD for generating high‐quality pharmacophores. ^[14] Based on the published data, the spike protein could bind to heparin/HS on the pocket of RBD involving amino acid residues of Arg346, Phe347, Ala348, Ser349, Ala352, Trp353, Asn354, Arg355, Lys356, Arg357, Lys444, Asn448, Asn450, Tyr451, Arg466, Arg509. ^[9] For Omicron variant, the positively charged amino acids (Arg346, Asn354, Arg355, Lys356, Arg357, Lys444, Arg466, Arg509) ^[9] that are involved in binding to ACE2 were selected. Then, we created two structure‐based preliminary pharmacophore models that are featured with 8 hydrogen bond receptors, 8 hydrogen bond donors, 2 hydrophobic centers, and 15 exclusion volume models by the “Interaction Generation” tool in the “Pharmacophore” module DISCOVERY STUDIO v4.5 software package.

The 3D pharmacophore models based on the RBDs of spike protein in SARS‐CoV‐2 and SARS‐CoV‐2 Omicron variant were applied to virtual screening of the HS analogue disaccharide library. Through mapping the fit hits for each pharmacophore and molecule using the “Screen Library” tool in the “Pharmacophore” module with Discovery Studio v4.5 software package, the FitValue was generated and the molecules showed highest FitValue were selected (the Minimum Features parameter is set to 4, the Maximum Features parameter is set to 22, and the Maximum Subset of Pharmacophores parameter is set to 250).

Mapping of the disaccharide interaction proteins and molecular docking

The identified disaccharides were screened for binding to other proteins using the open‐accessed PharmMapper online server (http://www.lilab‐ecust.cn/pharmmapper (accessed on 4 May 2022)). ^[15] After uploading the best‐fit disaccharides files identified by the RBD of SARS‐CoV‐2 pharmacophore model (in .mol2 format) as test molecule, PharmMapper extracts targets from its large internal pharmacophore database and accesses to the receptor‐based pharmacophore models to identify potential binding proteins. The jobs were submitted with each JOB ID generated. The status and the result of the submitted job were checked using the JOB ID.

The molecular docking studies of the disaccharides against target proteins (spike RBD of SARS‐CoV‐2/SARS‐CoV‐2 Omicron variant and interleukin‐1β (IL‐1β, PDB ID: 8I1B) and interleukin‐10 (IL‐10, PDB ID: 1LK3) were performed utilizing Autodock VINA in YASARA (YASARA Biosciences GmbH, Vienna, VIE, AUT, (accessed on 20 May 2022)). ^[16] To remove bumps and ascertain the covalent geometry of the ligands, the structures were all energy‐minimized with the NOVA force field. ^[17] The blind dockings were undertaken by defining simulation cell boxes of size 66.86 Å×66.86 Å×66.86 Å for the spike RBD of SARS‐CoV‐2, 75.70 Å×75.70 Å×75.70 Å for the spike RBD of SARS‐CoV‐2 Omicron variant, 57.33 Å×57.33 Å×57.33 Å for IL‐1β, and 77.04 Å×77.04 Å×77.04 Å for IL‐10. Then the docking studies on the aptamer were carried out through the built‐in docking simulation macro ‘dock_run.mcr’ using AMBER03 force field ^[18] with 25 poses and 9 clusters for each situation.

Protein‐ligand interaction

After molecular docking study, the interactions between the target proteins and the disaccharides were further studied utilizing “Ligand Interactions” tool in the “Receptor‐Ligand Interactions” module with DISCOVERY STUDIO v4.5 software package. The type of the non‐bond interactions and amino acid residues as well as the distances and the dihedral angles are labelled for characterizing and visualizing ligand binding sites within proteins and the interactions between amino acid residues and disaccharides.

