Identification of Inhibitors for Flagellar Assembly Protein FliN of Uropathogenic Escherichia coli using Virtual Screening and Molecular Dynamics Simulation Study

Abstract

Escherichia coli (E. coli) is a gram-negative bacterial pathogen that poses a significant clinical and epidemiologic challenge. The selection pressure brought by the insufficient use of antibiotics has resulted in the emergence of multi-drug-resistant E. coli in the past ten years. Computational and bioinformatics methods for screening inhibitors have significantly contributed to discovering novel antibacterial agents. One possible target for novel anti-virulence drugs is motility. Motility inhibitors are generally effective at concentrations lower than those required for the antibacterial properties of traditional antibiotics, and they are likely to exert less selective pressure than current medicines. Motility may be essential for bacteria to survive, find nutrients, and escape unfavorable environments and biofilm formation. The FliN is a protein forming the bulk of the C ring of the flagella and is present in multiple copies (more than 100) in bacteria. Its absence in mammals makes it an attractive drug target for drug discovery. Two-thousand seven hundred seventy-eight natural compounds from the ZINC library were screened against FliN (PDB ID: 4YXB) using PyRx AutoDock Vina, and the top compounds were selected for secondary screening after sorting the results based on their binding energy. Based on interactional analysis, binding energy (− 7.78 kcal/mol), and inhibition constant (1.98 µM), ZINC000000619481 was the best inhibitor. This compound binds exactly as per the defined active site residues of the receptor protein. Also, molecular dynamics was performed. The eigenvalue of the selected complex was 1.241657e−05. There were no ADME properties outside of the specified range for the identified hit; it fitted exactly to the binding site of the FliN receptor well and was found to be stable in MD simulation studies. Further in vitro and in vivo studies are needed to confirm its anti-bacterial activity and use as a potential antimicrobial drug against urinary tract infections caused by E. coli.

Introduction

Urinary tract infections (UTIs) rank among the most common infections worldwide [9]. Various microorganisms, including gram-negative and gram-positive bacteria and candida spp. cause UTIs. However, the most often identified bacteria responsible for both simple and complex UTIs is Uropathogenic Escherichia coli (UPEC) [9, 10]. Since carbapenems are the last resort of antibiotics, the rise of UPECs resistant to them, including NDM-1, has garnered attention and evolved into the biggest threat [32]. The potential of UPEC to develop biofilms on catheters and result in nosocomial infections is another issue. Treatment of UTIs is becoming a challenge as UPEC has developed resistance to major classes of antibiotics and, recently, to colistin [20]. So, there is a current need to find novel drugs by using novel drug targets that will activate against UPEC.

Flagella causes motility in bacteria. Motility may be essential for bacteria to survive, find nutrients, and escape unfavorable environments. FliN is a protein forming the bulk of the C ring of the flagella and is present in multiple copies (more than 100) in bacteria [2]. In E. coli, the FliN protein plays a role in the building and operation of flagella. The flagella motor protein comprises three proteins: FliM, FliN, and FliG. The primary subunit of the flagellar motor, FliN, is found in the basal body [36]. It reacts to different chemo-attractants by causing translational or rotational motion [23, 31]. In addition to its functions in motor switching and rotation, FliN is thought to have a role in the export of proteins that form the exterior structures of the flagellum (the hook, rod, and filament).

Along with movement, motility plays a critical role in the early stages of infection, colonization, and biofilm formation by several harmful bacteria [12]. Our previous study found FliN to be a unique target for screening novel inhibitors. It is non-homologous to the human host and has higher protein–protein interactions. From comparative analysis (preliminary work done by us), it was found to be a broad-spectrum drug target, essential for the pathogen's survival, interacting with a higher number of proteins and having no homologs with the human host [19]. Since ancient times, natural ingredients have been utilized to treat bacterial and viral illnesses in Indian and traditional Chinese medicine, and their use has previously been proven to be safe [3]. Virtual screening of natural substances could offer a different method for locating possible therapeutic candidates to fight infectious diseases [37]. In the current study, we propose using the FliN protein crystal structure to virtually screen natural compounds from the ZINC library. This early work employs molecular dynamics (MD) simulations and in silico analysis to assess the effects of natural substances on the FliN receptor protein by analyzing their binding postures and interactions.

