Genistein as a Potential Inhibitor of Tetrahydrofolate Biosynthesis in Combating Drug Resistance in Escherichia coli O157:H7

INTRODUCTION

There are 150 species in the Vigna genus of the Fabaceae family, including many essential food legumes.^[1] Vigna species are phylogenetically related to other agriculturally essential crops that belong to genera like Cajanus, Glycine, and Phaseolus.^[2] The most cultivated species of these genera are Vigna mungo (black gram), Vigna unguiculate (cowpea), Vigna radiata (green gram), Vigna subterranea (bambara groundnut), Vigna angularis (azuki bean), Vigna caracalla (snail bean), Vigna lanceolate (pencil yam), Vigna speciose (wondering cowpea), Vigna umbellate (red bean), etc. However, less genetic and genomic information is available on cowpea, mungbean, black gram, and azuki bean.^[3,4]

Different bioactive constituents like lectins, phytates, enzyme inhibitors, phenolic compounds, and oligosaccharides found in Vigna species are responsible for various metabolic functions against several diseases. Some other compounds in legumes exhibit antioxidant properties, including anthocyanidin pigments like delphinidin, cyanidin, pelargonidin, malvidin, and petunidin.^[5] The reduced cholesterol level and anticancer properties are attributed to saponins.^[6] It has been demonstrated that mucilage in black gram flour can help the sustained release of freely soluble drugs to develop new drug formulations.^[7] Green gram is rich in iron and phosphorus and highly nutritious.^[8] β-sitosterol, stigmasterol, soyasapogenol C, 1,4-butane diamine, 3-(carboxy methylamino) propanoic acid, 1H Imidazole, spermidine, spermine, amino acids, and peptides are among the pharmacologically significant chemicals identified from V. radiata seeds.^[9] Natural elicitors found in V. mungo sprouts include fish protein hydrolysates (FPH), lactoferrin (LF), and oregano extract (OE), all of which can activate the phenylpropanoid pathway (PPP) via the pentose phosphate and shikimate pathways.^[10] Earlier reports found that every legume component was high in trypsin inhibitors and that extracts of green gram seeds had immunostimulatory properties,^[11] which might be linked to an increase in humoral and cell-mediated responses, phagocytosis, and haematopoiesis in the treated rats.^[12]

The Gram-negative, rod-shaped bacterium Escherichia coli (E. coli) is a well-known member of the gut microbial flora. Gene transformation or foreign plasmid insertions are responsible for the pathogenicity of non-pathogenic E. coli strains.^[13] The bacteria that cause a haemolytic uremic syndrome (HUS) in humans are E. coli O157: H7, a Shiga-toxin-producing entero-haemorrhagic strain. It spreads primarily through meat contamination during slaughtering and packing. HUS causes haemolytic anaemia, thrombocytopenia, and renal failure by forming capillary clots and producing haemolytic anaemia, thrombocytopenia, and renal failure. The only therapy options are rehydration, anti-fever, and anti-pain medicines.^[14,15] Due to the outer membrane efficiency as a barrier and multidrug efflux pumps, the E. coli O157: H7 strain has substantially greater inherent resistance to several antibiotics. Efflux pumps reduce outer membrane permeability and antibiotic absorption, resulting in drug resistance.^[16] Efflux pumps affect nearly all antibiotic classes, including macrolides, tetracyclines, and fluoroquinolones. These antibiotics block DNA or protein synthesis; therefore, they must enter the cell to work. Some efflux pumps are drug-specific, but most are multidrug transporters that can excrete many structurally unrelated substances.^[17]

A multi-pathway-targeted strategy is required to create an effective therapy for HUS caused by E. coli O157: H7, rather than a single pathway-targeted approach. A better understanding of the aetiology and mechanisms of drug resistance is necessary to identify better targets and create novel treatments for HUS.^[18] Following a literature review, we found several prior findings on Vigna species focused mainly on antibiotic resistance.^[11] To our knowledge, no research demonstrates the antibacterial/antimicrobial mechanism of the Vigna species.

This study used the network pharmacology technique, which includes multi-compound, multi-target, and multi-pathway properties.^[19] The goal is to find possible bioactive chemicals and pharmacological pathways linked to antimicrobial action. This research might aid in discovering novel antimicrobial resistance mechanisms in E. coli O157: H7, which could assist in avoiding the bacterium’s pathogenicity.

MATERIAL AND METHODS

RESULTS

Targets of bioactive compounds

The 27 bioactive compounds were subjected to the STITCH database to retrieve the chemical-protein interactions. Using the Cytoscape 3.3.0 program, a bioactive compound-protein network has been created [Figure 1]. There were 207 nodes and 1637 edges in the network. The Network Analyser plug-in was used to determine the topological parameters. The mean degree of compounds per target was 2.83, and the average number of proteins per compound was 11.86. More than one protein from E. coli K-12 is linked to 13 active chemicals. This network reveals that Vigna species had promising therapeutic effects against E. coli O157:H7 targets [Table 2].

