Potential antifungal activity of novel carbohydrate derivatives validated by POM, molecular docking and molecular dynamic simulations analyses

ABSTRACT Our current study focuses on the molecular structures of the series of methyl 4,6-O-benzylidene-α-D-glucopyranoside 1–9 as potential antifungal agents to obtain more insight into the origin of their encouraging bioactivity. In vitro antimicrobial testing and the prediction of activity spectra for substances revealed that these carbohydrate derivatives have promising antifungal functionality compared to their antibacterial activities. In support of this finding, molecular docking was performed against lanosterol 14α-demethylase (CYP51A1) and Altemaria alternata (6LCC) by screening carbohydrate derivatives whereby significant binding affinities and non-bonding interactions have been observed against both microbial proteins. The majority of the derivatives studied here could bind near the critical catalytic residues. The molecular dynamics study has revealed that the complexes formed by these derivatives with the proteins lanosterol 14α-demethylase and Altemaria alternata can remain stable, both over time and in physiological conditions. The POM analysis shows the clear presence of an antifungal (O1δ-—O2δ-) pharmacophore site. ADMET and QSAR predictions were analysed to assess their pharmacokinetic and drug-likeness properties, showing promising results in silico. This work demonstrates that potential carbohydrate derivatives bind to fungal pathogens in an effort to circumvent their activities and open avenues for the development of newer antibiotics, which may target fungal pathogens.

Bioactive carbohydrate derivatives; antifungal agents; human and plant pathogens; POM (Petra/ Osiris/Molinspiration) analyses; identification of antifungal pharmacophore site Highlights . In the present study, carbohydrate derivatives were investigated to find potential antifungal agents through POM, in vitro biological and in silico computational approaches. . In vitro antimicrobial experiment revealed that alkyl chain and aromatic substituents can improve the antimicrobial efficacy of the carbohydrate structure which was also supported by PASS enumeration. . A molecular docking study against fungal targets exhibited a promising binding score and interaction in the catalytic active site. . A 100 ns MD simulation revealed the stable conformation and binding pattern in a stimulating environment of the studied derivatives. . POM analyses were used to calculate the physicochemical parameters of compounds and showed the presence of an antifungal (O1 d--O2 d-) pharmacophore site. . ADMET and QSAR properties have been calculated and provide safe use and low toxicity for both aquatic and non-aquatic species.

Introduction
Microbial infections affect individuals worldwide, leading to severe ailments that have serious consequences for public health. The need for novel antimicrobials, particularly antifungals, is increasing because of the current microbial resistance against major antimicrobials classes. Clinically studies have typically used antifungal agents, which are limited to the major types, for instance, azoles, which avert ergosterol synthesis, and polyenes. These combine with ergosterol in the fungal cell membrane, whereas echinocandins inhibit β-1,3glucose biosynthesis, and flucytosine which may inhibit DNA synthesis [1]. Since there are severe limitations in the clinical uses of these antifungal candidates, the development of brand new, effective, safe antifungal drugs, ones derived from novel antifungal sources, is of great and immediate importance [2]. Recently, the globe is facing unprecedented obstacles such as limited availability of drugs, inevitable toxicities, exorbitant prices, and the increasing emergence of drug resistance. Moreover, current researchers have found that nutrient metabolism disorders are finely connected to fungal virulence [3]. All microbial pathogens depend on gaining efficient nutrients to produce energy for cellular homeostasis, and glucose is one of the most important nutrients. Hence, investigations on glucose metabolism have also come to the forefront [4].
Carbohydrates are vital molecules that play important roles in several biological processes, including bacterial infections, cell growth and proliferation, cell-cell communication, and host immune responses. They are the major source of metabolic energy supply and the 'fine-tuning' of cell-cell interactions [5][6][7][8][9][10][11][12]. The modifications in the sugar moiety, including the changes in the sugar substituents, replacement of oxygen, the addition of heteroatoms, ring size variations, and the replacement with acyclic moiety [13][14][15][16][17][18][19], can produce remarkable changes in biological activity and the degree of toxicity [20][21][22][23][24]. It is known that carbohydrates with hydrophobic tails such as ester, amine, amide, or glycosidic connection may interact with the lipid bilayer of the bacterial cell membrane [25][26][27]. These compounds can be considered as membrane-targeting antibiotics. On the other hand, these compounds can bind to enzymes that play an important role in the biosynthesis of key carbohydrates in the microbial cell wall. They are also nonionic, non-toxic, taste-free, odorless, and biodegradable [28,29].
The current study focuses on the potential usage of a series of compounds, 1-9, which act as antifungal agents (Figure 1) [30]. Molecular docking was also performed to assess their activity against the lanosterol 14α-demethylase (CYP51A1) and Alternaria alternata (6LCC), along with the prediction of activity spectra for substances (PASS). The series was also analyzed to identify their pharmacophore sites.