Material and Methods

Protein Sequence and Tertiary Structure Prediction

The amino acid sequence of the FliN protein was saved from the NCBI as a FASTA format. The three-dimensional structure (3D) FliN protein model was modeled using the Phyre2 server (Protein Homology/analogy Recognition Engine V 2.0). A web-based tool package called Phyre2 was used to predict and analyze protein structure, function, and mutations. [21]. The web tool was accessed at http://www.sbg.bio.ic.ac.uk/phyre2.

3D Structure Validation

The tertiary structure validation step of the model generation process is crucial as it helps to detect any inaccuracies in the modeled structure. For 3D structure validation, the ProSA web server was initially utilized, which calculates a quality score for the precise input structure and displays it as a Z-score. Errors are likely present in the structure if the Z-score is not within the range of properties for native proteins [38]. An ERROR web server was used to estimate non-bonded atom–atom interactions. A Ramachandran plot was obtained through the RAMPAGE web service, which displayed the residues in allowed and restricted regions and explained the quality of the modeled structure.

Preparation of Macromolecule and Active Site Prediction

The crystal structure of the FliN of E. coli was modeled by the Phyre2 server. We eliminated additional molecules, including water and ligands, using AutoDock 4.0's protein preparation wizard [34]. An online tool called the Computed Atlas of Surface Topography of Proteins (CASTp) was used to identify, define, and quantify concave surface regions on three-dimensional protein structures [7].

Ligand Preparation (Natural Compounds)

A collection of 2778 natural compounds was obtained from version 15 of the ZINC database. [35]. These molecules' structural coordinates were obtained in the.sdf (structural data file) format by downloading them from the database. Each compound has a 2D format. As a result, they were transformed into a 3D format using “Open Babel software before the virtual screening, carefully ensuring that the ligands' structural coordinates were not altered in the process [30].

Virtual Screening

Virtual screening for selecting the probable inhibitors among the library of natural compounds was done using PyRx software [5]. Utilizing the Lamarckian genetic approach, we employed a PyRx-based sorting function [5]. A grid box was created to define the FliN protein's interaction region. The virtual screening results were selected based on the binding score.

ADMET Analysis

Following the initial screening, compounds were sorted using the Lipinski rule of five [13] and subjected to ADME screening using the SWISS ADME web tool [4]. The toxicity of those compounds with lead-like properties was then assessed using the Lazar toxicity module [4, 27].

Re-Docking via Molecular Docking

Molecules passing the primary and secondary filters were docked against the FliN to choose the best hit among the final hits. We used AutoDock4.0 to do the molecular docking [34]. Using a Lamarckian genetic process, lead molecules were docked into the active site residues designated by the grid box. AutoDock scoring function is based on the AMBER force field ff19SB” [39]. Hydrogen bond formation and the docked complex's binding score were studied.

Molecular Dynamics Simulation

Based on the docking results, the molecular dynamics simulation study was conducted for the best ligand-receptor complex molecules. For the molecular dynamics simulation study, the iMODS online server—a quick, free, and easily available molecular dynamics simulation server for defining and measuring protein flexibility was utilized [25].

Results

Tertiary Structure Modeling

Based on six templates, the Phyre2 server predicted six tertiary 3D structures of the FliN protein. We selected the top model based on the template c4YXBA. The confidence score of this model is 99.9%, whereas identity is 93% and coverage is 67%. The single highest-scoring template has modeled 89 residues (67% of the sequence) with 99.9% confidence (Fig. 1A). The template is a PDB molecule, flagellar motor switch protein fliM, flagellar motor switch. The PDB title of this template is FliN fusion protein.

Fig. 1.

Protein 3D modeling, refinement, and validation. A The 3D model of protein FliN modeled by Phyre2 server. B Validation by RAMPAGE AND VADAR tool analysis shows 88% in favored regions, 8% in allowed regions, 3% in the generous region, and 1% in outlier regions. C ProSA web with Z-score − 1.93

Validation of 3D Structure

The RAMPAGE web server exposed the improved structure to the Ramachandran plot analysis. The VADAR web tool plot exposed 88% in favored regions, 8% in allowed regions, 3% in the generous region, and 1% in outlier regions (Fig. 1B). ProSA's web server and ERRAT confirmed the 3D model's quality and possible flaws. Following refining, the selected model had an overall quality factor of 88.4%. The Z-score for the 3D model was estimated to be − 1.93 by the ProSA web server (Fig. 1C). The overall results from RAMPAGE, VADAR, ERRAT, and ProSA web servers have validated the 3D-modeled proteins’ outstanding quality.