Close

Figure 1: Compound-target interaction network of Vigna bioactives and E. coli O157:H7. This network maps connections between bacterial proteins (green) and bioactive metabolites (orange), highlighting clusters in stress response (cheA, arcB), transcription (rpoB), and metabolism (folA). Integrated ligands like genistein and methotrexate reveal multi-target effects across chemotaxis and biosynthetic pathways. This systems-level view illustrates how natural inhibitors disrupt bacterial homeostasis and resistance mechanisms.

Export to PPT

Table 2: Network topology profile of core network and subnetwork of Vigna species bioactive compound-proteins of E. coli O157: H7

Network parameters	Core or giant network	Subnetwork
Number of nodes	207	72
Number of edges	1637	902
Clustering coefficient	0.460	0.575
Network diameter	6	3
Shortest Path	42642	5112
Characteristic path length	2.774	1.374
Average number of neighbours	15.816	25.056

After topological analysis, 7 bioactive compounds (indole-3-aldehyde, quinolinic acid, genistein, p-coumaric acid, daidzein, methotrexate, and myristic acid) represented the critical bioactive compounds against bacterial proteins like ArcA, ArcB, FolB, and RpoB. The chemical-protein interaction network of these seven bioactive compounds is illustrated in Supplementary Figure 1. The other top proteins identified included PyrB, FolC, RpoC, TonB, Gor, and MenB [Figure 2, Table 3].

Supplementary Figure 1

Download

Close

Figure 2: Protein-protein interaction (PPI) subnetwork of regulatory hubs. This network highlights central nodes prioritized by colour: red (arcA, folA) and orange (arcB, rpoB/C, folC, pyrB) identify bottlenecks linking transcription, metabolism, and stress response. Yellow nodes (tonB, menB, gor) represent specialized modules for iron transport and redox homeostasis. Dense connectivity demonstrates a coordinated systems-level architecture integrating environmental signalling to regulate bacterial survival and antibiotic resistance.

Export to PPT

Table 3: The proteins in the backbone network showing the top ten protein based on the degree and betweenness centrality

Sr. no.	Name of the gene/protein	Description	Betweenness centrality (BC)	Closeness centrality (CC)	Degree (D)
1	arcB	Aerobic respiration control sensor protein	0.26006365	0.97260874	69
2	evgS	Sensor protein	0.3338865	0.81609195	55
3	barA	Signal transduction histidine-protein kinase	0.3338865	0.81609195	55
4	torS	Sensor protein	0.0571368	0.80681818	54
5	dos	putative aldose-1-epimerase	0.01143957	0.71717172	43
6	gmr	Modulator of Rnase II stability	0.01143957	0.71717172	43
7	ydaM (dgcM)	Diguanylate cyclase	0.00625494	0.69607843	40
8	adrA (yaiC)	Diguanylate cyclase	0.05578961	0.66981132	36
9	yedQ (dgcQ)	Probable diguanylate cyclase	0.00367128	0.65740741	34
10	yegE	Putative transport system permease protein	0.00184221	0.61206897	26

GO annotation

A total of 182 E. coli O157:H7 genes/proteins interact with compounds from Vigna species for gene enrichment and pathway function analysis. The GO-BP and MF describe a set of events assembled by one or more MFs. There were a total of 99 GO-BP and 32 GO-MF keywords enhanced in this study (p-value<0.01 corrected with Bonferroni). Figures 3a-b show the top significant GO-BP and GO-MF phrases. The top GO-BP terms include protein dephosphorylation, chemotaxis, response to temperature stimulus, cobalamin transport, DNA unwinding included in DNA replication, deoxyribonucleotide metabolic process, monocarboxylic acid biosynthesis process, hydrogen sulphide biosynthetic process, phosphorous metabolic process, signal transduction system, nucleotide metabolic process, oxoacid metabolic process, response to oxygen-containing compound, single-organism biosynthesis process, small molecule metabolic process, DNA topoisomerase type II (ATP-hydrolysing activity), oxidoreductase activity (acting on the CH-NH group of donors, NAD or NADP as acceptor), ribonucleoside-diphosphate reductase activity, thioredoxin disulfide as acceptor, GTP binding, histidine phosphotransfer kinase activity, haeme binding, and lipid-transporting ATPase activity (acting on a sulphur group of donors). The bioactive compounds linked with proteins in E. coli were discovered through several biological processes participating in the same functional module by analysing GO-BP and MF keywords.