Antifungal susceptibility testing
For determining 'mycelial growth' of the synthesised methyl 4,6-O-benzylidene-α-D-glucopyranoside derivatives against three fungi indicated in Table S1, the 'poisoned food' approach was selected. In each sanitized Petri plate, 20 mL of sterilized melted Potato Dextrose Agar medium (PDA at 45°C) was poured (90 mm). After the medium had solidified, the fungal inoculums (5 mm mycelial block) were placed in the centre of the Petri dishes.
Positive control was maintained with Nystatin, and negative control was also maintained without using any chemicals. Minimum inhibitory concentration (MIC) was also assessed by testing various concentrations of the derivatives against fungal cultures.

PASS prediction
The online web application PASS (http://www.pharmaexpert. ru/passonline/) was employed to calculate the antifungal activity spectrum of the carbohydrates derivatives [31]. Firstly, the thymidine analog structures were drawn and then changed to their smiles formats by using the renowned SwissADME free online application (http://www.swissadme.ch) to determine the antifungal spectrum using the PASS web tool. This server can surmise over 4000 types of antimicrobial functions, together with drug and non-drug activity, which helps to suggest the best potential outcomes with 90% validity. PASS outcomes are revealed by Pa (probability for active molecule) and Pi (probability for inactive molecule). Showing great potential, both, the Pa and Pi scores vary in the range of 0.00 to 1.00, and usually, Pa + Pi ≠ 1, as these potentialities are freely predicted. The biological actions with Pa > Pi are indicative of 'probable' titles for a selected drug molecule.

Protein selection and molecular docking
The crystal 3D structure of lanosterol 14α-demethylase (CYP51A1) and Alternaria alternata (pdb: 3LD6 and 6LCC) were recuperated in the pdb format from the protein data bank [32]. PyMol (version 1.3) software packages were employed to remove all heteroatoms and water molecules [33]. Energy minimisation of the protein was performed by using a Swiss-PdbViewer (version 4.1.0) [34]. Furthermore, a molecular docking study against the lanosterol 14α-demethylase (CYP51A1) and A. alternata (pdb: 3LD6, and 6LCC) was conducted on the optimised compounds. Finally, the PyRx application (version 0.8) was used to carry out molecular docking simulation [33], envisaging the target protein as a macromolecule and the carbohydrates derivatives as a ligand. The protein and ligands were inputted by converting the pdb format to pdbqt, and the AutoDock Tools of the MGL software package were employed. In the AutoDockVina, the size of the grid box was maintained as follows, (62.2091, 53.8631, and 64.3829 Å for 3LD6) and (61.0383, 51.1083, and 57.1417 Å for 6LCC) along the X, Y, and Z axes. After docking, both the structures of macromolecule and ligand were saved in pdbqt format, and Accelrys Discovery Studio (version 4.1) was employed to explore the results of docking and to predict the non-bonding interactions among carbohydrates derivatives and amino acid chains of the proteins [35].

Docking validation study
The reliability of the docking study was clarified by validation protocol. Validation was performed by extracting the co-crystallized structure of ligand (α-D-glucopyranose) of the lanosterol 14α-demethylase (3LD6) and re-docking it into the unchanged position. The pose of the re-docked ligand at the lowest energy level was superimposed on the co-crystallized ligand by PyMOL 2.3, and the score of root mean square deviation (RMSD) was determined for these two superimposed ligand structures. In order to verify the docking simulation, The RMSD score must be within the reliable range of 2 Å [36,37]. It was performed to improve ligand enrichment, which is essential to justify the docking methodology. PDBsum server (http:// www.ebi.ac.uk/thornton-srv/databases/cgi-bin/pdbsum/ GetPage.pl?pdbcode=3ld6&template=procheck_summary.htm l) was also employed to identify the validation of lanosterol 14αdemethylase (3LD6) with a Ramachandran plot ( Figure 2) which revealed 92.60% residues in the allowed region.