Protein Preparation and Active Site Identification

The Phyre2 server modeled the crystal structure of the FliN from E. coli. To eliminate other molecules like water and ligands, we employed the protein preparation wizard in “AutoDock 4.0 [35]. Locating, defining, and quantifying concave surface regions on three-dimensional protein structures can be done online with the help of the Computed Atlas of Surface Topography of Proteins (CASTp). CASTp identifies the active site residues with surface area (Fig. 2). Active site residues identified by CASTp server are GLN-52, ASP-53, ILE-57, MET-58, ILE-60, PRO-61, VAL-62, LEU-64, VAL-66, LEU 92, GLY 94, ILE-99, ILE-101, ILE-106, ALA-107, LYS-117, TYR-118, GLY-119, VAL-120, ILE-122, ILE-125, PRO-128, ARG-131, MET-132, LEU-135.

Fig. 2.

Active site prediction by CASTp server. The red surface area depicts the active sites (color figure online)

Virtual Screening (Natural Compounds Results)

Preliminary screening of the ZINC library containing 2778 natural compounds was conducted against FliN using the PyRx, a graphical interface containing AutoDock and AutoDock Vina wizard. After primary screening, compounds were sorted by their binding score. We selected the top 30 compounds with binding energy <  = − 7 kcal/mol from the initial screening.

Secondary Screening

After preliminary screening, the top 30 compounds were selected and further subjected to secondary filters like ADME studies and toxicity prediction. Out of these 30 compounds, all passed ADME studies and toxicity prediction.

Molecular Docking

Molecular docking was carried out for the 30 selected compounds (hits) onto the FliN protein. A Binding score was compared to find the best compound with hydrogen bonds and hydrophobic interactions and selected the top five compounds (Table 1).

Table 1.

Binding interactions of selected molecules with FliN

Sr. no	ZINC id	Binding energy (kcal/mol)	Inhibition constant (µM)	Hydrogen bond interactions	Hydrophobic interactions
1	ZINC000000619481	− 7.78	1.98	LEU-92, GLY-94	–
2	ZINC000001860195	− 7.15	5.75	ASP-90, GLY-91, GLU-95, ASP-98	–
3	ZINC000012933679	− 7.45	3.46	LEU-97, GLU-110, VAL-111	LEU-97, VAL-111
4	ZINC000072152828	− 7.56	2.89	LEU-89, LEU-92, PRO-96	LEU-89, LEU-97
5	ZINC000240122381	− 7.02	7.16	GLY-91, LEU-92, GLY-94, VAL-111	LEU-97, VAL-111

Validation of the docking protocol was also confirmed by re-docking the initial inhibitor of FliN, which has 0.00 Å RMSD between the docked pose and the crystal structure. We found a pattern similar to that of the experimental co-crystal structure. Interactional analysis of the top five compounds is shown in Figs. 3, 4, 5, 6, and 7.

Fig. 3.

A 2D interaction of FliN with ZINC000000619481. Green dotted lines represent hydrogen bonds. B 3D interaction of FliN with ZINC000000619481. Blue lines represent hydrogen bonds (color figure online)

Fig. 4.

A 2D interaction of FliN with ZINC000001860195. Green dotted lines represent hydrogen bonds. B 3D interaction of FliN with ZINC000001860195. Blue lines represent hydrogen bonds (color figure online)

Fig. 5.

A 2D interaction of FliN with ZINC000012933679. Green dotted lines represent hydrogen bonds. B 3D interaction of FliN with ZINC000012933679. Blue lines represent hydrogen bonds, and grey dotted lines show hydrophobic interactions (color figure online)

Fig. 6.

A 2D interaction of FliN with ZINC000072152828. Green dotted lines represent hydrogen bonds. B 3D interaction of FliN with ZINC000072152828. Blue lines represent hydrogen bonds, and grey dotted lines show hydrophobic interactions (color figure online)

Fig. 7.

A 2D interaction of FliN with ZINC000240122381. Green dotted lines represent hydrogen bonds. B 3D interaction of FliN with ZINC000240122381. Blue lines represent hydrogen bonds, and grey dotted lines show hydrophobic interactions (color figure online)

Molecular Docking Interactional Analysis of Natural Compounds (ZINC Library)

Five compounds, ZINC000000619481, ZINC000001860195, ZINC000012933679, ZINC000072152828, and ZINC000240122381, were selected for interactional analysis.