Close

Figure 3: Functional enrichment and GO distribution. Pie charts classify identified genes by (a) Biological process, showing dominance in phosphorus and nucleotide metabolism,

Export to PPT

Close

Figure 3: (b) Molecular function, highlighting coenzyme binding and specialized catalytic activities, and,

Export to PPT

KEGG pathway enrichment analysis

Seven significant KEGG pathways were obtained, including biotin metabolism, nicotinate and nicotinamide metabolism, pyrimidine metabolism, RNA polymerase, fatty acid biosynthesis, and bacterial chemotaxis [Figure 3c] [Table 4]. Multiple pathogeneses of E. coli O157: H7 include these critical signalling pathways and their related targets.

Close

Figure 3: (c) Pathway enrichment, featuring RNA polymerase and chemotaxis. These results provide a systems-level overview of metabolic priorities and regulatory mechanisms. GO: Gene Ontology, RNA: Ribonucleic acid.

Export to PPT

Table 4: KEGG pathway functional analysis

GO ID	GO Term	Nr. Genes	Associated genes
KEGG:00061	Fatty acid biosynthesis	5	accA, accB, accC, accD, fabA, fabB, fabD, fabF, fabG, fabH, fabI, fabZ, fadD
KEGG:00240	Pyrimidine metabolism	19	carA, carB, cdd, cmk, codA, cpdB, dcd, deoA, deoD, dnaE, dnaN, dnaQ, dnaX, dut, holA, holB, holC, holD, holE, hyuA, mazG, ndk, nrdA, nrdB, nrdD, nrdE, nrdF, pnp, polA, preA, preT, psuG, psuK, pyrB, pyrC, pyrD, pyrE, pyrF, pyrG, pyrH, pyrI, rihB, rpoA, rpoB, rpoC, rpoZ, rutA, rutB, rutC, rutD, rutE, rutF, tdk, thyA, tmk, trxB, udk, udp, umpG, upp, ushA, ydfG, yfbR, yjjG, yrfG
KEGG:00760	Nicotinate and nicotinamide metabolism	7	deoD, gabD, mazG, nadA, nadB, nadC, nadD, nadE, nadK, nadR, nudC, pncA, pncB, pncC, pntA, pntB, sad, sthA, umpG, ushA, yjjG, yrfG
KEGG:00780	Biotin metabolism	6	bioA, bioB, bioC, bioD, bioF, bioH, birA, bisC, fabB, fabF, fabG, fabI, fabZ, ynfK
KEGG:02020	Two-component system	34	acrD, aer, ampC, appY, arcA, arcB, arnB, atoA, atoB, atoC, atoD, atoE, atoS, baeR, baeS, barA, basR, basS, cbdA, cbdB, cheA, cheB, cheR, cheW, cheY, citA, citB, citC, citD, citE, citF, citG, citT, citX, cpxA, cpxR, creB, creC, crp, csrA, cusA, cusB, cusC, cusF, cusR, cusS, cydA, cydB, dctA, dcuB, dcuR, dcuS, ddpX, degP, dnaA, emrK, emrY, envZ, evgA, evgS, fdnG, fdnH, fdnI, fepA, flgM, flhC, flhD, fliA, fliC, frdA, frdB, frdC, frdD, glmY, glnA, glnB, glnD, glnG, glnL, glrK, glrR, gltI, gltJ, gltK, gltL, hyaC, iceT, kdpA, kdpB, kdpC, kdpD, kdpE, kdpF, maeA, mdtA, mdtB, mdtC, motA, narG, narH, narI, narJ, narL, narP, narQ, narV, narW, narX, narY, narZ, ompC, ompF, ompR, phoA, phoB, phoP, phoQ, phoR, pstS, qseB, qseC, rcsA, rcsB, rcsC, rcsD, rcsF, rpoN, rstA, rstB, sdiA, sfmZ, tap, tar, tolC, torA, torC, torD, torR, torS, trg, tsr, uhpA, uhpB, uhpC, uhpT, uvrY, yqeF, zraP, zraR, zraS
KEGG:02030	Bacterial chemotaxis	8	aer, cheA, cheB, cheR, cheW, cheY, cheZ, dppA, fliG, fliM, fliN, malE, mglB, motA, motB, rbsB, tap, tar, trg, tsr
KEGG:03020	RNA polymerase	4	rpoA, rpoB, rpoC, rpoZ

GO ID: g=Gene ontology identifier, GO: Gene ontology, Nr: Nuclear receptor, KEGG: Kyoto Encyclopedia of Genes and Genomes.

