Microbial pigments as potential anti-rift valley fever virus drugs

Abstract

Rift valley fever virus (RVFV) is among the WHO list of priority diseases, yet no effective vaccine or treatment is currently available. Microbial pigments (MPs) represent a promising small-molecules library which can be exploited for the drug discovery of anti-RVFV compound. In this study, thirteen MPs were in silico screened to identify candidates with acceptable drug-likeness and possible ability to cross the blood brain barrier. Next, the binding interaction of the filtered molecules were compared against key RVFV proteins for the selection of the optimum inhibitor. Molecular dynamics simulations were performed (200 ns) to further evaluate the interactions. The selected candidate (pyocyanin; PCN) was produced, purified and analytically characterized in-house. Finally, the antiviral potential of PCN was tested in vitro against RVFV using the tissue culture infection dose 50% (TCID₅₀) method. In silico screening studies revealed that prodigiosin and PCN exhibit ideal drug-likeness properties. PCN exerted a promising in silico interaction with the key RVFV proteins as revealed by the molecular docking and dynamic studies. Results showed that PCN may be effectively produced and purified from bacterial cultures. Its cell-safe concentration (0.49 μg/mL) demonstrated a promising 2.89 log₁₀ reduction in TCID₅₀ when incubated with RVFV infected cells. A lower effect was observed (2.00 log₁₀ reduction) in cells treated with PCN prior to RVFV infection. These results suggest that PCN may represent a potential effective low-cost molecule to combat RVFV.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13065-025-01680-2.

Introduction

Rift valley fever virus (RVFV) is currently on the WHO list of potential next pandemics. It is recognized as a viral threat that is neglected by healthcare systems [1]. It infects live stocks and causes increased neonatal mortality and abortion (up to 100%) [2]. RVFV can be transmitted to humans, commonly via mosquito bites, contact with blood and/or organs of infected animals or the consumption of unpasteurized milk [3]. Recent studies highlighted that RVFV spread have broken the geographical barriers. The prevalence of RVFV, in human is rising to 5.24% [4]. The disease implications range from mild fever to fatal hemorrhagic disease [5]. Alarmingly, RVFV outbreaks were already recorded with a high (13%) mortality rate among humans and high probability to cause pandemics [4, 6]. To date there is no specific cure nor an approved vaccine against RVF infection despite its multiple outbreaks [7]. Against this background, the WHO placed RVFV on its list of prioritizing diseases for research and development in emergency contexts and on its R&D blueprint [8].

The availability of active pharmaceutical ingredients (API) is a rate limiting step for securing access to medications. Their chemical production is time consuming and expensive, posing a challenge even to countries with developed infrastructures [9, 10]. Discovery of a new molecule may require up to 10–15 years of research and has a significant carbon footprint. Despite this significant investments, the process has a limited success rates for small molecules (7%) and biological products (11%) [11].

The drug development process becomes more complicated for the development of antiviral agents due to the rapid development of resistance and the fast-viral mutations. This renders the development of a new antiviral agent less financially appealing to pharmaceutical companies which may contribute to the lack of effective medications for many viruses [12]. It is important to note that to date, only a limited number of known viruses has clinically approved medications [9, 13–15]. In the same context, accelerated development of antiviral agents is frequently required in situations of viral emergencies and outbreaks [16].

A plethora of identified microbial pigments (MPs) exists in the drug likeness space which supports their further investigation as potential drug candidates [17–24]. In this context, investigating evidence of their promising antibacterial [25–27], anticancer [28, 29] and antimalarial [30, 31] and antifouling [31, 32] activities, is emerging [18].

To the best of our knowledge, limited research exists on the investigation of MPs as antiviral agents. Therefore, this study aimed to investigate multiple MPs as potential antiviral agents against RVFV. In silico comparison of the antiviral potential of best MP candidates, in terms of drug likeness, was conducted on key RVFV proteins for identification of those with best binding affinity. Furthermore, the stability of the MP-RVFV protein interaction was assessed. Finally, the antiviral activity of the selected MP was conducted.

Material and methods

Chloroform was purchased from Fisher Scientific (Leicestershire; UK). The RVF virus model and the Vero cell (CCL-81) were supplied by the Egyptian company for Biological products and vaccines VACSERA (Giza; Egypt). Docking was performed using the AutoDock Vina software 1.2.5. The LC/MS–MS analysis was performed on a Shimatzu Nexera HPLC–MS/MS system (Tokyo; Japan). A Jasco UV–Vis spectrophotometer (Tokyo, Japan) was employed in this study.

Computational studies

ADME pharmacokinetic screening

The canonical SMILES of thirteen MPs (canthaxanthin, astaxanthin, zeaxanthin, β-carotene, staphyloxanthin, deinoxanthin, prodigiosin, undecylprodigiosin, heptyl prodigiosin, PCN, violacein, indigoidine and bikaverin) were introduced into the Swiss-ADME web server (http://www.swissadme.ch/index.php) to calculate the pharmacokinetic properties and estimate the drug-likeness of the 13 tested candidates as per variable filters (Lipinski , Ghose, Veber, Egan, Muegge rules) and to select the potential drug-like molecules. The criteria were set to select candidates with no violations of the selected filter rules. From the compounds with zero violation, only those with the ability to cross blood brain barrier was further investigated.

Molecular docking

Target preparation

The viral genome contains three negative segments large (L), medium (M) and small (S). The L segment encodes the viral RNA dependent RNA polymerase which is identified as the L protein (Table 1). This protein is essential for the viral replication within the host cell via binding and cleavage of the 5’ N7-methylated guanosine cap of host mRNAs as an initial step for viral transcription which render the protein as an active target for the small molecule inhibitoris [61]. Other RVFV proteins are the glycoprotein N (Gn), the glycoprotein C (Gc) and two nonstructural (NS) proteins, namely the NSm1 and NSm2. These proteins are encoded by the M segment. The Gn and Gc are positioned in the viral lipid coat to facilitate its fusion and entry into the host cells. Finally, the S segment of the RVFV genome encodes the viral virulence factor protein (nonstructural protein; NSs) and the nucleocapsid (N) protein which encapsulates the viral genome. Specifically, Gn, Gc, N and L proteins are attractive targets for the development of RVFV protein due to its role in the viral transcription [62].

