Marine Sulfated Polysaccharides as Promising Antiviral Agents: A Comprehensive Report and Modeling Study Focusing on SARS CoV-2

SARS-CoV-2 (severe acute respiratory syndrome coronavirus-2) is a novel coronavirus strain that emerged at the end of 2019, causing millions of deaths so far. Despite enormous efforts being made through various drug discovery campaigns, there is still a desperate need for treatments with high efficacy and selectivity. Recently, marine sulfated polysaccharides (MSPs) have earned significant attention and are widely examined against many viral infections. This article attempted to produce a comprehensive report about MSPs from different marine sources alongside their antiviral effects against various viral species covering the last 25 years of research articles. Additionally, these reported MSPs were subjected to molecular docking and dynamic simulation experiments to ascertain potential interactions with both the receptor-binding domain (RBD) of SARS CoV-2’s spike protein (S-protein) and human angiotensin-converting enzyme-2 (ACE2). The possible binding sites on both S-protein’s RBD and ACE2 were determined based on how they bind to heparin, which has been reported to exhibit significant antiviral activity against SARS CoV-2 through binding to RBD, preventing the virus from affecting ACE2. Moreover, our modeling results illustrate that heparin can also bind to and block ACE2, acting as a competitor and protective agent against SARS CoV-2 infection. Nine of the investigated MSPs candidates exhibited promising results, taking into consideration the newly emerged SARS CoV-2 variants, of which five were not previously reported to exert antiviral activity against SARS CoV-2, including sulfated galactofucan (1), sulfated polymannuroguluronate (SPMG) (2), sulfated mannan (3), sulfated heterorhamnan (8), and chondroitin sulfate E (CS-E) (9). These results shed light on the importance of sulfated polysaccharides as potential SARS-CoV-2 inhibitors.


Introduction
Mother Nature remains an outstanding hub for valuable natural compounds that have been used for multiple purposes since ancient times by our ancestors, including pharmaceutical, biomedical, nutritional, and cosmetics applications. Among the dozens of natural product groups, polysaccharides are macromolecular polymeric carbohydrate molecules comprised of large chains of monosaccharide units. They are widely found in animals, plants, and microorganisms, and their functions are mainly either structureor storage-related. Many studies have revealed that natural polysaccharides and their chemically modified derivatives have significant inhibitory activities against viral diseases such as human immunodeficiency virus (HIV) and herpes simplex virus (HSV) [1,2].
Sulfated polysaccharides (SPs) are a class of negatively charged polysaccharides comprising natural or modified sulfate moieties in their structural carbohydrate backbone. They possess significant biological activities such as antioxidant, anti-allergic, antiviral, anticancer, and anticoagulant abilities; hence, the study of SPs has significant importance for drug discovery campaigns [1,3]. SPs are mainly found in the cell walls of marine algae or seaweeds; they are less common in some mammals, such as fish skins, and rare in mangrove plants. The seaweed cell wall comprises about 40% of sulfated polysaccharides, which is relatively higher than the average content in other sources. The most interesting marine algal SPs are sourced from ulvans from green macroalgae, carrageenans and agar from red macroalgae, and fucoidans and laminarians from brown macroalgae. These SPs have shown antiviral activity against herpes simplex virus (HSV), human immunodeficiency virus type-1 (HIV-1), chikungunya virus, cytomegalovirus (CMV), influenza virus, and hepatitis virus, in addition to other enveloped and non-enveloped viruses. Moreover, the antiviral activity of SPs against the current COVID-19 pandemic has also been reported [1,4,5].
Marine sulfated polysaccharides (MSPs) have recently received increasing attention due to their antiviral activity. In particular, carrageenans have exhibited promising inhibitory effects on many viral strains, effectively preventing the internalization of virus particles by interfering with the interactions between the virus and host cell receptors. Carrageenan nasal spray (Boots Dual Defence ® in the UK market) has been proven effective in patients with common cold infected by human coronaviruses beta and alpha. In contrast, the impact was attributed to the increased viral clearance and the reduced relapses of symptoms in children and adults [6]. Another nasal spray formulation (xylometazoline HCl) containing iota-carrageenan (4) can effectively relieve nasal congestion of the upper respiratory tract and protect the respiratory mucosa against viral infection [7]. Coldamaris ® lozenges, comprising iota-carrageenan (4) as the active pharmaceutical ingredient, cause denaturation of glycoproteins on the coronavirus surface, inhibiting the virucidal effects of coronavirus [8].
It remains challenging to develop novel drugs within a limited timeframe, as drug discovery is time-and resource-consuming; however, the process of drug discovery is immensely enhanced by the modern era of computational technology, which has accelerated drug discovery and drug repurposing. Computer-assisted or in silico design employs computational methods in drug discovery and is being applied to streamline and speed up hit-to-lead optimization and hit identification. Several computational techniques have been introduced to predict and select therapeutic targets, study the interactions between drug and receptor, characterize and determine ligand binding sites on the targets, and determine hit compounds using ligand-and structure-based virtual screening [9,10].
Molecular docking and dynamic simulations are the most intriguing among the computational tools employed for structure-based drug discovery, an approach that evaluates the binding affinities between two candidates: small molecules and macromolecular targets (protein). Many drugs currently in the market were developed based on in silico strategies, such as zanamivir (used to treat influenza), nelfinavir, and saquinavir (used in the treatment of HIV); therefore, computational methods have been receiving popularity in the pharmaceutical industry as being crucial in the drug discovery process as reliable and effective techniques [11][12][13].
In the present investigation, we aimed to shed light on the antiviral potential of MSPs, illustrating their modes of interaction against different targets with the aid of several in silico tools, highlighting the most potential candidates for further investigation against SARS CoV-2.