Two hydrogen bonds stabilized the complex between ZINC000000619481 and FliN. Amino acid residues involved in hydrogen bonding with the ligand were LEU-92 and GLY-94. Other interactions involved in stabilizing the complex were pi-sigma and pi-alkyl interactions. The binding score of the receptor with ZINC000000619481 was found to be − 7.78 kcal/mol, and the inhibition constant is 1.98 µM. 2D and 3D plots of the ZINC000000619481 with FliN are shown in Fig. 3A, B, respectively.

Two hydrogen bonds stabilized the complex between ZINC000001860195 and FliN. Amino acid residues involved in hydrogen bonding with the ligand were ASP-90, GLY-91, GLU-95, and ASP-98. Other interactions involved in stabilizing the complex were pi-sigma and pi-alkyl interactions. The binding score of the receptor with ZINC000001860195 was found to be − 7.15 kcal/mol, and the inhibition constant is 5.75 µM. 2D and 3D plots of the ZINC000001860195 with FliN are shown in Fig. 4A, B, respectively.

Two hydrogen bonds stabilized the complex between ZINC000012933679 and FliN. Amino acid residues involved in hydrogen bonding with the ligand were LEU-97, GLU-110, and VAL-111, as well as hydrophobic interactions with LEU-97 and VAL-111. Other interactions involved in stabilizing the complex were pi-sigma and pi-alkyl interactions. The binding score of the receptor with ZINC000012933679 was found to be − 7.45 kcal/mol and the inhibition constant is 3.46 µM. 2D and 3D plots of the ZINC000012933679 with FliN are shown in Fig. 5A, B, respectively.

Two hydrogen bonds stabilized the complex between ZINC000072152828 and FliN. Amino acid residues involved in hydrogen bonding with the ligand were LEU-89, LEU-92, and PRO-96, and hydrophobic interactions with LEU-89 and LEU-97. Other interactions involved in stabilizing the complex were pi-sigma and pi-alkyl interactions. The binding score of the receptor with ZINC000072152828 was found to be − 7.56 kcal/mol and the inhibition constant is 2.89 µM. 2D and 3D plots of the ZINC000072152828 with FliN are shown in Fig. 6A, B, respectively.

Two hydrogen bonds stabilized the complex between ZINC000240122381 and FliN. Amino acid residues involved in hydrogen bonding with the ligand were GLY-91, LEU-92, GLY-94, and VAL-111, and hydrophobic interactions with LEU-97 and VAL-111. Other interactions involved in stabilizing the complex were pi-sigma and pi-alkyl interactions. The binding score of the receptor with ZINC000240122381 was found to be − 7.02 kcal/mol, and the inhibition constant is 7.16 µM. 2D and 3D plots of the ZINC000240122381 with FliN are shown in Fig. 7A, B, respectively.

Molecular Dynamics Simulation of the Best Compounds

The results of molecular dynamics simulation and normal mode analysis (NMA) of the best complex ZINC000000619481 and FliN are shown in Fig. 8A. The molecular dynamic simulation study was conducted to know the movements of atoms and molecules in the drug molecule ZINC000000619481. The deformability graph of the complex shows the peaks in the graphs representing the protein region with deformability (Fig. 8B). The eigenvalue of the complex is 1.241657e−05, as shown in Fig. 8C. The higher frequency represents localized displacements, whereas low energy modes show collective conformational changes. The variance graph represents the cumulative variance in green color and individual variance in red color, as shown in Fig. 8D. The B-factor graph visualizes the relation of the docked complex between the NMA and the PDB, as shown in Fig. 8E. The co-variance map of the complex where the correlated motion between a pair of residues is indicated by red color, anti-correlated by blue color, and uncorrelated by white color as depicted in Fig. 8F. The complex’s elastic map shows the relation between the atoms, and dark grey regions represent stiffer regions” (Fig. 8G).

Fig. 8.

Molecular dynamics simulation result of FliN and ZINC000000619481 docked complex. NMA mobility (A), deformability (B), eigenvalues (C), Variance (D), B-factor (E), co-variance map (F), and elastic maps (G) (color figure online)