Molecular dynamics

To justify whether the models constructed from protein‐ligand docking are reliable, molecular dynamics (MD) simulations of binding between spike RBD of SARS‐CoV‐2 and SARS‐CoV‐2 Omicron variant, IL‐1β and IL‐10 with LBJ‐1 and LBJ‐2 were performed. The AMBER14 force field was utilized for the MD simulation as implemented in the YASARA program. For the treatment of long‐range coulomb forces beyond 8 Å cutoff, the MD simulation used periodic boundary conditions and the particle‐mesh Ewald method. NaCl was used at the concentration of 0.9 %, and the density of HOH was 0.997 g/mL in the MD cell. No restraints were applied during the MD simulation using the settings employed in the second equilibration dynamics. The energies and coordinates every 100 ps were saved at constant temperature (298 K) and pressure was uncontrolled in NVT ensemble. Structural stability of the RBD‐disaccharide and interleukin‐disaccharide complex was examined by analyzing the average values of potential energy with root mean square deviation (RMSD) throughout the trajectory. The RMSD profiles of all MD structures (Figure S1) show that the variation of the RMSD values tends to be stable after the long‐range dynamics simulation, which means that the structures at equilibrium have been obtained and the last MD structure was chosen for the analysis.

Quantum chemical analysis

The structural analysis of the disaccharides was carried out using Gaussian 09 (accessed on 8 June 2022), ^[19] while the orbitals were shown using Multiwfn (accessed on 12 June 2022) software. ^[20] The optimization of 3D structure of the disaccharides was performed by using the B3LYP‐D3/6‐31+G(d,p) level of density functional theory (DFT).

ADMET studies

The potential druggability of the disaccharides was evaluated by ADMET (absorption, distribution, metabolism, excretion and toxicity) using the online software program ADMETlab2.0 (https://admetmesh.scbdd.com/service/evaluation/index (accessed on 11 July 2022)). ADMETlab2.0 ^[21] utilizes a series of high‐quality prediction models trained by multi‐task graph attention framework to conveniently and efficiently implement the calculation and prediction the physicochemical properties, pharmacokinetics, drug‐likeness and drug chemical friendliness of the leading compounds. Molecules were uploaded in SMILES format. Then, the parameters including physicochemical properties, medicinal chemistry measures, and the ADMET behaviors are depicted by the structure of the disaccharides.

Results and Discussion

Characterization of the pharmacophore models

Figure 2 shows the alignment of spike protein RBD of SARS‐CoV‐2 and the Omicron BA.1 variant (the 15 mutations of Omicron are indicated with *). The structure of Omicron variant was virtually modelled by alphafold2. ^[22] The highest predicting local distance difference test (pLDDT) score (Table 1) indicates the satisfactory quality of the model (Figure 3).

Figure 2.

SARS‐CoV‐2 and its Omicron BA.1 variant spike RBD sequence alignment. * represents the mutation site.

Table 1.

Omicron RBD models and corresponding pLDDT score.

Omicron RBD model name	pLDDT
rank 1	78.3
rank 2	76.3
rank 3	75.6
rank 4	69.6
rank 5	69.1

Figure 3.

Omicron RBD homology model in details. a. Predicted IDDT per position. b. Model confidence measurement. The quality of the model and its corresponding pLDDT score. c. The predicted aligned error (PAE) is displayed as an image for each of the structure predictions.

The pharmacophore models of SARS‐CoV‐2 consisted of 32 features (denoted as F) including 8 hydrogen bond receptors (F1 and F2 interaction with R346, F3, and F4 interaction with A348, F5, F6, F7, and F8 corresponding to interaction with N354, K356, N448, and N450), 8 hydrogen bond donors (F9, F10, F11, F12, F13, F14, F15 and F16 corresponding to interaction with E340, R346, Y351, A352, N354, N450, Y508, and V511), 2 hydrophobes, and 14 exclusion volume models. Likewise, the pharmacophore models of the SARS‐CoV‐2 Omicron variant consisted of 24 features including 6 hydrogen bond receptors (F17, F18, F19, F20, F21, and F22 corresponding to interaction with R346, R509, N354, K356, S399, and R466), 5 hydrogen bond donors (F23, F24, F25, F26 and F27 corresponding to interaction with E340, R346, N354, R355, and S399), 2 hydrophobes, and 11 exclusion volume models (Figure 4).

Figure 4.

RBD pharmacophore models. a. SARS‐CoV‐2 spike protein RBD pharmacophore model. b. SARS‐CoV‐2 Omicron variant spike protein RBD pharmacophore model. The green arrows represent hydrogen bond receptors, purple arrows represent hydrogen bond donors, blue spheres represent hydrophobes, and grey spheres represent exclusion volume models.