Results and Discussion
Herein, we have summarized the findings of approximately 80 research studies published in the last 25 years on MSPs that exhibited potential antiviral effects against a total of 22 different viral strains (Table 1, Figure S1). These antiviral MSPs were sourced from various marine sources, including brown algae, red algae, green algae, blue-green algae, microalgae, sea cucumber, and squid cartilage. According to the findings, 40% of the antiviral MSPs were isolated from the red algae, followed by 24% from brown algae, 14% from green algae, 10% from blue-green algae, 9% from microalgae, 2% from squid cartilage, and 1% from sea cucumber (Figure 1). The red-algae-derived MSPs exhibited greater efficacy against several viruses; however, the distribution was highly skewed towards herpes simplex viruses (HSV), which received most of the research effort ( Figure S1).

MSPs from Brown Algae
Brown algae come in second place after red algae in terms of antiviral activities and research attention (Figure 1). The main reported MSPs from this genus were fucoidans, a group of heterogeneous sulfated polysaccharides that represent about 25-30% of the algae dry weight and are made up of a backbone of α-(1→3)-L-fucopyranose residues or alternating α-(1→3) and α-(1→4)-linked L-fucopyranosyls with a sulfate group mainly substituted on C-2 or C-4 fucopyranose residues [4]. Fucoidans were found to be effective against 7 different viruses (HSV, SARS-CoV-2, InfAV, HIV-1, DENV-2, NDV, and HCMV) (Table 1, Figure 3 and Figure S1). Generally, fucoidans block viral infection by preventing viral entry through competing for the positive charge attachment site of the envelope glycoproteins. The extent of antiviral activity is related to the number of sulfate groups present in the fucoidan structure [102]. As illustrated in Table 1, fucoidans have been tested in vivo using mice or mouse models for their activity against InfAV and HSV-2 and showed promising results. A recent in vitro study of two different fucoidans revealed that these compounds could be potent inhibitors of SARS-CoV-2 [17,18].

MSPs from Green Algae
Green-algae-derived MSPs come in third place as the most-investigated algae. Most of the examined compounds for antiviral activities belong to Ulvanes (Monostroma, Ulva, Enteromorpha), Codium, and Caulerpa. Different green-algae-derived MSPs were studied in vitro and in vivo against 11 viruses, including avian InfAV, HSV, HCMV, DENV-2, EV71, MeV, NDV, MuV, hCV, HIV, and JEV (Table 1, Figure 4). Song and his colleagues examined the antiviral activity of Ulva pertusa against avian InfAV and found that it exhibited a mild antiviral effect (40% viral inhibition). When combined with a vaccine against the same virus, it generated a synergistic effect, which significantly enhanced the production of antibodies by more than two-fold (~100%) [86]. Although the antiviral activities of green algae have received less attention than the red and brown, they possess unique antiviral properties against diverse viruses, such as NDV [78] and JEV [85].