Molecular dynamic simulations
Molecular dynamics simulations (MDS) were conducted in YASARA dynamics [38] with the aid of the AMBER14 force field [39]. The docked complexes were initially cleaned and optimised, and the hydrogen bonds were oriented. The TIP3P water solvation model was used with a periodic boundary condition [40]. The physiological conditions of the simulations were set at a temperature of 298 K, pH 7.4, and 0.9% NaCl. The initial energy minimizations were conducted by steepest gradient algorithms using a simulated annealing method (5000 cycles) [41]. The long-range electrostatic interactions were calculated by the Particle Mesh Ewalds by a cut off radius of 8.0Å [42,43]. The time step of the simulations was set at 2.0 fs. The simulation trajectories were saved after every 100 ps. By following the Berendsen thermostat and constant pressure, the simulations were extended for 100 ns [41]. The simulation trajectories were utilised to analyze the root mean square deviations (RMSD), solvent accessible surface area, hydrogen bond, and radius of gyration (Rg) [44][45][46][47][48].

Pharmacokinetic and drug-likeness prediction
The prediction of ADMET properties in drug development is important to prevent drug failure during the clinical stages.
For this reason, the best-identified esters were evaluated using pkCSM [49]. The absorption in the human intestine, the blood-brain barrier, and the central nervous system, along with its metabolism, indicates the chemical biotransformation of a drug by the body, total clearance of drugs, and the toxicity levels of the molecules were all predicted. The multiple linear regression (MLR) equations are then utilised to obtain the QSAR and pIC50 values [50].

POM theory
POM (Petra/Osiris/Molinspiration) Theory was employed to identify and to optimise most of the antifungal pharmacophore sites one by one on the basis of their different physico-chemical parameters and their different electronic charge repartition of the corresponding heteroatoms. POM Theory was extended, with success, to other various and different biotargets [51][52][53][54][55][56][57][58][59][60].

The carbohydrates derivatives
In continuation of carbohydrate research in the Laboratory of Carbohydrate and Nucleoside Chemistry (LCNC), we intended to test a series of methyl 4,6-O-benzylidene-α-D-glucopyranoside (1) derivatives [30] for PASS, POM, molecular docking, molecular dynamics simulations pharmacokinetic profile, drug-likeness, and pIC50 analysis.

Assessment of antimicrobial activities by PASS and bioactivity
The antimicrobial spectrum was predicted by applying the web server PASS to all the carbohydrate derivatives 1-9. The PASS results are expressed as Pa and Pi, which may be found in Table 1. The carbohydrate derivatives 1-9 showed 0.41 < Pa < 0.60 for antibacterial, and 0.42 < Pa < 0.62 for antifungal. These results revealed that these molecules could be more efficient against bacteria and fungi. The attachment of additional aliphatic acyl chains (C2 to C14) increased the antifungal activity (Pa = 0.624) of (1, Pa = 0.241), whereas the insertion of chloro and -C(CH 3 ) 3 substituted aromatic groups markedly improved the activity. The same scenario was observed for the antibacterial activity, where the aromatic analogs revealed an improved activity compared to the acyl chain analogs. However,

Docking validation study
The confirmation of the protein-bound ligand and re-docked co-crystallized ligand was employed to validate the accuracy of the docking performance. Figure 5 displays the superimposed view between the docked ligand and the co-crystallized ligand, whereby the RMSD score is <2Ǻ. The docked complex was observed thereafter to interact with the identical active site in comparison to the ones mentioned in the current study. The small molecules can gain a lower RMSD score easily, even if placed randomly, whereas the bulky symmetric molecules can be exchanged in the binding site during docking. Literature review [61][62][63] has suggested a novel way to know the quality of docking poses based on visual analyses. Figure 5 shows the 2D visualisation of the interactions between a predicted docking pose and the tested ligand conformation. The outcomes of the visual analyses exhibited the same interactions as in the experimental binding mode, as found in Figures 6 and 7. This study revealed that predicted data alone is not a reliable parameter for validation of the docking process together with the use of visual analyses is a new way that must be considered as essential.