Pharmacophore‐based virtual screening

For SARS‐CoV‐2 spike protein RBD, the disaccharides library was virtually screened based on 34 pharmacophores. the best hit disaccharide (named LBJ‐1, Figure 5a) is 2‐O‐sulfated‐GlcA‐β‐(1→4)‐3,6‐O‐sulfated‐GlcN, which represents GlcN with C3 and C6 hydroxyl group sulfonation, C4 links GlcA with C2 hydroxyl group sulfonation by a glycosidic bond. The features of LBJ‐1 included F2, F3, F8, F12 and F13. For Omicron variant, virtual screening based on 24 pharmacophores in the model identified the best hit (named LBJ‐2, Figure 5b) of 3,6‐O‐sulfated‐GlcNS‐α‐(1→4)‐2‐O‐sulfated‐GlcA, which represents GlcNS with C3 and C6 hydroxyl group sulfonation, C1 links GlcA with C2 hydroxyl group sulfonation through a glycosidic bond. The features of LBJ‐2 included F19, F20, F22, F24, F25 and F26. The results also indicated that each best hit compound showed a good binding affinity with the corresponding RBD proteins at their active sites. LBJ‐1 obtained the highest FitValue 3.26775 against SARS‐CoV‐2 RBD pharmacophore model, while LBJ‐2 obtained the highest FitValue 4.13885 against Omicron variant pharmacophore model, respectively.

Figure 5.

Structure of the best hit compounds. a. LBJ‐1, b. LBJ‐2.

Analysis of molecular interactions by molecular docking

To understand the interaction mechanisms, molecular docking ^[23] was performed. LBJ‐1 (the best hit for SARS‐CoV‐2 spike RBD pharmacophore) and LBJ‐2 (the best hit for Omicron SARS‐CoV‐2 spike RBD pharmacophore) were docked into the protein models (Figure 6), respectively. As a comparison, selected antiviral compounds (Table 2) were analyzed in parallel by molecular docking. The results indicate that LBJ‐2 binds more tightly with SARS‐CoV‐2 Omicron variant spike RBD (with the binding energy of −7.9770 kcal/mol) and SARS‐CoV‐2 spike RBD (with binding energy −6.4560 kcal/mol) than LBJ‐1 (with the binding energy of −7.1650 and −6.2170 kcal/mol, respectively). The results also show that both LBJ‐1 and LBJ‐2 have a higher binding activity with RBDs compared with the candidate drugs against SARS‐CoV‐2 (Table 2).

Figure 6.

Interactions of the disaccharides with RBDs. a. Interaction between LBJ‐1 and SARS‐CoV‐2 spike protein RBD. b. Interaction between LBJ‐2 and SARS‐CoV‐2 spike protein RBD. c. Interaction between LBJ‐1 and SARS‐CoV‐2 Omicron variant spike protein RBD. d. Interaction between LBJ‐2 and SARS‐CoV‐2 Omicron variant spike protein RBD. The green dashed lines represent conventional hydrogen bond, the pale green dashed lines represent carbon hydrogen bond or Pi‐donor hydrogen bond, and the orange ones represent attractive charge or Pi‐sulfur.

Table 2.

Results of molecular docking studies.

Ligand name	Binding free energies [kcal/mol]
	SARS‐CoV‐2 spike RBD Omicron variant	SARS‐CoV‐2 spike RBD	Interleukin‐10	Interleukin‐1β
LBJ‐2	−7.9770	−6.4560	−5.8180	−6.1700
remdesivir [25]	−7.8540	−6.3050	−7.7160	−6.3590
posaconazole [26]	−7.7090	−7.0720	−8.3810	−6.9730
vaniprevir [27]	−7.3490	−6.8920	−9.3040	−6.3750
LBJ‐1	−7.1650	−6.2170	−5.9630	−5.6570
cepharanthine [26]	−7.1540	−6.7460	−9.9530	−6.6390
grazoprevir [28]	−7.1230	−6.8230	−8.6280	−6.6150
emetine [29]	−7.1170	−6.0780	−8.1680	−5.7860
boceprevir [30]	−5.8580	−5.1540	−7.5690	−5.0220
6‐thioguanine [31]	−4.3880	−4.2110	−4.8790	−3.6430