Table 1.

Binding site coordinates

Target Protein	PDBID	X	Y	Z	Exhaustiveness	Reference
Crystal Structure of DC-SIGN in complex with a triazole-based glycomimetic ligand (DC-SIGN)	7Nl7	16.400260	− 9.320900	26.466780	8	[90]
Crystal structure of Gn from Rift Valley fever virus (Gn)	6F8P	7.522000	85.667000	16.258000	32	[91]
Crystal structure of glycoprotein C from Rift Valley Fever Virus (Gc)	4HJ1	134.791000	14.219000	111.455000	32	[92]
Structure of the cap-binding domain of Rift Valley Fever virus L protein (L-protein)	6QHG	− 0.456355	6.257740	100.042439	32	[70]
Crystal Structure of the Rift Valley Fever Virus Nucleocapsid Protein (N)	3LYF	− 3.969686	18.438412	2.109294	32	[93]

Active site selection

The active site of the DC-SIGN (7NL7) was selected to be that of the co-crystallized ligand (3-Aminopropyl 2-deoxy-2-(4-phenyl-1,2,3-triazol-1-yl)-alpha-d-mannopyranoside). The dimensions of the site were X = 16.40, Y = − 9.32, and Z = 26.47. For the RVF Cap-binding domain of L protein (6QHG), the ligand (7N-Methyl-8-Hydroguanosine-5-Triphosphate) binding site, as presented in the X-ray crystallographic structure, was selected as the docking site with the coordinates X = − 0.456355, Y = 6.257740, Z = 100.042439.

No co-crystallized ligand was identified within the PDB database for proteins Gn, Gc and N protein. Accordingly, the largest pocket was deduced by the Biovia studio software for these proteins. The coordinates of docking were 6F8P (X = 7.522000, Y = 85.667000, Z = 16.258000), 4HJ1 (X = 134.791000, Y = 14.219000, Z = 111.455000) and 3LYF (X = -3.969686, Y = 18.438412, Z = 2.109294) as determined from PDB site records. The docking radius was kept at 20 for all proteins.

Docking procedures and pose selection for molecular dynamics

Docking was conducted using the AutoDock Vina software version 1.2.5 [33]. The docking experiment was performed on the selected MPs which have shown optimum drug-likeness according to the ADME screening (PCN, prodigiosin). The receptors were kept rigid, and the pigments were flexible to select the most suitable binding pose. The docking exhaustiveness for each protein is presented in Table 1 and the number of modes was kept to default (9 modes).

Molecular dynamics

Molecular dynamics (MD) was performed for further verification and the assessment of the binding stability of PCN to selected protein (Rift Valley Fever virus L protein). The docked pose of the selected molecule (PCN) with each target protein was utilized for unbiased, classical MD simulations of 200 ns duration in GROMACS 2021 to assess the stability and interaction strength of the complexes [34]. The CHARMM-GUI web server's solution builder module facilitating input file preparation was utilized [35–38]. Each PDB file was uploaded, solvated with TIP3P water in a cubic box with 1 nm padding, and neutralized with Na⁺ and Cl⁻ ions to 0.154 M concentration mimicking physiological conditions. CHARMM36m force field parameters were employed for the proteins, TIP3P water, and neutralizing ions. Moreover, the CHARMM general force field CGenFF within CHARMM-GUI was used to parameterize the small molecule.

The procedure was initiated with potential energy minimization to prevent atomic collisions with a maximum force on any atom set at 100 kJ.mol⁻¹.nm⁻¹ as a convergence criterion. Next, the equilibration was performed via an NVT ensemble for temperature adjustment followed by an NPT ensemble for pressure adjustment. Both steps employed V-rescale thermostat and Berendsen barostat, with target values of 310 K and 1 atm, respectively, and lasted for 2 ns each [39]. Each equilibration stage utilized position restraint on the backbone, heavy atoms of the ligand, and side chain heavy atoms with force constants of 400 kJ.mol⁻¹.nm⁻¹, 400 kJ.mol⁻¹.nm⁻¹, 40 kJ.mol⁻¹.nm⁻¹, respectively. H-bond lengths were constrained by the Linear Constraint Solver (LINCS) [40], and electrostatics were treated with Particle Mesh Ewald (PME) with a threshold 1.2 nm [41]. Integration of New­tonian equations of motion used the leap-frog algorithm with a 1 fs time-step for the two equilibration steps while 2 fs for the production step. During the production run, the Nose–Hoover thermostat and Parrinello-Rahman barostat were utilized to keep the temperature at 310 K and pressure at 1 atm. Frames were collected every 0.1 ns (2000 frames total). Pre-analysis involved removing periodic boundary conditions (PBC) and making the protein whole again using the GROMACS gmx trjconv command. After that, VMD TK scripts were used for analysis [42]. Separate root mean square deviation (RMSD) values were calculated for each protein and molecule, as were the distance from the centers of mass of ligand to protein, root mean square fluctuation (RMSF), solvent accessible surface area (SASA), radius of gyration (RoG), and H-bond counts between the ligand and the corresponding protein. All frames were used for these analyses. To manage data complexity, trajectory clustering was performed using TTClust and obtained a representative frame for each cluster [43]. Protein–ligand interactions were identified and counted in each frame using the protein–ligand interaction profiler (PLIP) [44].