MSPs from Miscellaneous Marine Sources
Blue-green algae, microalgae, sea cucumbers, and squid cartilage (Table 1, Figure 4) are studied much less than the former algae for their MSPs as potential antiviral agents. Among them, the novel sulfated polysaccharide calcium spirulan (Ca-SP) was isolated from the blue-green alga Spirulina platensis and found to be an inhibitor for several viruses, including HSV-1, HCMV, InfAV, MeV, HIV-1, and Muv [88]. Recently, a clinical trial confirmed the potential of Ca-SP (Spirularin ® HS) against herpes viruses through inhibiting the attachment and penetration of HSV-1 into mammalian epithelial cells and blocking the entry of Kaposi sarcoma-associated herpesvirus HSV-8 [90]. The highly sulfated polysaccharide p-KG03 isolated from the microalgae Gyrodinium impudium showed potent inhibitory activity against InfAV by blocking the early stage of replication and entry [95]. Sea cucumber sulfated polysaccharide (SCSP) is one of the most recent MSPs extracted from the sea cucumber Stichopus japonicus. Song and colleagues confirmed its activity against SARS-CoV-2 in an in vitro study through binding to the S-glycoprotein, preventing SARS-CoV-2 host cell entry [18].

In Silico Investigation of MSPs against SARS CoV-2
Molecular modeling of polysaccharides is not an easy task due to their high diversity, complexity, and flexibility; however, the recent advances with in silico tools can relieve the complexity of the process to a greater extent. Recently, heparin (10), an example of sulfated polysaccharide, was reported to exhibit a promising antiviral activity against SARS CoV-2 (EC 50 = 36 µg/mL) through S-protein binding; hence, it suppresses the viral attachment to ACE2 (SARS CoV-2 S-protein receptor) and subsequently its entry inside the host cell ( Figure 5) [17,103]. Similarly, most of the reported MSPs were found to mediate their antiviral activity via the exact mechanism, particularly those reported as anti-SARS CoV-2. Accordingly, in this study, we shed light on MSPs as potential SARS CoV-2 antiviral agents by speculating on their plausible mode of action at the molecular level using a series of molecular docking and dynamic simulation experiments. Firstly, we determined the possible binding sites on both the S-protein receptor-binding domain (S-RBD) and ACE2 for MSPs. To do so, we utilized ClusPro [105], a software specialized in the prediction of heparin (10) (i.e., as an example of sulfated polysaccharide) binding sites in any given protein via molecular docking. The predicted heparin-S-RBD and heparin-ACE2 complexes were then subjected to 50 ns molecular dynamic simulation (MDS) experiments to select the most stable binding modes with each protein (S-RBD and ACE2). As shown in Figure 6, heparin was predicted to achieve stable binding with S-RBD at two sites (sites 1 and 2). Site 1 is located in a region that can interact with ACE2 directly and has a moderate positive charge. In contrast, site 2 is shallower than site 1 and is located in a wide positively charged region. Heparin-S-RBD complexes in these two sites were significantly stable over the course of MDS with low deviations from the original poses (RMSD~2.5 Å) and minimal fluctuations (site 1 and site 2, Figure 6). This apparent stability resulted from the networks of H-bonds and ionic interactions formed between heparin and each binding site (Tables 2 and 3).  On the other hand, heparin was predicted to interact with a small pocket (site 3) on ACE2 located near the S-RBD binding region ( Figure 6). Similar to sites 1 and 2, heparin achieved stable binding with site 3 during the MDS (RMSD~1.9 Å) through extensive H-bonds and ionic interactions (site 3, Table 4 and Figure 6).  The complex stability of RDB-ACE2 was also studied upon heparin binding to each site. As depicted in Figure 4, the distance between S-RBD and ACE2 (calculated as the distance between GLN-493 and GLU-35, respectively) remained constant (~5.1 Å) over 30 ns of MDS, while binding of heparin to S-RBD or ACE2 via sites 1, 2, or 3 led to significant instability of the complex and gradual dissociation (i.e., increased distance between GLN-493 and GLU-35). Heparin binding to site 1 on S-RBD or site 3 on ACE2 showed the greatest effects as S-RBD dissociated almost completely from ACE2 after 20 ns of simulation (Figure 7). Accordingly, it can be concluded that stable binding of SPs to sites 1, 2, or 3 destabilizes the S-RBD-ACE2 complex. It was recently reported that heparin can bind to S-RBD, preventing SARS CoV-2 from reaching ACE2 and entering the host cell; however, according to our modeling results, heparin can also bind to and block ACE2, acting as a protective agent against SARS CoV-2 infection. To study the mode of action of all reported MSPs against SARS CoV-2 to