Moreover, the results of molecular docking show that the interactions between the spike RBD of SARS‐CoV‐2/SARS‐CoV‐2 Omicron variant with the disaccharides mainly involve the sulfonic groups, e. g. the sulfonic groups (3S, 6S) and amino groups (NH₂) on GlcN unit, and sulfonic group (2S) on GlcA unit of LBJ‐1; and sulfonic groups (NS, 3S, 6S) and amine group (NH−R) on GlcNS unit, sulfonic group (2S) on GlcA unit for LBJ‐2. In addition, the hydroxyl groups on disaccharide skeletons also participate in the binding to the RBDs (Tables S2–S5).

Furthermore, the molecular docking study revealed the amino acid residues that interact with the antivirus drug candidates. For the Omicron RBD, Trp353 is involved in the interaction with boceprevir, grazoprevir, posaconazole, remdesivir, vaniprevir; Arg355 is involved in the interaction with boceprevir, cepharanthine, grazoprevir, posaconazole, remdesivir, vaniprevir; Phe464 is involved in the interaction with boceprevir, cepharanthine, grazoprevir, posaconazole, remdesivir; and Arg466 is involved in the interaction with boceprevir, grazoprevir, posaconazole vaniprevir (Figures S4–S11).

And earlier studies reported that the amino acid residues of Trp353, Arg355, and Arg466 of SARS‐CoV‐2 RBD are involved in interaction with receptor ACE2. ^[9] We further identified the involvement of amino acid Phe464 can be new target site for SARS‐CoV‐2 and its Omicron variant, and the potential for additional new mutants.

Since SARS‐CoV‐2 infection often causes inflammatory reactions, and it is known that a number of inflammatory cytokines interact with HS, we wanted to see whether the RBD‐binding disaccharides also bind to other proteins. In this approach, we probed the disaccharides against the largest collection of databases (http://www.lilab‐ecust.cn/pharmmapper) that contained 23236 proteins covering 16159 pharmacophore models predicted as druggable binding sites. Among the highest score, we selected two interleukins, IL‐1β and IL‐10, for molecular docking analysis. Results show that LBJ‐2 has a relatively higher binding affinity with IL‐1β (−6.1700 kcal/mol) and a lower binding affinity with IL‐10 (−5.8180 kcal/mol), while LBJ‐1 has a lower binding affinity to IL‐1β (−5.6570 kcal/mol) and a higher binding affinity to IL‐10 (−5.9630 kcal/mol). In comparison, almost all the small molecule drugs (except for 6‐thioguanine) had higher binding energy values with IL‐10 than both disaccharides; whereas emetine boceprevir and 6‐thioguanine had lower binding values with IL‐1β than LBJ‐2. This is of interesting as SARS‐CoV‐2 infection can lead to severe inflammatory reactions. So, binding of the disaccharides to both RBD and IL‐1β may indicate the possibility for attenuation of inflammation in COVID‐19 patients.

Molecular docking analysis shows that both disaccharides have higher binding activity compared with the candidate drugs against SARS‐CoV‐2. The reliability of the models constructed from protein‐ligand docking was verified by molecular dynamics (MD) simulations. Both methods gave similar complex structures. Quantitatively, the RMSD values of the structures from the MD and docking are small (Figure S1), which indicates that the structural variation is minor. Interestingly, we found that the disaccharide binding sites of RBDs from SARS‐CoV‐2 and its Omicron variant are shifted, which may be due to altered conformation of RBD caused by mutations. Nevertheless, both disaccharides are still located around the binding interface between ACE2 and RBDs of both Omicron variant and SARS‐CoV‐2 (Figure 7). Moreover, MD simulations results of interleukin and disaccharide indicate that LBJ‐1 and LBJ‐2 bind tightly to pro‐inflammatory cytokine IL‐1β through non‐bonded interactions under physiological conditions after stabilization (Figure S2a, Figure S2b) of the complex structure, which may inhibit its pro‐inflammatory effects mediated by binding to the IL‐1β receptor. Meanwhile, LBJ‐1 and LBJ‐2 interacted weakly with IL‐10 and moved away from IL‐10 after the complex structures were stabilized (Figure S2c, Figure S2d), and therefore might not affect the inhibitory effect of IL‐10 on inflammation mediated by its receptor binding (Figure S3).