Binding free energy calculation using MM-GBSA

Binding free energy was computationally estimated using the MM-GBSA approach implemented in gmx_MMPBSA tool and key residue contributions were determined using a decomposition radius focused on amino acids within 1 nm of the ligand to accommodate all amino acids related to the interactions [45, 46]. Both ionic strength and solvation (igb) parameters were adjusted to 0.154 M and 5, respectively [47, 48]. The internal and external dielectric constants were adjusted to 1 and 78.5, respectively. The following equations show the method used to calculate the binding energy between the protein and the complex:

1

where < > depicts the average of the enclosed free energies of the ligand protein complex, the receptor, and the ligand over the frames used in the calculation. In this approach, the stable part of the trajectory (the last 400 frames) was selected. Different energy terms were calculated using Eqs 2, 3, 4, 5, 6 as follows:

2

3

4

5

6

where:

The ΔH is the enthalpy as estimated from the sum of the gas-phase energy (E_gas) and the solvation-free energy (E_sol). The –TΔS is the entropy contribution to the free binding energy. The E_gas is the sum of electrostatic (E_ele) and van der Waals (E_vdW) terms. The E_solv is the total of adding the polar sol­vation energy (E_GB) and nonpolar solvation energy (E_SA) . E_SA is estimated from the solvent-accessible surface area. γ.SASA is the solvent accessibility surface area.

Protein Ligand Interaction Fingerprint (ProLIF) analysis

To elucidate the nature and persistence of intermolecular interactions at the residue level, the Python-based Protein–Ligand Interaction Fingerprints (ProLIF) tool was employed [49]. This approach enabled the systematic quantification and classification of non-covalent interaction types formed between PCN and specific amino acid residues throughout the simulation.

Production and extraction of PCN

The extraction and purification of PCN from Pseudomonas aeruginosa cultures was performed according to previously reported protocol [50, 51]. Briefly, PCN producing P. aeruginosa suspension (provided by the Microbiology Department, Faculty of Pharmacy, Ahram Canadian University, Egypt) was incubated in nutrient LB broth for 48 h at 37 °C. Post incubation, PCN was extracted from the liquid LB broth (100 mL) by liquid–liquid extraction. The organic phase (chloroform, 100 mL) was mixed with an equal volume of broth in a separatory funnel. The two phases were vigorously shaken for three minutes to extract PCN, a pigment with a characteristic blue color, into the organic phase. The broth was discarded, and the chloroform layer was separated by the separating funnel for further purification. A second cycle of liquid–liquid extraction was performed where the blue chloroform layer (100 mL) was mixed with an equal volume of 0.1 N HCl. The now ionized PCN was transferred into the aqueous medium, turning the aqueous phase to red. For further purification, the aqueous layer was neutralized by 0.1 N NaOH, added dropwise, to return PCN to its unionized state. At this point the blue un-ionized PCN was re-extracted into new chloroform.

Analytical and purity testing

Analytical and purity testing

Identification and purity testing were performed by UV–Vis and mass spectroscopy (MS). The purified PCN was dissolved in chloroform and the UV–VIS spectrum recorded in the range 400–800 nm to investigate the characteristic λ_max of PCN. The mass of the purified PCN was detected by MS scan in the direct scanning mode in range from m/z 50–1000. The fragmentation of the extract was performed by MS/MS for confirming the structure of PCN. Different collision energies and conditions were applied to verify the identity of the fragments.

In vitro studies

Cell viability assay

The viability assay was performed on Vero cells (CCL-81) that were cultured in Eagle’s minimum essential medium (Hyclone, Logan City, Utah, USA) containing 10% fetal bovine serum (10%), 100 μg/mL streptomycin, 100 μg/mL penicillin and 2 mM L-glutamine. Vero cells were incubated in a humidified atmosphere and 5% CO₂. Upon confluency, the growth medium was discarded from the 96-well plates. Next, a two-fold serial dilution of PCN (500 μg/mL) was performed in Eagle’s minimum essential medium to achieve a concentration range of 500–0.49 μg/mL and 100 μL of each concentration were added in the 96-well plates. The cell culture plates were incubated at 37 °C for 24 h followed by media discarding and washing with phosphate buffer saline, pH 7.4 (250 μL/well). MTT (0.5 mg/mL; 50 μL) was added to each well and the plates were re-incubated for another 4 h. The plates were washed with PBS and the formed formazan blue crystals was dissolved in DMSO (50 μL) and the absorbance was measured at 570 nm. The cell viability (%) was calculated relative to the control (cells treated with PCN free media). The equation was absorbance of treated cells/absorbance of control cells*100 [52]. The cell-safe concentration was determined and used for the assessment of the antiviral effect [52].

Antiviral assay

The chloroform extracted PCN was dried, weighed and used to estimate its antiviral potential. The antiviral potential of extracted PCN against RVFV was evaluated by determining the reduction of tissue culture infectious dose 50% (TCID₅₀) which is the viral dilution that can induce cytopathogenic effect (CPE) to 50% of Vero cells [53, 54]. Two procedures were adopted for estimating the antiviral potential of PCN; the direct and indirect antiviral assays using the cell-safe concentration of PCN [55].

Indirect viral assay

To determine the ability of PCN to prevent viral infection, the indirect antiviral activity assay was performed. In this assay, the cells were inoculated with the safe concentration of PCN (0.49 μg/mL) for 24 h at 37 °C. Next the media were decanted, and Vero cells were washed then the tenfold serial dilutions of the virus (10^–1–10^–8) were added and incubated with the cells with daily examination for detection of CPE development and determination of the viral titer.

Direct antiviral assay

In the direct method the Vero-cells were inoculated with variable tenfold serial dilutions (10^–1–10^–8) of RVFV for 1 h. Next, the media were discarded, and the plates were washed to remove the un-adsorbed viruses. Next the plates were inoculated with the cell safe concentration of PCN (0.49 μg/mL) for 24 h and examined microscopically for CPE. In both cases the viral titer was recorded and compared to interferon (IFN-α)-treated and untreated plates as positive and negative controls, respectively.

Results and discussion

This manuscript describes the first report to explore MPs for the discovery of RVFV inhibitor. In this study, thirteen MP were screened by Swiss-ADME to assess their drug-likeness. The molecules which passed the medicinal chemistry filters (Lipinski, Ghose, Veber, Egan, and Muegge) with zero violations were further investigated by computational study (molecular docking) to predict their potential activity on RVFV. Finally, the in vitro antiviral activity was assessed for PCN, the most promising compound according to computational screening studies (Fig. 1).