Homology modelling of target proteins

The tertiary structure determines its capacity to interact with other molecules or perform diverse tasks. Because the target proteins AcrA, AcrB, FolA, and RpoB do not have a crystal structure in the PDB, their 3D structures have been modelled using existing protein sequences for homology-based modelling. The homology modelling of these proteins was done using Phyre2, a web-based service. Phyre2 generates alignments of a provided protein sequence to proteins with published structures using the hidden Markov method. The homology modelling indicated that the single highest scoring template had high confidence in modelling AcrA, AcrB, FolA, and RpoB. [Figures 4a-d and Table 5]. ModRefiner and DeepView v4.04 were used to refine the structure and minimise the energy of the projected models of all four proteins. The AcrA, AcrB, FolA, and RpoB overhaul geometries were corrected and warped by energy reduction.

Close

Figure 4: Homology modelling of key resistance-related proteins. These 100% confidence models illustrate (a) AcrA, an elongated periplasmic bridge, (b) AcrB, a massive transmembrane multidrug efflux pump, (c) FolA, a globular enzyme with a central β-sheet and inhibitor target, and (d) RpoB, the complex β-subunit of RNA polymerase’s catalytic core. Together, these structures provide a basis for analysing protein-ligand interactions and mechanisms of antibiotic resistance.

Export to PPT

Table 5a: Templates used in homology modelling of acrA, acrB, folA and rpoB using Phrey2 server with percentage of confidence and identity

Protein name	Template used (Phrey2 fold library ID)	Name of template	Percentage of confidence	Percentage of identity
acrA	d5v5sH_	Multidrug efflux pump subunit acra	100	100
acrB	d2v50A	Multidrug resistance protein mexb	100	55
folA	d1ra9a	Dihydrifolate reductase-like	100	99
rpoB	c3IuOC	DNA-directed RNA polymerase subunit Beta	100	100

Table 5b: Molecular docking scores for potential active biological compounds with protein targets

Target protein	Biological compounds	Docking score (Kcal/mol)
acrA	Indole-3-aldehyde	-5.981
	Quinolinic acid	-5.68
	Genistein	-5.313
	p-Coumaric acid	-5.195
	Daidzein	-5.08
	Methotrexate	-4.582
	Myristic acid	0.189
acrB	Daidzein	-6.152
	Methotrexate	-5.801
	Genistein	-5.628
	Quinolinic acid	-5.088
	Methotrexate	-4.994
	Indole-3-aldehyde	-4.226
	p-Coumaric acid	-3.766
	Myristic acid	1.973
folA final	Genistein	-6.653
	Methotrexate	-6.548
	Daidzein	-6.323
	Methotrexate	-5.608
	Indole-3-aldehyde	-5.325
	Quinolinic acid	-4.461
	p-Coumaric acid	-3.318
	Myristic acid	-0.004
	Daidzein	-5.636
	Genistein	-5.613
	Indole-3-aldehyde	-5.437
rpoB	p-Coumaric acid	-4.909
	Quinolinic acid	-4.698
	Methotrexate	-4.343
	Methotrexate	-3.764
	Myristic acid	0.484

In protein structural prediction, model validation is critical since the modelled protein structure is eventually utilised to plan subsequent tests and comprehend the protein’s biological function. The PROCHECK module of SAVES was used to calculate the Ramachandran plot for all four protein models. Only the FolA protein model indicated that no residues were found in forbidden locations and 90.5 % of residues were found in the most favourable regions, indicating that the 3D model’s quality was expected to be highly significant [Figure 5a]. In comparison to the other three proteins, AcrA, AcrB, and RpoB, which have 45.31%, 70.18%, and 60.08%, respectively, ERRAT analysis revealed that the overall structural quality of the predicted structure of FolA is 94.66, which is very good experimentally and computationally [Figure 5b]. The VERIFY 3D study revealed that 100% of FolA protein residues have a 3D-1D score of > 0.2, indicating that the primary sequence and tertiary structure are compatible [Figure 5c].

Close

Figure 5: Stereochemical and environmental validation of FolA Model. Structural quality was confirmed via (a) Ramachandran plot, showing high-quality backbone geometry with residues in favoured/allowed regions, (b) ERRAT plot, verifying reliable local environments with scores below the 95% and 99% error thresholds, and (c) Verify3D plot, where scores above 0.2 confirm correct 3D-1D sequence compatibility and folding. These results validate the model as a robust template for computational drug design

Export to PPT

The structural quality of FolA is assessed using ProSA-web, and its ProSA Z-score value of 8.39 fits within the range of native conformations determined using the X-ray crystallography technique, which is depicted as an encircling huge black dot [Figure 6a]. The results show that most residues in the predicted protein FolA lie in the negative energy minimum area, indicating solid structural quality and low energy levels [Figure 6b]. As a result of the different techniques, the anticipated folA 3D structure is of extremely high quality and stability.