Molecular docking analysis
The binding energy and interaction modes of the carbohydrate derivatives with the active site of lanosterol 14αdemethylase (3LD6) and A. alternata (6LCC) were studied in silico using the AutoDock Vina software (Tables 2 and  S4). The results that were obtained have showed that the aliphatic acyl derivatives displayed better binding scores than the aromatic derivatives. The docking conformation of the most active molecules (2, 4 and 6-9) was displayed in Figure 6. The interactions between the inhibitor and bordering residues of both proteins were illustrated in 2D schematics. These were attained by importing the docking results into the Discovery Studio Visualizer. The pattern of interactions between the ligand and enzyme, including the involved amino acids and the contribution of the total energy of the interaction, was demonstrated in Figure 7.
Most of these interactions included hydrophobic contacts, Van der Waals interactions, hydrogen bonds, electrostatic, carbonyl, and one specific atom-aromatic ring, plus they   also provide an insight into understanding molecular recognition.
Along with Phe234, all the derivatives displayed maximum π-π interactions with the Phe77, Phe139, Phe146, Phe152 and Trp239, denoting tight binding with the active site. Reports suggest that Phe234 is the PPS's principal component, and PPT is responsible for the accessibility of small molecules at the active site [64]. Binding energy and binding modes were improved in some of the derivatives (1, 2 and 5) because of the significant hydrogen bonding. It was observed that the alterations of the -OH group in thymidine increased the π-π interactions with the amino acid chain at the binding site, whereas their improved polarity resulted in the formation of hydrogen bond interactions. The maximum numbers of H-bonds were observed for derivative (5) with Arg264, Asn307 and Asp176 residues. H-bonds execute a vital function in shaping the specificity of ligand binding with the receptor. The H-bond surface and hydrophobic surfaces of derivative (5) with both targets were represented in Figure S1.
The distance of the ligands, along with the change in the accessible area of the two important catalytic residues within the active site of the target, is shown in Table S4. The calculated binding affinities suggest that the synthesised molecules can spontaneously interact within the binding site of the target.

Molecular dynamics simulations
A molecular dynamics simulation (MDS) study was conducted to analyze the structural stability of the docked complexes. The root mean square deviations (RMSD) of the C-alpha atoms from docked complexes were evaluated where the 3LDC complexes showed an initial upper trend ( Figure 8A-D). The complexes-6 showed higher RMSD at the initial stage of the complexes, which indicates higher flexibility of the complexes across the amino acid residues.
The complexes-6 and 7 were both stabilised after 20 ns, which was followed by a lower RMSD until the rest of the simulation time. Both complexes exhibited RMSD less than 2.5Å, indicating the stability of the complexes. The complexes from 6LCC ( Figure 9A-D) showed a similar trend with an RMSD under 2.5Å. Therefore, the solvent accessible surface area of the complexes demonstrated the changes in the surface area of the protein. The complexes from the 3LDC protein indicated that the complex-7 exhibited a higher SASA compared with complex-6, indicating that the complex-7 increased the total surface area upon ligand binding. The complexes from 6LCC also demonstrated a stable nature and lower fluctuations in SASA trajectories. Moreover, 3DLC complexes exhibited lower fluctuations in the Rg trajectories, which indicates the stability of the complexes. The Rg from 6LCC complexes exhibited lower deviations and a more stable nature. The radius of gyrations of the complexes determined the mobile nature of the complexes, where the higher Rg correlated with, the more labile nature. In comparison, the lower Rg correlated with those having greater stability of the complexes.
Therefore, the hydrogen bond of the complexes indicated the stable nature of the complexes, whereas the complexes with 3DLC and 6LCC showed lower fluctuations and fewer changes, which is related to the more stable nature of the complexes.