Fig. 1.

The workflow for the assessment of anti-RVFV microbial pigment (MP) molecules. a) Five medicinal chemistry filters were applied for selection of MP molecules with zero violations. This step resulted in identification of 4 candidates out of 13 screened candidates. b) The ability to cross BBB was assessed and 2/4 candidates exhibited ability to cross the BBB. c) Molecular docking was performed to select the candidate with best binding interaction to viral target proteins. d) The stability of binding of the most promising candidate (PCN) was assessed by molecular dynamics. e) In vitro production of PCN. f) In vitro assessment of PCN anti-RVFV activity

In silico pharmacokinetic assessment of MP

A panel of MPs (n = 13) were screened for the selection of potential candidate in terms of good drug likeness (Fig. 2). The screened candidates were canthaxanthin, astaxanthin, zeaxanthin, β-carotene, staphyloxanthin, deinoxanthin, prodigiosin, undecyl-prodigiosin, heptyl prodigiosin, PCN, violacein, indigoidine and bikaverin. The average molecular weights were 455.54 ± 164.79.

Fig. 2.

a) Chemical structures of the investigated MP. The 4 selected candidates, which have acceptable drug-likeness criteria according to different medicinal chemistry filters, highlighted. b) The boiled egg illustration of membrane absorption of the 4 compounds. The four compounds exist in the white zone indicating high intestinal absorption potential. PCN and prodigiosin exist in the yellow zone which indicate they have the ability to cross the blood brain barrier which is crucial for RVFV replication. All compounds are not expected to be substrates of the Pgp as indicated by the red dots

Five medicinal chemistry filters were applied in this study: namely Lipinski, Ghose, Veber, Egan and Muegge. The Lipiniski states that for compound to be drug like it should maintain a limit of 5 hydrogen bond donors, 10 hydrogen bond acceptors and molecular mass less than 500 Da, and a partition coefficient not greater than 5 [56]. The Ghose filter on the specify that the successful candidate should have a molecular weight in range 160–480 Da, the logP should range from − 0.4 to 5.6, the number of atoms should be between 20 and 70, while the molar refractivity should range from 40 to 130 [57]. The Veber filter ensures that the compound should adhere to total polar surface area ≤ 140Ų, and the number of rotatable bonds should be ≤ 10 to achieve acceptable bioavailability. Similarly, the Egan rule states that the logP should be ≤ 5.88 while total polar surface area should be maintained ≤ 131.6 Ų⁵⁷. The Muegge rule states that for a drug like chemical compound should have molecular weight between 200 and 600 Da, XLogP in range 2–5, a total polar surface area below 150, the limit of rings should be less than 7, number of heteroatoms should be more than 1, number of carbons should be more than 4, number of rotatable bonds less than or equal to 15, the number of hydrogen bond acceptors and doners should be maintained below 10 and 5, respectively [58].

According to ADME screening results, four candidates (out of 13 screeded compounds) passed the Lipinski, Ghose, Veber, Egan, and Muegge drug likeness filters with zero violations. These candidates were prodigiosin (Mwt: 323.43 Da), PCN (Mwt: 210.23), violacein (Mwt: 343.34) and bikaverin (Mwt: 382.32). In silico pharmacokinetic assessment revealed that all the four candidates may be orally absorbed. However, only two candidates, prodigiosin and PCN, exhibited higher probability of crossing the blood brain barrier as inferred from in silico studies. This is an interesting and important predictor for successful anti-RVFV drug because the virus propagates mainly in the brain causing brain tissue damage [59, 60] (Table 2 and supplementary file 1).

Table 2.

In silico screening of the ADME-Tox properties

Molecule	GI absorption	BBB permeant	Lipinskia	Ghosea	Vebera	Egana	Mueggea	Bioavailability Score
Canthaxanthin	Low	No	2	4	0	1	1	0.17
Astaxanthin	Low	No	2	4	0	1	1	0.17
Zeaxanthin	Low	No	2	4	0	1	1	0.17
β-Carotene	Low	No	2	4	0	1	1	0.17
Staphyloxanthin	Low	No	2	4	1	1	3	0.17
Deinoxanthin	Low	No	2	4	1	1	1	0.17
Prodigiosinbc	High	Yes	0	0	0	0	0	0.55
Undecylprodigiosin	High	No	0	1	1	1	1	0.55
Heptyl Prodigiosin	High	Yes	0	0	0	0	1	0.55
Pyocyaninbc	High	Yes	0	0	0	0	0	0.55
Violaceinb	High	No	0	0	0	0	0	0.55
Indigoidine	Low	No	0	1	1	1	0	0.55
Bikaverinb	High	No	0	0	0	0	0	0.55

^aData represents number of violations

^bCompounds with no violations of medicinal chemistry filters

^cCompounds with BBB permeation ability

Proteins involved in RVF infectivity were selected as potential targets for RVFV control. These include the C-type lectin receptor dendritic cell-specific intercellular adhesion (DC-SIGN), the viral cap-binding L-protein, Gn, Gc, and N proteins. The X-ray crystal structures of the target proteins were downloaded from the protein databank (https://www.rcsb.org) with PDBIDs of 7NL7, and 6QHG, 6F8P, 4HJ1 and 3LYF, respectively. The proteins were prepared via removal of water molecules and cofactors, the addition of polar hydrogens, and fixing potential and charges using the Biovia discovery studio software (Table 1).

Molecular docking and dynamic studies

The viral infection involves multiple stages starting by the viral adsorption and attachment onto the cell surface followed by penetration into the cell and later employing cellular resources for replication. Targeting cellular or viral proteins involved in this cycle may contribute to reduction of viral infectivity [63].