Close

Figure 6: The structural integrity and reliability of the modelled protein were validated using ProSA-web and standard stereochemical assessment tools. (a) The Z-score plot displays the overall model quality (represented by the black dot within the red circle) falling within the range of scores typically found for native proteins of similar size, as determined by X-ray crystallography (light blue) and NMR (dark blue), (b) The energy plot illustrates the local model quality across the sequence position, with both window sizes (10 and 40 residues) remaining largely below the zero-energy threshold, indicating favourable folding energetics. ProSA: Protein Structure Analysis tool, NMR: Nuclear magnetic resonance.

Export to PPT

The QMEAN6 server derives a quality estimate based on the geometrical analysis of single models. The clustering-based scoring function QMEANclust calculates a global and local quality estimate based on a weighted all-against-all comparison of the models from the ensemble provided. The global quality of models is assessed using six structural descriptors. They are (a) a torsion energy potential based on three successive amino acids used to test local geometry, (b) two distance-dependent potentials based on C atoms and all atoms used to assess long-range interactions, (c) a solvent potential, and (d) solvent accessibility. The QMEAN Z-score calculates the absolute quality by comparing it to similar-sized reference structures in the PDB that were solved using experimental methods (Srivastava et al., 2008). The ‘degree of nativeness’ of the predicted structure was calculated using the QMEAN Z-score. QMEAN server results employing energy-minimised AcrA, AcrB, FolA, and RpoB 3D structures. The AcrA, AcrB, FolA, and RpoB models had overall QMEAN scores of 0.71, 0.64, 0.87, and 0.64, respectively. The model of folA acquired from the Phyre2 server has superior C-beta interaction energy, all-atom pairwise energy, solvation energy, torsion angle energy, and secondary structure agreement [Supplementary Table 2]. As a result, the model of FolA derived from the Phyre2 server was superior to the models generated for the other three protein targets (ArcA, ArcB, and RpoB).

Supplementary Table 2

Download

Molecular docking

Four targets were selected based on phytochemical-target interaction network analysis and pathway enrichment analysis for the seven active ingredients screened according to ADME parameters. The generated active components (ligands) and selected targets were then subjected to molecular docking studies. After the ingredient-target interaction validation, one component returned probable successful docking scores [Table 5].

The results indicated that these active ingredients showed a potential binding affinity with the selected targets. In general, a negative score indicates an excellent binding activity. The ingredients indole-3-aldehyde, quinolinic acid, and genistein showed docking scores of -5.981, -5.68, and -5.313, respectively, against the AcrA target [Supplementary Figure 2a]. Likewise, the ingredients daidzein, methotrexate, and genistein showed docking scores of -6.152, -5.801, and -5.628, respectively, against the AcrB target [Supplementary Figure 2b]. Furthermore, the ingredients daidzein, genistein, and indole-3-aldehyde exhibited docking scores of -5.636, -5.613, and -5.437, respectively, against the RpoB target [Supplementary Figure 2c]. Out of four targets, FolA exhibited good binding affinity towards the three phytochemicals genistein, methotrexate, and daidzein, with docking scores -6.653, -6.548, and -6.323, respectively [Figures 7a-7c].

Supplementary Figure 2

Download

Close

Figure 7: Molecular docking analysis of ligands with target protein. 3D poses and 2D maps show (a) Genistein stabilized by hydrogen bonds and hydrophobic contacts, (b) Methotrexate forming complex polar interactions via its pteridine ring, and (c) Daidzein utilizing distinct residues for stabilization. Panels progress from 3D ribbon-and-stick views (left) to solvent-accessible surfaces (middle) and 2D interaction schematics (right), detailing specific amino acid contacts crucial for binding affinity. 3D: Three-dimensional, 2D: Two-dimensional.

Export to PPT

The best hit target protein FolA was used for redocking with three phytochemicals, genistein, methotrexate, and daidzein, with site-specific and blind docking approaches to ensure that ligands bind to the same active site and interact with the same amino acid residues as before. Here, both methods provided valuable information about the interaction of the three ligands with the target protein FolA. Table 6 illustrates the binding affinity and amino acid residues involved in the formation of the hydrogen bonds and other bonds with the three compounds used in the study. It is observed that the residues Thr46, Ala7, Phe31, and Asp27 were observed to have good interaction with daidzein and genistein in both site-specific and blinded docking [Supplementary Figures 3a-b]. However, we observed that methotrexate interacts with FolA by forming hydrogen bonds with Asp27 and Ile5 in site-specific docking. In contrast, in blinded docking, it forms hydrogen bonds with Ile94, Ser49, Asn18, and Asp122 [Supplementary Figures 3a-b]. The overall molecular docking experiments and ADMET analysis of the three protein-ligand complexes, suggested that genistein is a potential compound against the identified therapeutic target FolA in antibiotic-resistant E. coli O157: H7 strain.

Supplementary Figure 3

Download

Table 6: The binding affinity and amino acid residues in folA involved in the formation of the hydrogen bonds and other bonds with the three compounds genistein, methotrexate and daidzein observed by redocking experiment (site-specific and blinded docking).