suggest previously unreported candidates, we subjected all collected compounds (Table 1) to dock against the proposed binding sites (sites 1, 2, and 3). Top-scoring hits (Figure 8) were selected according to the following criteria: (i) Docking score < −5 kcal/mol. Scores >−5 (i.e., −4.9 to −1.2 kcal/mol) showed unstable binding with the corresponding protein (RMSD > 20 Å at the first 20 ns); (ii) ∆G value < −5 kcal/mol, (iii) the docking pose remains stable over 50 ns of MDS. Some compounds achieved docking scores <−5 kcal mol (e.g., −6.4 kcal mol), although they were significantly unstable during the MDS experiments (RMSD > 15 Å at the first 20 ns) and presented significantly higher ∆G values (~−1.3 kcal/mol). As such, we made the binding stability at least 50 ns alongside ∆G value < −5 kcal/mol another selection criterion to discriminate between binders from non-binders, ensuring that all selected hits can achieve stable binding with the corresponding protein. Additionally, we took into consideration the newly emerged SARS CoV-2 variants during our docking experiments. Upon reviewing the recent mutations of the viral Sprotein, we found that the UK variant (i.e., B.1.1.7) has a mutation in its S-RBD (i.e., N501Y) that was associated with its higher affinity to the ACE2 receptor [106]. This mutation was the replacement of the amino acid ASP-501 with TYR. We applied this mutation to the S-RBD structure to study its effect on the binding with MSPs upon docking and MDS. Table 2, docking against site 1 (with and without the N501Y mutation) resulted in nine SPs with scores <−5 kcal/mol (Figure 8). Four of these hits were previously reported to exert antiviral activity against SARS CoV-2. Further MDS and ∆G experiments revealed that the best binding modes for iota-carrageenan (4), kappa-carrageenan (6), sulfated galactan (7), and sulfated heterorhamnan (8) (Figure 8) with site 1 were significantly unstable (their average RMSD values were higher than 15 Å and their ∆G values were higher than −5 kcal/mol) and they could easily dissociate from site 1 (with and without the N501Y mutation). Sulfated galactofucan (1), sulfated mannan (3), and chondroitin sulphate E (CS-E) (9) were among the best-scoring hits that were significantly stable during the MDS experiments, either with the mutated or non-mutated RBD. Additionally, they had not been previously reported for their antiviral efficacy against SARS CoV-2; hence, they were considered good candidates for future evaluation.

As shown in
As shown in Figures 9-14, the prevalent interactions of these MSPs and RBD site 1 in its mutated and non-mutated forms were H-bonds and water bridges, where the sulfate esters were the main contributors in these interactions with site 1 key amino acid residues (Table 2). Moreover, their fluctuations and deviations from the starting docking poses were almost identical to both the mutated and non-mutated form of site 1 during the MDS, except for chondroitin sulphate E, which achieved slightly more stability and less deviation from the starting binding pose upon binding to the non-mutated form of site 1 ( Figure 13).
Regarding site 2, only five sulfated polysaccharides presented docking scores <−5 kcal/mol, of which iota-carrageenan (4) was unstable on MDS (got an average RMSD >15 Å) and presented a ∆G value higher than −5 kcal/mol (i.e., 1.7 kcal/mol). Sulfated galactofucan (1), sulfated polymannuroguluronate (SPMG) (2), and sulfated mannan (3) were among the best MSPs that achieved stable binding with S-RBD site 2 and were not previously reported against inhibitory effects of SARS CoV-2 (Table 3). In this site (i.e., site 2), ionic interactions between the negatively charged sulfate moieties and the positively charged key amino acid residues of this site were crucial, along with the other polar interactions (e.g., H-bonding and water bridging), in maintaining stable binding with such shallow binding site (Figures 15-17).
On the other hand, docking against the ACE2 binding site (site 3) revealed that both (SPMG) and lambda-carrageenan were the best-performing MSPs. Additionally, they were also stable over the course of MDS, achieving low deviations from their starting binding poses (RMSD~2.9 Å) and ∆G values <−5 kcal/mol (~−5.2 kcal/mol). Lambdacarrageenan was previously identified to bind to SARS CoV-2 S-protein [44]; however, there have been no reports on the ACE2 binding potential of MSPs so far. As such, focusing on this human protein target will be of great interest in future investigations of SPs as anti-COVID-19 therapeutics. Similarly to site 1 and 2, the dominant interactions between these MSPs and the key amino acid residues inside site 3 were also of the polar type (e.g., H-bonds, water bridges, and ionic interactions) (Figures 18 and 19).