Pharmacokinetic profile and drug-likeness analysis
In order to predict the pharmacokinetic properties, including absorption, distribution, metabolism, excretion (ADME), and toxicity of the compounds, pkCSM ADMET descriptors algorithm protocol was used. Drug absorption depends on factors such as membrane permeability, intestinal absorption, skin permeability thresholds, and substrate or inhibitor of Pglycoprotein. A value of intestinal absorbance below 30% suggests poor absorbance. Our results indicated that all analogs showed excellent absorption with values exceeding 30% (Table 3). Skin permeability is an important consideration for improving drug efficacy, and it is particularly of interest in the development of transdermal drug delivery. A molecule will barely penetrate the skin if log Kp is more than −2.5 cm/h. All analogs exhibited good skin penetrability and the Kp of the thymidine esters was −2.517 to −2.967 cm/h (< −2.5) ( Table 3). For the pkCSM predictive model, high Caco-2 permeability is translated into predicted log Papp values >0.90 cm/s. As shown in Table 3, the value of the Caco-2 permeability (log Papp) of the thymidine esters ranged from 0.318 to 0. 570 cm/s, log Papp < 0.9 cm/s, indicating that they all have low Caco-2 permeability. For the discovery of orally administrative drugs, solubility is one of the major descriptors. High water solubility is useful for delivering active ingredients in sufficient quantities in small volumes of such pharmaceutical dosage. The values for water solubility are given in log (mol/l) (Insoluble ≤ −10 < poorly soluble < −6 < Moderately < −4 < soluble < −2 < very soluble < 0 ≤ highly soluble). As indicated in Table 3, it can be observed that, the tested esters are soluble.
Distribution volume (Vd) is a pharmacokinetic parameter that reflects the tendency of an individual substance to either stay longer in plasma or redistribute to another tissue compartment. According to Pires et al. [49], VDss is considered low if it is below 0.71 L/kg (log VDss < −0.15), and high if it is above 2.81 L/kg (log VDss > 0.45). Table S5 showed that the value of carbohydrate derivatives whereby VDss ranged from −0.410 to 0.274, with only four derivatives (1-3 and 7) having a VDss value of < −0.15.
Blood-brain partitioning and brain distribution are critical properties for drugs targeting the central nervous system. Compounds exhibiting a logBB < −1 are considered poorly distributed to the brain. The logPS (the central nervous system (CNS) permeability) value of the carbohydrate derivatives ranges from −3.453 to −2.303, logPS < −3 (Table  S5), indicating that the carbohydrate derivatives (1-9) are unable to penetrate the CNS. The model provided by pkCSM pharmacokinetics predicts the total clearance log (CLtot) of a given compound in log(mL/min/kg). The larger the CLtot value of the compound, the faster the excretion process. Table S5 showed the log CLtot value of the carbohydrate derivatives ranging from 0.058 to 1.557 mL/ min/kg.
The toxicity of the carbohydrate derivatives, as described in Table S6, shows high LD50 values (2.11-2.45), suggesting that the esters are lethal only at very high concentrations. The negative result in the AMES test suggested that the ester is not   TAACF, was duly employed to validate the activity of the synthesised compounds. Hence, the series of compounds 2-9 were assessed to identify their pharmacophore sites according to the POM organigram ( Figure 10). The identification of the type of pharmacophore sites of these compounds was derived from the physical and chemical properties of the tested compounds by using the bioinformatics POM platform (Figure 10). Results   of pharmacokinetic properties and bioactivity scores analyses are shown in Figure 11 and Table 5. The Molinspiration calculations of compounds 1-9 show that most of the compounds are suffering from weak hydrosolubility because of the cLogP of compounds 3-9 depasse 5 value (Tables 6 and 7). So, the reduction of the length of alkyl substituents along with the number of the aryl groups will be beneficial in favour of an improved bioavailability. This series presents 7/9 compounds that present two violations (NV = 2) because the MW and cLogP depasse the limits of Lipinski 5 rules. This has a direct impact on the bioactivity scores ( Table 7). The calculations of atomic charges of compounds 1-9 show that most compounds bearing two esters groups in germinal (1, 2) positions can able to inhibit fungus (Fig. 7).
The identification of the pharmacophore sites of the best Hit (Compound 8) as the most active candidate as an antifungal agent becomes evident in the presence of combined (O1, O2) pharmacophore sites ( Figure 12).

Conclusion and perspectives
This study revealed that carbohydrate derivatives 5, 6 and 8 showed more potential activity against fungal organisms with better pharmacokinetic and biological spectra. These observations were validated by a molecular docking study, which revealed a promising antimicrobial efficacy for the synthesised carbohydrate derivatives. Compounds (1-9) showed promising binding interactions and binding energy with lanosterol 14α-demethylase and A. alternata. This finding is strongly and unequivocally supported by the MDS over 100 ns, confirming the binding stability of the docked complexes in the trajectory analysis. This implies that the protein-ligand complexes are highly stable in the biological system. Additionally, the pharmacokinetic properties of these compounds indicated that most of the designed molecules showed improved kinetic parameters, maintained all drug-likeness rules, and improved biological activity. The QSAR data revealed that with increasing the molecular weight and chain length, the pIC50 increased but stayed below the standard.

Author contributions
SMAK conceptualisation, methodology, article writing, and supervision; FAA, TBH and HL perform the POM study; MAH and SM performed the computational and interpretation of the results; MARK, AA, NMPM, FH and SSMS resources, validation of article, edited and improve the manuscript. All authors reviewed and approved the manuscript.