Target protein	Compound name	Site-specific docking			Other interactions
		Binding free energy (kJ/mol)		Hydrogen bonds
folA	Daidzein	-7.62		Thr46, Ala7, Phe31, Asp27	Gky97, Met20, Trp22, Ala6, Thr113, Trp30, Phe153, Leu28, Ile5, Ile14, Ile94, Gly96, Gly96
	Genistein	-7.63		Thr46, Ala7, Phe31, Asp27	Leu28, Ile5, Ile94, Gly95, Gly96, Gly97, Ile14, Tyr100, Met20, Trp22, Ala6, Thr113, Trp30, Phe153
	Methotrexate	-4.00		Asp27, Ile5	Leu28, Met20, Ala6, Trp22, Ala7, Trp30, Thr113, Phe31, Gly96, Thr100, Ile14, Ile94, Gly95, Ser49, Thr46, Ile50, Leu54, Pro23
	Blind docking
	Daidzein	-8.46	Thr46, Ala7, Phe31, Asp27		Gly97, Gly96, Met20, Ala6, Ile5, Trp30, Thr113, Phe153, Leu28, Ile94, Ile14, Gly95, Tyr100
	Genistein	-8.28	Asp27, Ala7, Tyr100, Gly96		Phe15, Ile5, Phe125, Ala6, Gly95, Gly96, Ser49, Ile94, Leu28, Trp30, Phe31, Thr113
	Methotrexate	-9.16	Ile94, Ser49, Asp18, Asp122		Ile5, Lie50, Gly15, Thr123, Ala19, Met16, Gly17, Gly97, Gly96, Thr46, Ile14, Gly95, Thr100, Phe31

DISCUSSION

The network pharmacological analysis of the bioactive compounds in Vigna species has identified potential molecules with antimicrobial properties, even on antibiotic-resistant E. coli O157: H7 strain. Preliminary topological analysis of the bioactive compounds-candidate target network revealed about 27 compounds. Further, GO and KEGG enrichment analysis of the candidate targets reported key signalling pathways like a two-component system; nicotinate and nicotinamide metabolism; phenylalanine, tyrosine, and tryptophan biosynthesis; starch and sucrose metabolism.

The functional enrichment analysis of all the interactions was carried out using the ClueGO tool [Supplementary Table 3]. This leads to the identification of genes in the network playing a significant role in biological processes like signal transduction (GO:0007165), single organism signalling (GO:0044700), phosphorelay signal transduction system (GO:0000160), cell communication (GO:0007154), and cellular response to the stimulus (GO:0051716). Molecular functions such as molecular transducer activity (GO:0060089), signal transducer activity (GO:0004871), receptor signalling protein activity (GO:0005057), receptor activity (GO:0004872), catalytic activity (GO:0003824), molecular function (GO:0003674), transferase activity (GO:001674), and small-molecule binding (GO:0036094). Enriched KEGG pathways include a two-component system, biotin metabolism, microbial metabolism in a diverse environment, and fatty acid metabolism and biosynthesis.

The genes responsible for fatty acid biosynthesis, a two-component system, and biotin metabolism were identified by GO enrichment. The genes fabD, fabF, folA, fabG, and fabH are associated with fatty acid biosynthesis, and the genes accA, accB, accC, accD, fabA, fabB, fabL, and fabZ with fatty acid synthesis. The genes bioA, bioB, bioC, bioD, bioF, bioH, birA, bisC, fabB, fabF, fabG, fabL, fabZ, and ynfK are found to be associated with biotin metabolism. Biotin synthesis in E. coli O157:H7 is tightly regulated in response to biotin supply/demand by the biotin protein ligase BirA, which serves as both an enzyme that funnels biotin into metabolism and a negative transcriptional regulator of the biotin synthesis operon (bio-operon), controlling bioA and bioBFCD.^[45]

Further, the genes cpxA, cpxR, rcsB, rcsC, acrD, aer, ampC, appY, arcA, arcB, arnB, atoA, barA, basR, cheA, cheB, cheR, cusB, cusC, cusF, cusR, and others are involved in the two-component system.^[46] The bacteria, including E. coli, can sense, respond, and adapt to changes in their environment or intracellular conditions thanks to a two-component signal transduction system. The two-component systems are the sole regulatory components shared by a wide spectrum of virulence systems.^[47] Many virulence factors are necessary for bacteria to thrive in a foreign host, and two-component systems control ‘essential’ genes in some bacteria.^[48,49] The CpxR-CpxA and RcsB-RcsC are a couple of two-component systems in E. coli involved in virulence mechanisms. Capsular polysaccharides (K-antigen) aid adhesion in homologous activities. It has also been reported that these factors, which encode for regulating or facilitating the transfer of antibacterial or antibiotic resistance, may also contribute to the bacterium’s shift to the pathogenic state. Thus, bacterial virulence factors contribute to the bacterium’s survival and growth at the infection site. Various compounds will be synthesised by the plant, which belong to various chemicals. The flavonoids synthesised in plants in response to microbial infection possess potential antimicrobial activity.