Protein targets involved in viral entry to cells

Dendritic cell-specific intercellular adhesion molecule-3-grabbing non-integrin (DC-SIGN) is reported to mediate viral entry into cells via interaction with the RVFV envelope N-Glycans [64, 65]. Competitive interaction with the DC-SIGN may protect the cells against viral entry [66]. The shallow carbohydrate binding site of the receptor has been a challenging target for drug discovery and development of pharmacological inhibitors. However small-molecule pyrazolone, thiazole, or quinoxalinone and quinolone inhibitors were reported to block the DC-SIGN mediated adhesion [67]. Additionally, ligands that hamper a bicyclic core with a heteroatom and a carbonyl group were suggested as good candidates for DC-SIGN inhibition [66, 67]. These are common features to the phenazine scaffold which have motivated our investigation of the activity of PCN as a phenazine based compound. It is worth pointing out that phenazines are gaining a large research interest as drug-like scaffold. Phenazine based compounds are common in multiple pharmacologically active compounds with anticancer, antimalarial, and antimicrobial activities [68].

Our results show that, PCN and prodigiosin exerted moderate interactions of − 5.5 kcal/mol and − 5.1 kcal/mol at the ligand binding site, respectively. The binding interaction was achieved via interaction of the carbonyl group of PCN and the ASN 349 via H-bonding with a bond distance of 2.28 A°. Similarly, the nitrogen atom of the pyrazine ring exerted a H-bonding with ASN 365 with a bond length of 2.78 A°. Several other binding modes were involved in the interaction via phenyl group Pi-interactions with multiple amino acids in the DC-SIGN binding site including the ASP 367, GLU358 which are shared with the co-crystallized ligand (Fig. 3). The net binding affinity of PCN to the DC-SIGN was weaker than that of the co-crystallized triazole-based glycomimetic ligand (− 6.2 kcal/mol) which suggests that DC-SIGN might not be suitable target for the antiviral potential of either molecule.

Fig. 3.

The binding of PCN to the DC-SIGN. a) 7nl7 prepared protein for docking. The docking site is represented by the red sphere. b) The docking interaction of PCN to 7nl7. c) the binding interaction of the co-crystallized ligand (3-Aminopropyl 2-deoxy-2-(4-phenyl-1,2,3-triazol-1-yl)-alpha-d-mannopyranoside to the protein (DC-SIGN)

Glycoprotein N (Gn), glycoprotein C (Gc) are two RVFV proteins that are positioned in the viral lipid coat and facilitate its fusion/entry into the host cells [69]. In this study, PCN exerted a binding affinity of − 6.4 and − 7 kcal/mol to Gn and Gc, respectively. Prodigiosin binding affinities were − 5.5 and − 7.4 kcal/mol to Gn and Gc, respectively (Fig. 4). Although the Gc binding score seems relatively high, the assessment of antiviral potential (as evidenced from the direct and indirect viral assays discussed below) revealed that the inhibitory potential is mainly dependent on inhibiting replication rather than preventing cell entry.

Fig. 4.

Binding affinity of PCN to the host proteins involved viral attachment to the cell. I) Binding interaction of Gn. a) figure of 6f8p prepared for docking. The red sphere represent the docking site. b) the 2D diagram for PCN binding to the Gn. c) the binding score for different poses. II) Binding interactions of Gc. d) Figure of 4HJ1 (chain c). The orange sphere represents the docking site. e) The docking interaction of PCN with Gc. f) The docking score of the binding poses (n = 9)

Protein involved in viral replication

Following viral attachment and cellular entry, the virus starts its replication process. An integral protein in RVFV replication cycle is the cap-binding domain of RVFV L protein (PDBID: 6QHG). This protein is involved in viral replication. It binds to the cellular mRNAs and cleaves off its cap structures to initiate the viral transcription. This protein is gaining a large research interest as a viral target to inhibit RVFV infectivity [61, 70].

Molecular docking simulation revealed that PCN binds to the cap binding protein with an affinity of -6.4 kcal/mol. The binding affinity was higher than that of prodigiosin (− 6.3 kcal/mol) as well as the co-crystallized ligand (7N-methyl-8-hydroguanisine-5'-triphosphate) which recorded a binding affinity of − 5.8 kcal/mol [60]. The strong binding was achieved via H-bonding with the ARG 1716 with a bond distance of 2.22435 A°. This is in addition to hydrophobic interactions between the flat area of the receptor amino acids PHE 1713 and the phenyl group of PCN. Interestingly, the binding mode involved the ARG1716 and the TYR1728 which are shared as the binding amino acids with the co-crystallized ligand which further support the binding capacity and subsequent inhibition (Fig. 5).

Fig. 5.

Binding interaction of PCN to the RVFV cap-binding protein. a) the prepared 6QHG for docking. The orange sphere represents the docking site. b) The 2D docking interaction of PCN to 6QHG. c) The 2D diagram of the co-crystallized ligand, 7N-methyl-8-hydroguanosine-5'-triphosphate binding interaction with 6QHG

Other potential target of the RVFV is the Nucleocapsid Protein (N protein) [62]. It plays an integral part in viral replication. The in silico screening results revealed that the binding affinity was found to be − 5.9 and − 5.6 kcal/mol for PCN and prodigiosin, respectively (Fig. 6). Table 3 summarizes the binding affinities achieved by PCN and prodigiosin.

Fig. 6.

Binding interaction of PCN to the RVFV nucleocapsid protein (N protein). a) prepared 3LYF protein for docking. The sphere represents the docking site. b) The 2D diagram for binding interaction of PCN to protein. c) the interaction scores of the 9 binding modes

Table 3.