Figure 7.

Superposition of two complexes including the last structures obtained from the MD and the complexes obtained from docking. The RMSD values of the structures aligned by PyMOL including the last structure from the MD (colored in light teal) and the complex obtained from docking (colored in wheat) are 1.192, 1.468, 1.591, and 1.805 respectively. a. SARS‐CoV‐2 spike protein RBD+LBJ‐1. b. SARS‐CoV‐2 spike protein RBD+LBJ‐2. c. SARS‐CoV‐2 Omicron variant spike protein RBD+LBJ‐1. d. SARS‐CoV‐2 Omicron variant spike protein RBD+LBJ‐2 obtained. And the amino acid residues of the binding interface between ACE2 and RBD are colored in light pink.

Since the outbreak of SARS‐CoV‐2, extensive studies on the virus inhibitory activity of heparin and its derivatives have been carried out. Studies have found that enoxaparin has higher activity in comparison with fully‐desulfated or N position desulfated enoxaparin, which demonstrates the necessity of sulfation. ^[24] Some studies have shown that the binding between oligosaccharides and SARS‐CoV‐2 spike protein RBD increased with the degree of 6‐O‐sulfated in glucosamine residues, which further illustrates the importance of 6‐O‐sulfation in determining the interaction between HS and SARS‐CoV‐2 spike protein RBD. ^[32] The identified disaccharide structures are well in the line with these findings.

Frontier molecular orbitals analysis of the disaccharides

To explore the chemical stability of the disaccharides, we performed frontier molecular orbitals (FMOs) analysis on LBJ‐1 and LBJ‐2. FMOs are usually referred to the highest occupied molecular orbital and lowest unoccupied molecular orbital. The structural properties of ligands used for developing the docking interaction with selected protein molecules have significant importance in explaining the density distribution pattern on frontier molecular orbitals. ^[33] The energetics of FMOs are crucial to the reactivities and chemical stability of ligand molecules. The electronic and structural properties of the selected ligands are computed by optimizing their structures using the B3LYP−D3/6‐31+G(d,p) level of density functional theory. The HOMO−LUMO energy difference is important in determining the reactivity and stability of molecule. The energetic results of FMOs show that the ▵E _L‐H values of HOMO‐LUMO gap of LBJ‐1 and LBJ‐2 are 6.60 and 6.37 eV, representing the structures of LBJ‐1 and LBJ‐2 are stable (Table 3 and Figure 8).

Table 3.

FMOs results of the LBJ‐1 and LBJ‐2.

Ligand name	HOMO (eV)	LUMO (eV)	ΔEL‐H=ELUMO‐EHOMO (eV)
LBJ‐1	−6.733997	−0.138265	6.595732
LBJ‐2	−6.368412	0.006074	6.374486

Figure 8.

The FMOs including HOMO and LUMO for the disaccharides as calculated at B3LYP‐D3/6‐31+G(d,p) level of DFT.

Druggability assessment by ADMET studies

Having seen the high affinity interaction of the disaccharides with RBD proteins as well as the cytokines, we were curious about the druggability of the compounds. The ADMET analysis shows similar physicochemical properties of LBJ‐1 and LBJ‐2 (Table 4). Both disaccharides possess a high TPSA due to the polar groups of hydroxyl and sulfonic oxygens as well as the amino and amine nitrogen. Furthermore, the disaccharides are considered to have a proper Plasma Protein Binding (PPB) because of the predicted value <90 % which may lead to a relatively high therapeutic index. The results also indicate their proper volume distribution (in the range of 0.04–20 L/kg) as well as the poor blood‐brain barrier permeability which might be required to avoid central nervous system side effects. Nevertheless, it is predicted none of the compounds will affect the five major isoforms of cytochrome P450 (CYP) including CYP1A2, CYP2C19, CYP2C9, CYP2D6, and CYP3A4. ^[34] The toxicity prediction results indicated that both disaccharides show negative responses in AMES toxicity, which means a negative mutagenicity and carcinogenicity response. And the rat oral acute toxicity result which is related to acute mammalian toxicity also shows low‐toxicity. The results also showed that disaccharides are non‐corrosive/non‐irritant chemicals to eyes. Furthermore, by Pfizer Rule, ^[35] compounds with a low log P (≤3) and high TPSA (≥75) are likely to be non‐toxic. And synthetic accessibility score (SAscore) ^[36] is designed to estimate ease of synthesis of drug‐like molecules, based on a combination of fragment contributions and a complexity penalty, and compounds with a low SAscore (<6) are likely easy to synthesize. Thus, both sugars are predicated as drug‐likeness compounds by ADMETlab2.0 through Pfizer Rule and SAscore.