Discussion

Motility is a potential target for new antivirulence agents. Unlike existing antibiotics, motility inhibitors are anticipated to exert less selective pressure. They are usually effective at concentrations lower than those required for traditional antibiotics to have their antibacterial effects [12]. In Campylobacter jejuni (C. jejuni), the significance of motility as a virulence factor was initially recognized when only the motile strain of C. jejuni was isolated from an infection site where a combination of motile and nonmotile C. jejuni phase variations was employed to initiate infection [12]. Flagellum-mediated motility has since been shown to increase the pathogenicity of several Gram-negative bacteria, such as “Bordetella bronchiseptica, Bordetella pertussis [14], Helicobacter sp. [17], Legionella pneumophila [14], pathogenic E. coli strains [16], Pseudomonas aeruginosa [6], and Vibrio cholerae” [28]. In addition to aiding in host attachment, flagella are involved in the creation of biofilms by various bacterial pathogens, which facilitate the development of antibiotic resistance and chronic and recurrent infections [16]. Since biofilm formation is an important pathogenetic mechanism of UTIs caused by UPEC, we explored the Flagellar assembly protein FliN as a selective drug target. Previous studies have shown that migration and rotation of the bacteria in semisolid and liquid media are affected adversely by the inactivation of the FliN gene [24]. Previously, these motor switch proteins were discovered and used as Vibrio cholerae and Leptospira interrogans therapeutic targets [11, 12]. The most well-known inhibitors of sodium-driven bacterial flagellar motors in a variety of Vibrio species, enterohemorrhagic Escherichia coli (EHEC), and uropathogenic Escherichia coli (UPEC) strains include amiloride and its analogs [12]. These days, natural products represent a major source of new pharmaceuticals or act as models for creating new synthetic medications, such as antibiotics and anticancer treatments [1].

ZINC is a freely available database containing around 230 million purchasable compounds (natural compounds, FDA-approved drugs, drug-like chemical compounds, synthetic drugs, etc.) for virtual screening in ready-to-dock three-dimensional formats. In an investigation by Henriksen et al. (2010), inhibitors of S. aureus were found by screening the ZINC library of natural chemicals against three distinct enzymes of the histidine biosynthesis pathway. Of the 49 chemicals that they obtained, 13 had an impact on the traditional disc diffusion approach. Another study by Kalia et al. [15] screened a library of natural compounds against LasR and found six novel potential QS-inhibiting compounds [15]. Similarly, a study by Navinraj et al. showed the antibacterial and antifungal properties of Nimbolide, which was extracted from leaves [29]. A study by Singhal L [33] screened the zinc database to identify MurB enzyme inhibitors for the Staphylococcus aureus pathogen and found ZINC 34230491 to be one of the best candidate drugs [33].

We consider FliN a valuable target as its mutational analysis has shown a strong polarizing effect on virulence [12]. Two-thousand seven hundred seventy-eight natural compounds from the ZINC library were screened against FliN (PDB ID: 4YXB) using PyRx AutoDock Vina, and the top compounds were selected for secondary screening after sorting the results based on their binding energy. All the selected compounds passed the secondary filter ADME and toxicity. Five compounds, ZINC000000619481, ZINC000001860195, ZINC000012933679, ZINC000072152828, and ZINC000240122381, were found to be non-carcinogenic in nature. Molecular dynamics simulations and Root mean square deviation (RMSD) were computed for the top selected five inhibitors. Of the final five hits, ZINC000000619481 was one of the best inhibitors against FliN.

A binding score of the receptor with ZINC000000619481 was found to be − 7.78 kcal/mol, and the inhibition constant is 1.98 µM. Amino acid residues involved in hydrogen bonding with the FliN and ligand were LEU-92 and GLY-94. This compound binds exactly as per the defined active site residues of the receptor protein. The eigenvalue of the selected complex was 1.241657e−05. The higher frequency represents localized displacements, whereas low energy modes show collective conformational changes [26].

Based on docking poses, interactional analysis, and molecular dynamics simulation, ZINC000000619481, a natural compound, was identified as the best inhibitor which fit the FliN receptor binding site well, showed no ADME attribute outside the allowed range, and was stable in the MD simulation testing. This compound might be examined for antibacterial activity and utilized as a starting point to find antimicrobial medications to treat E. coli-caused urinary tract infections.

Conclusion

The current study is the first report on utilizing the FliN protein as a drug target for screening natural compounds for treating urinary tract infections caused by E. coli. Based on docking poses, interactional analysis, and molecular dynamics simulation, compound ZINC000000619481 was the best inhibitor. Before the selected inhibitors can be used in clinical settings, further in vitro and in vivo validations are required.

Author Contributions

HK, and NT, Conceptualization. HK, Data curation. HK, Formal analysis. HK, Investigation. HK, Methodology. HK, Software. NT, Supervision. HK, and NT, Validation. HK, and NT Writing—original draft. HK and NT, Writing—review & editing.

Funding

None.

Declarations

Conflict of interest

None.