Binding affinity of MP with acceptable drug likeness to variable targets in the RVFV protein

MP	Binding affinity Kcal/mol
	L protein	Gn	Gc	N protein	DC-SIGN
Prodigiosina	− 6.3	− 5.5	− 7.4	− 5.6	− 5.1
PCNa	− 6.4	− 6.4	− 7	− 5.9	− 5.5

^a expected blood brain barrier crossing ability

The results indicated that PCN and prodigiosin may have the potential to exert antiviral activity against RVFV. In this study, PCN as produced from Pseudomonas aeruginosa was selected to investigate its antiviral potential. PCN is a phenazine based compound that is produced as a secondary metabolite and virulence factor of P. aeruginosa [59]. The molecule has gained attention over the past few years for its biological attributes. This is mainly due to its ability to cross cell membranes as supported by its low molecular weight and Zwitter ion characteristic along with its redox-modulating activity which have contributed to its anticancer and antimicrobial abilities [59, 71].

MD simulations

MD simulation of the PCN binding to RVFV protein was performed to further confirm the in silico screening results. The RVFV cap binding protein was selected for the MD simulation. This is based on the in silico docking results which showed that the binding affinity of PCN exceeded that of the co-crystallized ligand suggesting potential protein inhibition. This was supported by the observation that PCN in vitro inhibition was more likely to be induced via halting viral replication rather than viral cell entry.

It is worth noting that among the proteins involved in viral replication, RVFV cap-binding protein showed the stronger binding capacity over the N protein suggesting its potential role in inhibiting viral replication by PCN accordingly it was chosen to further investigate its stability of binding.

The MD simulation studies came in good agreement with the docking study though the ligand showed a dissociation from the protein then rebinding to the same active site Fig. 7A–J. As shown in Fig. 7C, the RMSD value of the protein shows a fluctuating trend during the first 100 ns before stabilizing at an average of 2.5 Å in the next 80 ns then fluctuating again in the last 20 ns. The reason for these fluctuations is shown in the RMSF analysis. As can be seen, the N-terminal amino acids are highly fluctuating with the rest of the protein showing stability with an RMSF < 2.5 (Fig. 7B). The RMSD of the protein was recalculated while excluding the N-terminal region from 1704 to 1711 which showed that the protein is stable and equilibrated during the whole simulation (see supplementary material 2). Figure 7A, B show the protein RMSD and RMSF. In Fig. 7C, the ligand RMSD was calculated. The molecule showed an initial stability during the first 37 ns with an average around 3.5 Å before dissociating (high RMSD values) and then rebinding again to the same site starting from 160 ns to the end of the simulation. Similarly, the distance between the center of mass of the ligand to the center of mass of the protein shows a similar trend indicating the dissociation and rebinding (Fig. 7D). The radius of gyration, and SASA (Fig. 7E, F) show minimal fluctuations with averages of 15 Å, and 7838 Ų, respectively. The number of H-bonds formed between PCN and the protein was calculated (Fig. 7G). During the time when the ligand was binding to the protein there was one H-bond formed (Table 4).

Fig. 7.

Analysis performed on the trajectory of cap-binding L protein _PCN complex. a) RMSD values from the trajectory for the protein. b) Protein RMSF values of the carbon-alpha atoms. c) Ligand RMSD values. d) Change in the distance between the center of mass of PCN and that of the protein. e) Radius of Gyration (RoG) for the protein. f) SASA values. g) Number of H-bonds formed between the PCN ligand and the protein

Table 4.

Number and types of interactions detected by the PLIP webserver for the interaction between PCN and the cap binding protein that showed the binding of PCN to the protein, name the protein

Cluster Number	Number of Hydrophobic Interactions	Amino Acids in receptor	Number of Hydrogen Bonds	Amino Acids in receptor
C2	4	Gln1717—Tyr1728 (3)	1	Arg1716
C3	5	Phe1713 (2)—Tyr1728 (2)—Ile1747	0	None
C5	2	Phe1713—Pro1755	2	Arg1716—Gln1717

The MM-GBSA technique is a widely used computational method for estimating the binding free energy of ligands to proteins. This approach combines molecular mechanics energies with continuum solvation models to approximate the binding affinity. This method decomposes the total binding free energy into several components, including van der Waals interactions, electrostatics, polar solvation, and nonpolar solvation terms [48]. Figure 8 depicts the various components that contribute to the MM-GBSA binding free energy analysis.

Fig. 8.

Energy components of cap-binding L protein -PCN complex. Bars represent the standard deviations. The figure shows that the binding was achieved mainly by Egas, which is the summation of the summation of van der Waals interaction (hydrophobic interaction; vdw) and the electrostatic interaction (Eel). The E solv interaction composed of the non-polar surface area (surf) and the solvation energy (gb) did not contribute to the binding interaction. The estimated total binding energy was − 13.55 kcal/mol

The average total binding energy of the PCN compound is − 13.55 ± 3.32 (standard deviation) kcal/mol during the last 40 ns (400 frames) during which the ligand was bound to the protein. With an average value of − 17.17 ± 3.01 kcal/mol, the van der Waals energy was found to be the most favorable, this was followed by the electrostatic interaction energy with an average value of − 8.6 ± 7.66 kcal/mol. The decomposition analysis (Fig. 9) was employed to detect amino acids that exist within 1 nm of the PCN molecule and contribute to the interaction in a value smaller than − 1 kcal/mol. Only two amino acids were found namely Phe1713 (− 1.28 kcal/Mol), and Met1782 (− 1.52 kcal/Mol).

Fig. 9.

Binding free energy decomposition of the cap-binding L protein -PCN complex. The figure shows that the PCN exhibit the strongest binding through its interaction with Phe1713 which is shared with the co-crystallized inhibitor indicating efficient binding

The trajectory was then grouped into five clusters, using TTClust, and a representative frame was generated for each cluster. Next, the number and types of interactions were detected using PLIP webserver for only the representatives that showed the binding of PCN to the protein (three cluster representatives out of the five; Table 4). It is worth pointing out that PLIP provides a way to thoroughly analyze and visually represent how protein–ligand complexes interact on atomistic level, through distance and angle-based thresholds reported in literature. Therefore, it is useful in this study as it identifies the interacting amino acids [44]. A total of 3 H-bond interactions and 11 hydrophobic interactions were detected from PLIP. One amino acid (Phe1713) is common in two cluster representatives and formed hydrophobic interaction with the PCN. Moreover, the results from the ProLIF python library indicated that the Phe1713 and Met1782, the two amino acids with the highest interaction energy from MM-GBSA, had frequencies of occurrence of 37.3% and 36.4%, respectively, indicating their role in stabilizing the interaction (Fig. 10). Figure 11 shows the 3D conformations and the interactions detected by PLIP of the ligands inside the binding pocket. These results support the strength and stability of PCN binding to the RVFV cap binding protein with a potential subsequent inhibition (supplementary material 3).