Structure-Activity Relationship
The structural complexity of SPs has made the study of their structure-activity relationship quite challenging, and until now it has not been entirely understood; however, certain key bioactivity-related structural features could be concluded from the previously reported studies and the present modeling study.
The first structural determinant that may affect the antiviral potential of this class of compounds is the number of negatively charged groups (e.g., sulfates or carboxylates). Polysaccharides with higher numbers of sulfate or carboxylate groups per monosaccharide were more bioactive. Highly sulfated glucans (either natural or chemically synthesized derivatives) were found to be far more bioactive as antiviral agents than those with lower degrees of sulfation, while non-sulfated glucans were practically inactive [107,108]. Moreover, in our modeling study, MSPs with a single sulfate or carboxylate group per monosaccharide did not achieve stable binding with either S-RBD or ACE2 (∆G values > −5 kcal/mol); hence, we could conclude that electrostatic and other polar interactions mediated by these negatively charged groups play a significant role in stabilizing the binding of this class of compounds to S-RBD and ACE2.         Secondly, the distribution of these negatively charged moieties played an essential role in their antiviral activity. In previous studies, carrageenans with similar degrees of sulfation (50 mol%) but isolated from different sources showed varying antiviral activities [39,109,110]. Different charge densities were proposed to explain these findings, and accordingly to explain the antiviral activity [71]. Our modeling study found that neither iota-nor kappa-carrageenan achieved stable binding with any proposed binding sites. This could be attributed to their relatively low degree of sulfation (one sulfate group per onosaccharide); however, both were reported to be bioactive against SARS CoV-2 and other pathogenic viruses. This could be explained by the unique distribution of these limited sulfate groups, which might cause them to be bioactive via different modes of action. In contrast, sulfated galactan (7) has a high degree of sulfation ( Figure 8); however, it did not achieve stable binding with the proposed binding sites (∆G value >−5 kcal/mol). Accordingly, the presence of many negatively charged moieties does not guarantee good antiviral activity without the proper orientations and distribution over the polysaccharide backbone.  Finally, the general chemical structure of the polysaccharides (e.g., the stereochemistry of the glycosidic linkage and that of monosaccharides and the presence of branching points in the main backbone) has a general impact on the degree of their antiviral activity [111].

Molecular Modeling
We utilized the carbohydrate modeling protocol used by Sapay and his co-workers in our in silico study of MSPs [112].

Molecular Docking
The docking study was carried out against both the receptor-binding domain (RBD) of SARS CoV-2 s spike protein (S-protein) and human angiotensin-converting enzyme-2 (ACE2) (PDB codes: 6lZG and 1R42, respectively) [113,114]. An Autodock Vina docking machine [105] was used for docking experiments. To determine the best binding site for docking, we docked heparin (i.e., sulphated poly saccharide analogue) against both S-protein and ACE2 using the ClusPro server [14] by uploading each crystal structure to the website and choosing heparin as a ligand. The retrieved docking poses were then arranged from the highest to the lowest scores. We selected the docking pose of the highest score and subjected it to a short molecular dynamic simulation (25 ns) experiment to study the stability of heparin in this pose. Heparin showed acceptable stability in this orientation, where it deviated from the initial pose by an average RMSD of 2.1 Å ( Figure 6); hence, the binding site of heparin in this pose was used for docking in the subsequent docking experiments. Heparin docking was carried out using its tetrasaccharide form (i.e., the default form in the ClusPro software). Accordingly, we also used the tetrasaccharide form of each MSP for docking experiments to reduce the computational cost required for docking and molecular dynamic simulations. All results were visualized using PyMol software [115].