Table 4.

Physiochemical parameters of LBJ‐1 and LBJ‐2.

Parameters	LBJ‐1	LBJ‐2
No. of H bond acceptors	21	24
No. of H bond donors	10	9
Topological Polar Surface area, TPSA ([Å]2)	345.66	382.88
Lipophilicity, log P	−4.482	−4.647
Water Solubility, log S	1.648	2.182
SAscore	5.215	5.179
Pfizer Rule	Accepted	Accepted
Plasma Protein Binding (PPB)	36.32 %	51.27 %
Volume Distribution (VD) (L/kg)	0.521	0.454
Blood‐Brain Barrier (BBB) Penetration	0.043	0.002
CYP1A2 inhibitor	No	No
CYP2C19 inhibitor	No	No
CYP2C9 inhibitor	No	No
CYP2D6 inhibitor	No	No
CYP3A4 inhibitor	No	No
AMES Toxicity	Negative	Negative
Rat Oral Acute Toxicity	Low‐toxicity	Low‐toxicity
Eye Corrosion/Irritation	No	No
Drug‐likeness	Yes	Yes

Despite the good druggability of our screened disaccharides, in wet lab, the organic synthesis of HS analogues is complicated, including chemical protection, activation, coupling, and de‐protection, which often require much optimization and customization for each step. And the control of product formation is less than ideal. Meanwhile, the enzyme tools (such as glucuronyl C5‐epimerase, hexuronyl 2‐O‐sulfotransferase, glucosaminyl 6‐O‐sulfotransferase, and glucosaminyl 3‐O‐sulfotransferase) are used together in the HS analogues formulation, which has exquisite stereoselectivity and regio‐selectivity for connecting monosaccharide units but lack the precise control in introducing O‐sulfo groups and epimerizing uronic acid residues. Therefore, it is challenging to synthesize HS analogues as our screened disaccharides. Guided by this computationally determined mechanism, we hope that our work could provide some evidence for other research groups involved in polysaccharide synthesis to achieve HS analogues with the parts of our designed disaccharides allowing working in the active binding pocket of the targeting proteins to favor intracellular conformations.

Conclusion

In this study, aimed at exploring sugar‐based anti‐SARS‐CoV‐2 drug candidates, we have applied several software tools to screen a HS‐analogue disaccharide library that covers all possible structures of 1–4 linked glucosamine and hexuronic acid with and without sulfate substitutions. Using the information from database, pharmacophore models of RBDs of SARS‐CoV‐2 as well as its Omicron variant were built and used for virtual screening. Screening of 96 different structures of disaccharides identified two disaccharides that bind to the RBDs pharmacophore with high affinity. Moreover, molecular docking of these two disaccharides to a pharmacophore model database found that the disaccharides also interacted with IL‐1β and IL‐10. The multiple interactions and druggable properties of the disaccharides should be of interest for further experimental biology studies. It should be highlighted that heparin is composed of the same sugar units as LBJ‐1 and LBJ‐2, so it should be highly relevant to identify the fine structures in heparin that exhibited SARS‐CoV‐2 inhibitory activity. ^[37] Collectively, the accumulated studies support the antiviral potential of hydroxy‐sulfonated oligo‐ and polysaccharides, while disaccharides should be the minimum active molecules. Computer aided study is an efficient and economical way for rational design of drug candidates thus, this established procedure may be applied to a broad‐spectrum study to explore potential drug candidates of other saccharides and HS‐mimetics.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Li B., Zhang T., Li J.-p., Yu M.-j., ChemBioChem 2022, 23, e202200461.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.