Fig. 10.

Types and frequency of different interactions produced from ProLIF python library for the simulation time of 200 ns

Fig. 11.

Types and number of different ligand-protein interactions produced from PLIP for each cluster representative. H-bond is represented by a blue solid line, hydrophobic inter-action is represented by a dashed grey line, amino acids is represented by blue sticks, and PCN is represented by orange sticks

Collectively, the results indicate that PCN exerts stable interaction with the RVFV cap-binding L protein. This stable binding suggests a stronger inhibition of the viral transcription over inhibition of viral attachment and providing a potential mechanism of antiviral effect of PCN.

In vitro studies

Extraction, purification and analysis of PCN

The promising potential of PCN as a drug candidate for RVFV was validated by in vitro studies to support the in silico data. PCN was extracted from P. aeruginosa culture by liquid–liquid extraction using chloroform as the organic phase [72]. Chloroform was selected as an efficient organic solvent for PCN extraction in terms of yield and purity [72, 73]. Acid–base alteration was performed for purification of the blue pigment each with fresh water and chloroform [50]. In the acidic medium (0.1 N HCl), PCN is protonated and converted to water soluble red pigment that is extracted solely in the aqueous phase. Upon converting the aqueous phase to basic pH by the addition of 0.1 N NaOH, PCN returns to its unionized state causing salting out from water and moving to the chloroform layer. The procedure was repeated for 3 cycles to ensure further purification of PCN. The purified PCN yield was 30 mg from 100 mL of bacterial culture medium.

Spectral analysis of the blue pigment (Fig. 12) revealed that PCN exhibited a characteristic absorption at 699 nm (Fig. 12). This was concordant with previous report indicating the UV–Vis absorbance of PCN in organic medium [51, 74, 75]. The MS spectrum showed that the extracted PCN corresponding peak appeared at m/z 211.05 which agrees with the molecular weight of acidified PCN (Fig. 12). Additional bands appeared at m/z 233 and 249 which correspond to the sodium and potassium adducts of PCN. The tandem MS fragmentation of PCN showed that the highest abundance of the fragments was observed at m/z 196, 182, and 168. These bands correspond to the loss of CH₃, NH₂ and O⁻ and are explained by the chemical structure of PCN (Fig. 12). Similar fragmentation patterns were observed when different collision energies were utilized indicating consistency of the fragmentation [51].

Fig. 12.

Spectral analysis and chromatographic analysis of the extracted PCN. a) the UV–Vis spectrum of PCN extracted from culture media. b) the MS spectrum of PCN showing the m/z 211.05 peak indicating the accurate mass spectrum of the extract along with its sodium (m/z 233) and potassium adduct (m/z 249) adducts. c) The abundance (%) of PCN m/z 211.05 fragments showing major peaks appearing at m/z 168, 183 and 196. d) proposed fragmentation pattern of PCN showing the fragments corresponding to bands at m/z 168, 183 and 196

Antiviral assay

In this study the tissue culture infectious dose 50% (TCID₅₀) method was applied for the assessment of the antiviral potential of PCN. The assessment was based on the observation of the ratio of CPE in Vero cells as an indication of the viral infectivity in PCN treated cells versus IFN-α treated and untreated cell cultures [53].

Initially, the MTT assay was performed to determine the sole effect of PCN on the viability of Vero cells. The concentration that maintained viability of the Vero cells was considered cell-safe and was further recruited for the estimation of the antiviral potential of PCN. The MTT assay revealed that 0.49 μg/mL of PCN maintained the Vero-cell viability (100.97 ± 3.30) accordingly it was assigned as the cell-safe concentration and was used for further investigation of the antiviral assay.

The antiviral assay results showed that the incubation of PCN (0.49 μg/mL) with the Vero-cells for 24 h prior to the viral infection resulted into a 2.00 Log₁₀ reduction in the TCID₅₀ versus untreated cells. Similarly, treatment of previously RVFV infected Vero-cells with PCN (0.49 μg/mL) induced a 2.89 Log₁₀ reduction in the viral titer. The observed reduction in the TCID₅₀ than negative control (untreated cells) indicates the possible antiviral potential of PCN against RVFV. The viral load which was capable of inducing CPE (infecting 50% of cultured cells) became ineffective in presence of PCN. INF-α was used as a positive control for being the potent inhibitor of RVFV [76]. The results showed that INF-α induced a 5.1 ± 0.163 Log₁₀ reduction when it was applied to RVFV infected cells while a 4.1 ± 0.127 Log₁₀ reduction in TCID₅₀ when RVFV infection was induced on a prior INF-α treated cell culture. Unfortunately, despite being effective, INF-α based treatment is a high-cost regimen that is not applicable in situations or geographical regions where RVFV is expected [77]. There are intense research efforts for the discovery of an effective small-molecule based therapy for RVFV. Literature review revealed that ribavirin induced a 1.9 Log₁₀ reduction in RVFV TCID₅₀ when high drug concentration (62.5 μg/mL) of the drug was used [78]. In the same context, suramin, an FDA approved antiviral agent, exhibited a one log reduction in the RVFV titer [79]. A close reduction in the RVFV titer was observed when favipiravir was tested pending the use of high concentration (> 20 μM) [80, 81]. The study presents PCN as a potential inhibitor of RVFV that may be achieved at a low production cost (Fig. 13).