Molecular Dynamic Simulation
MD simulation experiments were carried out using the MDS machine in Maestro software, Desmond v. 2. [116][117][118], using its default force field (i.e., OPLS-AA). Proteinligand systems were constructed using the System Builder function. Thereafter, these systems were embedded in an orthorhombic box consisting of TIP3P water and 0.15 M Na+ and Cl-ions (the default dimensions were used). Subsequently, the prepared systems were energy-minimized and equilibrated for 10 ns. Ligand parameterization was carried out during the system building step according to the OPLS force field. For MD simulations carried out using NAMD software, the parameters and topologies of the ligands were computed using the Charmm36 force field. The online software Ligand Reader and Modeler (http://www.charmm-gui.org/?doc=input/ligandrm, accessed on 15 May 2021) [119] and the VMD plugin Force Field Toolkit (ffTK) [120] were utilized for this regard. The generated parameters and topology files were then loaded into VMD to read the proteinligand complexes without errors, then the simulation step was performed. MD simulations were run for 50 ns at 310 K in the NPT ensemble with the Nose-Hoover thermostat and Martyna-Tobias-Klein barostat using anisotropic coupling. We selected the best binding poses for each compound as starting co-ordinates to investigate their binding stability and mode of interaction.

Binding Free Energy Calculations
Binding free energy calculations (∆G) were performed using the free energy perturbation (FEP) method [121]. Briefly, this method estimates the binding free energy (i.e., ∆G binding ) according to the following equation: ∆G binding = ∆G Complex − ∆G Ligand . These estimations were derived from separate simulations (NAMD software was used for these experiments). All input files required for simulation by NAMD were papered using the online website Charmm-GUI (https://charmm-gui.org/?doc=input/afes.abinding, accessed on 30 April 2021). Subsequently, we loaded these files into NAMD in order to produce the required simulations. The FEP function in NAMD was used to accomplish this experiment. The equilibration step was achieved in the NPT ensemble at 300 K and 1 atm (1.01325 bar) with Langevin piston pressure (for "complex" and "ligand") in the presence of the TIP3P water model. Then, 10 ns FEP simulations were carried out for each ligand and the last 5 ns of the free energy values was measured for the final free energy estimation [121]. All resulting trajectories were visualized and analyzed by VMD software.

Conclusions
To challenge highly virulent SARS CoV-2 and its emerging mutations, which have cost millions of lives, there is an urgent need to ramp up drug discovery pipelines within a short timeframe. In order to shorten the drug development cycle, screening of existing chemical libraries holds greater potential than starting from scratch. The levels of structural and functional diversity of MSPs have been widely discussed for their antiviral properties; however, the depth of these activities in terms of possible mechanisms against potential drug targets has been poorly studied. A large number of members belong to the class of SPs being overlooked as potential candidates to suppress the virulence of SARS CoV-2.
In the present investigation, we aimed to point out the most probable binding sites on S-protein's RBD and ACE2 using a well-known sulfated polysaccharide (i.e., heparin) as a molecular probe. Interestingly, the binding of heparin to these sites led to dissociation of the RBD-ACE2 complex. These MDS-derived results indicated and validated these proposed binding sites as potential targets for further ligands from the same chemical class. Accordingly, these identified binding sites were used in a series of docking experiments using most of the collected MSPs in this report. A number of MSPs were then found to be potential binders to these sites. Most of the identified hits have been reported to exert antiviral activity against SARS CoV-2; however, some of them have not, meaning they are considered very promising candidates for experimental testing. Overall, the present investigation shed light on the huge potential of MSPs as antiviral chemical entities and provided the scientific community hints about their potential against SARS CoV-2, with some important structural information to be utilized in further experimental research.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/md19080406/s1, Figure S1-S8 indicating details about the reported SMPs in addition to the reported SMPs that were reported in the manuscripts.