Furthermore, these molecules can reverse antibiotic resistance while improving the effectiveness of already available antibiotics. As a result, developing and using flavonoid-based medicines to treat antibiotic-resistant illnesses might be a potential strategy. In animals and humans, bacterial biofilm-based infections play a crucial role in all infectious and chronic infections.^[50] Biofilm formation makes the bacteria more resistant to antimicrobial agents.^[51] In E. coli, quorum sensing, namely autoinducer-2-mediated cell-cell communication, has been postulated as a key regulator of biofilm formation.^[52] Hydrophilic flavonoids can interact with the membrane surface, protecting it against harmful substances and biofilm development.^[53] Flavonoids found as secondary metabolites from plants, such as daidzein and genistein, suppress the production of E. coli O157: H7 biofilms.^[54] The antibiotics carbenicillin and levofloxacin were potentiated by daidzein against E. coli, presumably circumventing the efflux resistance mechanism.^[55] The replicative helicases DnaB and RecBCD helicase/nuclease of E. coli were inhibited by myristic acid/myristin.^[56] Daidzein has been postulated as a possible inhibitor of various DNA and RNA polymerases, viral reverse transcriptases, and telomerase.^[57] The proteins involved in the two-component system were found to be interacting with genistein (gyrA, phoQ, torS, arcB) and other proteins, HiuH, Z1444, YadH, Z1629, CysJ, HisS, CarB, VarA, PyrB, and GyrB are involved in the various processes, including response to antibiotics [Supplementary Table 3], are found to be first neighbours of the network obtained for genistin. Furthermmore, for the compound indole-3-phosphate, the interacting neighbour proteins were KatE (catalase), CheY (chemotaxis response regulator), EvgS (sensor protein), and BarA (signal transduction histidine-protein kinase). The gene evgS is a component of efflux pumps and two-component systems like EvgAS, PhoPQ, and BaeSR, and increases the expression of emrKY, yhiUV, acrAB, mdfA, and tolC.^[58] Further, the first neighbours of the p-coumaric acid also found some proteins involved in two-component systems (ECs_0417, RcsC, NarQ, and ZraS) and other proteins with various functions, like an expression of master biofilm regulator CsyD [Supplementary Table 3].

The network topology analysis of the chemical-protein interaction network based on degree, betweenness centrality and closeness centrality revealed that seven compounds (myristic acid, methotrexate, quinolinic acid, daidzein, phaseolin, genistein, indole-3-aldehyde, and p-coumaric acid) have a good interaction with four proteins (AcrA, AcrB, FolA, and RpoA) identified as potential therapeutic targets. In E. coli, when components of the efflux pump (e.g. AcrB) are deleted or inactivated, there is feedback that increases acrA/acrB operon expression via regulators like MarA and SoxS.^[59] In Escherichia coli, deletion of mgrB led to upregulation of folA, alongside phoP/phoQ. This increased expression of FolA conferred increased resistance to TMP (trimethoprim).^[60] Some evolution/adaptation studies find that rpoB (or more broadly RNAP genes) are upregulated in certain mutants/stress environments. For example, in double mutants in E. coli, components of RNAP, including rpoB, were sometimes found upregulated.^[61]

Further, the molecular docking studies of all four targets with all seven phyto-ingredients revealed that only FolA protein shows better interaction with the phytochemicals methotrexate, genistein, and daidzein compared to the other targets identified [Supplementary Figures 2a-c]. Thus, better molecular interaction was observed for genistein against folA with a docking energy of -6.653 kcal/mol compared to methotrexate (-6.548 kcal/mol) and daidzein (-6.323 kcal/mol). Here, the residues of the binding pocket in FolA show interaction with the genistein. FolA is a DNA synthesis and cell maintenance enzyme associated with novo glycine and purine synthesis.^[62] This is especially critical during infancy and pregnancy when cells divide and develop rapidly. To replicate DNA, FolA is required. As a result, a lack of FolA disrupts DNA synthesis and cell division, which is a location of fast cell turnover.^[62] Trimethoprim was used to treat various E. coli infections. However, the rapid growth of TMP-resistant bacteria limits its effectiveness. TMP is designed to inhibit the bacterial enzyme dihydrofolate reductase (DHFR), expressed by the folA gene. However, recent evolutionary studies have discovered resistance-inducing mutations in the folA gene, specifically mutation L28R.^[63] Due to the rapid spread of resistant E. coli, trimethoprim, a popular therapy for urinary tract infections, is becoming outdated. Although direct resistance mechanisms such as overexpression of a mutant folA and dfr enzymes have been widely studied, the accompanying changes that induce or sustain resistance are unclear.