Fig. 13.

a) Estimation of cellular viability of Vero cells upon treatment with different concentrations of PCN. The figure shows that the cell-safe concentration was 0.49 μg/mL. b) comparing the viral titer in untreated cells (TNT) versus cells treated with PCN before infection (TBT; indirect) and cells treated post infection (TPT; direct).c) the percentage reduction in TCID₅₀ post treatment

The observed inhibitory pattern suggests that PCN may have a higher potential to interfere with viral replication, as indicated by a 2.89 log₁₀ reduction in TCID₅₀, rather than the ability to manifest a cytoprotective effect against RVFV (2.00 log₁₀ reduction). These observations support the in silico results where the molecular docking indicated a moderate potential of PCN to bind to the cellular receptor (DC-SIGN) which mediates viral entry. On the other hands, the higher tendency to inhibit the viral replication (2.89 log₁₀ reduction) can be explained by the deduced binding affinity and stability of the PCN binding to the RVFV cap-binding protein (Fig. 13). These results pose PCN as a promising inhibitor of RVFV. Its activity may be better manifested in reducing the RVFV propagation in infected cells and to a lesser extent in protecting the cells against RVFV infection.

While the study demonstrates promising in silico and in vitro effect of PCN against RVFV, it has some limitations. The ADME screening and brain uptake were evaluated in silico using the ADME-TOX studies of the compound using the Swiss-ADME tool. Although computational screening during this stage limits sacrificing animals, the promising results from the computational and in vitro investigations warrant future in vivo studies to investigate the mechanism of action of PCN, its effective dose, and brain uptake. Results of these investigations could support the utility of PCN as an anti-RVFV molecule. PCN used in this study was extracted from P. aeruginosa cultures. Although purification of PCN from bacterial cultures is most relevant to the efforts of low- and middle-income countries to combat viral pandemics, future studies should compare antiviral activity of PCN purified from bacterial culture against that of high purity commercial PCN.

Conclusions

According to the WHO, the RVFV is listed as one of nine prioritized diseases and is predicted to become the next pandemic [4]. Efforts are currently underway to develop RVFV inhibitor or vaccine. However, the design and de novo synthesis of a new RVFV inhibitor via the conventional methods are time consuming. They encompass a research journey from compound design to organic synthesis and clinical approval that may take several years to complete. In addition, the success rates of de novo drug synthesis and validation are relatively low and the development of viral resistance is highly likely to occur. In this context, multiple computational screening studies for the discovery of RVFV inhibitors from natural sources were conducted [63, 82]. Alternatively, the repurposing of previously approved FDA drug and chemical libraries has been investigated [83–85]. Despite the potential of such strategies to reduce the drug discovery time and cost, to date, there is no remedy or effective vaccination for RVFV [86]. It is worth mentioning that computational and in silico screening were proven successful for the fast drug development especially during public health concerns and epidemics. Examples include the discovery of raltegravir, vemurafenib, and molnupiravir for treating HIV, melanoma, and Covid-19, respectively [87]. On the other hand, MPs present a plethora of readily available small molecules which could be exploited for drug discovery [88]. Recruitment of MPs as drug candidates has multiple advantages. These include a lower carbon footprint production compared to that of molecules prepared using organic synthesis, short synthesis route, and a more economic scaling than drugs prepared by organic synthesis [89].

This study investigated the discovery of an anti-RVFV small molecule from MPs as a green and low-cost drug development approach. Thirteen MP were screened for the discovery of a potential inhibitor of RVFV. Four out of the thirteen studied MP candidates showed acceptable drug-likeness as indicated by multiple medicinal chemistry filters. Two of which (PCN and prodigiosin) may have a higher probability to cross the BBB rendering them promising candidates for the treatment of RVFV infection which mainly replicates in the brain. The in silico studies revealed that PCN had stronger interaction potential against the key proteins involved in the RVFV infectivity and thus it was selected for additional investigations.

PCN was efficiently produced from bacterial culture at a yield of 3 mg/mL with good purity as indicated by UV and MS. The in vitro antiviral activity of PCN was investigated in two experiments: (1) when PCN was incubated with RVFV infected cells, and (2) when it was incubated with the cells prior to the infection to give insight on the ability of PCN to inhibit RVFV replication and protect cells against viral infection; respectively. In both cases the TCID₅₀ was recorded and compared to positive (INF treated cells) and negative controls (untreated cells). The results suggest that PCN may exhibit a significant inhibition of RVFV infection when incubated with cells post infection (2.89 log₁₀ reduction in TCID₅₀) rather than a cyto-protective effect (2.00 log₁₀ reduction). These observations agree with the in silico molecular simulation studies where PCN exhibited a strong and stable interaction with the cap-binding L protein involved in RVFV transcription. The data demonstrates the possible anti-viral effect of PCN. Meanwhile it highlights that PCN may be easily and effectively produced from bacterial cultures at a low-cost of production.

In conclusion, the study highlights the potential of MPs as antiviral agents with multiple molecules exhibiting acceptable drug-likeness properties. The data suggests that the MP, PCN can be effectively produced, extracted and purified from bacterial cultures. Importantly, this is the first report to show the remarkable in silico and in vitro antiviral activities of PCN against key RVFV proteins and its potential as an effective low-cost tool to combat RVFV. These results present PCN as a possible effective and low-cost small molecule that could represent an effective tool to combat RVFV. The promising results from the computational and in vitro investigations warrant future in vivo studies to further evaluate the utility of PCN as an anti-RVFV molecule. Furthermore, animal testing to show brain uptake and effective dose of PCN will be required in future studies to evaluate the utility of PCN as an anti-RVFV drug. Future studies could also investigate the potential of other MPs such as prodigiosin, violacein and bikaverin which demonstrated good binding affinity.

Author contributions

F.F has contributed to the experimental design, project management, methodology, writing the manuscript and revision. I.M.I has contributed to methodology, writing the manuscript and revision. H.A. has contributed in project management, writing the manuscript and revision.

Funding

Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooperation with The Egyptian Knowledge Bank (EKB).

Data availability

This manuscript does not report data generation or analysis, All related data are submitted in the paper. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.