DHFR is an important enzyme in the de novo production of purine and thymidine.^[64] Antibiotics and anticancer drugs are effective when small molecules targeting this enzyme are used.^[65] However, this enzyme develops quick resistance to available antifolates by gaining mutations on binding residues. Clinical-level resistance to known antifolates can be established after only three rounds of directed-evolution attempts.^[66] Attempts to decipher the evolutionary processes for antibiotic resistance development in DHFR revealed that resistance develops through the progressive fixation of mutations via organised routes. The most common trimethoprim-resistant mutations were on the promoter (9G>A; 35C>T) or on the DHFR protein (P21, A26, L28R, W30, and I94), as well as their combinations.^[67] Physical-chemical tests on these mutants have also revealed that the reduced drug affinity comes at the expense of catalytic efficiency and protein stability. Several types of chemicals have been investigated for their possible anti-folate action to address the problem of fast drug resistance acquisition. The folA gene encodes the primary DHFR activity in the tetrahydrofolate production pathway. Transfering hydride from NADPH to the pteridine ring C6 catalyses the conversion of dihydrofolate to tetrahydrofolate. Tetrahydrofolate is crucial for protein and nucleic acid production. DHFR is a therapeutic development target since it is required for cell division and proliferation. Several anticancer, antibiotic, and antimalarial medicines, including methotrexate and trimethoprim, can inhibit.^[68,69] DHFR has also been employed to investigate drug resistance mechanisms^[67] and assess mutations’ evolutionary impact on protein homeostasis.^[70,71] All cells have the DHFR, which is responsible for keeping intracellular folate stores in a biochemically active reduced form. Intracellular reduced folates, which are required for one-carbon transfer processes, are depleted due to inhibition. Thymidylate, purine nucleotides, methionine, serine, glycine, and many other chemicals required for RNA, DNA, and protein synthesis need one-carbon transfer processes. As a result, DHFR is a promising target for developing novel antibacterial and anticancer drugs.

Genistein (C15H10O5) belongs to the multifunctional natural flavonoid class of flavonoids with a 15-carbon skeleton [4′, 5, 7-trihydroxyisoflavone or 5, 7-dihydroxy-3-(4-hydroxyphenyl) chromen-4-one]. Genistein has a molecular structure that is comparable to estradiol.^[72] One of the main goals of quantitative structure-activity relationships (QSAR) is to gather the information that may be used to make more active or less hazardous chemicals. Before their creation, QSAR correctly predicted a substantial number of chemicals.^[73] Taking all aspects from the literature survey and the above network pharmacology results, we studied the binding affinity of genistein to DHFR (folA) to develop a potent DHFR inhibitor. Genistein is a 7-hydroxyisoflavone with other hydroxy groups at positions 5’ and 4’. In our study, these two hydroxy groups are found to form hydrogen bonds with amino acid residues ILE14 and SER49, and the other hydroxy group forms a hydrogen bond with ASP24 of the DHFR, with an overall docking score of -6.653 kcal/mol, suggesting a stable interaction between ligand and target protein.

In summary, we attempted to reveal the interaction of the Vigna species compounds with E. coli O157:H7, which causes common diarrhoea to HUS in humans. We observed that the compounds, methotrexate, daidzein, and genistein show interaction with protein target FolA, which participates in two-component systems, biofilm metabolism, fatty acid synthesis, and others involved in developing antibiotic resistance by the pathogen [Table 7]. However, certain limitations, like data collection from various databases for bioactive compounds, targets, and screening of bioactive compounds based on ADME parameters, may be more comprehensive. Therefore, in vivo and in vitro experiments would validate in silico analysis of these active compounds against antibiotic-resistant E. coli O157: H7.

CONCLUSION

The systems pharmacology approach was employed to uncover the relation between Vigna species bioactive molecules and target proteins of E. coli O157: H7 imparting antibiotic resistance in the bacterium. Homology modelling and validating 3D structure for the four proteins resulted in only folA protein with an acceptable structure. The analysis identified compounds like methotrexate, daidzein, and genistein, demonstrating potent interaction with protein target FolA. The bioactive compound genistein exhibited good potency of interaction with the amino acid residues of FolA, demonstrating its inhibitory activity against the DHFR. Several compounds have been explored for their potential anti-folate activity to address the concern of rapid drug-resistance acquisition. The folA encodes the main DHFR activity in the tetrahydrofolate biosynthesis pathway. Hence, further experimental analysis may give insights into the genistein’s mechanism of action against folA to prevent antibacterial resistance.