In silico screening and in vitro validation of phytocompounds as multidrug efflux pump inhibitor against E. coli

Abstract Multiple drug resistance (MDR) in bacteria has increased globally in recent times. This has reduced the efficacy of antibiotics and increasing the rate of therapeutic failure. Targeting efflux pump by natural and synthetic compounds is one of the strategies to develop an ideal broad-spectrum resistance-modifying agent. Very few inhibitors of AcrB from natural sources have been reported till date. In the current study, 19 phytocompounds were screened for efflux pump inhibitory activity against AcrB protein of E. coli TG1 using molecular docking studies. The molecular dynamics simulation provided stability the protein (AcrB) and its complex with chlorogenic acid under physiological conditions. Moreover, the detailed molecular insights of the binding were also explored. The Lipinski rule of 5 and the drug-likeness prediction was determined using Swiss ADME server, while toxicity prediction was done using admetSAR and PROTOX-II webservers. Chlorogenic acid showed the highest binding affinity (−9.1 kcal mol−1) with AcrB protein among all screened phytocompounds. Consequently, all the phytocompounds that accede to Lipinski’s rule, demonstrated a high LD50 value indicating that they are non-toxic except the phytocompound reserpine. Chlorogenic acid and capsaicin are filtered out based on the synergy with tetracycline having FIC index of 0.25 and 0.28. The percentage increase of EtBr fluorescence by chlorogenic acid was 36.6% followed by piperine (24.2%). Chlorogenic acid may be a promising efflux pump inhibitor that might be employed in combination therapy with tetracycline against E. coli, based on the above relationship between in silico screening and in vitro positive efflux inhibitory activity. Communicated by Ramaswamy H. Sarma.


Introduction
Antimicrobial resistance (AMR) is becoming more prevalent in the community and nosocomial environments, posing a serious alarm to human health and accounting for a significant number of deaths and morbidities around the world (Founou et al., 2017;Naylor et al., 2018). The synthesis of b-lactamases enzymes (e.g. extended-spectrum b-lactamases [ESBLs], carbapenemases and metallo b-lactamases) has increased in the last two decades, contributing resistance to third-generation cephalosporin and carbapenem antibiotics (Blair et al., 2015;Hsu et al., 2021;Martinez et al., 2019). The drug resistance mechanisms has been classified into three categories: mutation of the antibiotic target site, modification of the enzyme and enhancement of the efflux pump's activity (Christaki et al., 2020). Bacterial efflux pumps are huge weapons of drug-resistant pathogens, and they raise and sustain AMR by extruding or reducing intracellular antibiotic concentrations, often in a non-specific manner (Chitsaz & Brown, 2017). Efflux pumps (EP) have emerged as key drivers of AMR in Gram-positive including Gram-negative bacterial species, making them potent and universal targets for eradicating drug-resistant phenotypes (Neuberger et al., 2018). The physiological processes that EPs play a role include stress adaptation, transportation of important nutrients, virulence and the pathogenicity (Costa et al., 2013;Fern andez-Cuenca et al., 2013;Holmes et al., 2016). Another important physiological function endorsed to EPs is in the formation of pathogenic biofilms (Alav et al., 2018). Despite the severe threats posed by Gram-negative bacteria and their drugresistant nature, more research aimed at targeting them with new, alternative and successful methods including natural product exploration. Efflux pump inhibitors are substances that can minimise resistance or completely reverse AMR to already ineffective antibiotics by inhibiting EPs. According to the World Health Organization (WHO), conventional herbal medicines are used by 85-90% of the world's (World Health Organization, 2017).
Medicinal plants have been recognised as a source of phytoconstituents capable of generating alternative and effective drug leads to combat AMR by focussing on the main determinants of drug resistance, such as EPs (Prasch & Bucar, 2015). Efflux pump inhibiting capabilities have been tested on a variety of synthetic and natural compounds (Ciulla & Kumar, 2018;Shin et al., 2018;Spengler et al., 2017). Finding new and effective EP inhibitors (EPIs) as drug lead to reverse AMR is thus gaining importance. The inner membrane protein AcrB, from the AcrAB-TolC drug transport complex of Escherichia coli, is an excellent model system in evaluating compounds having EPI characteristics in this regard. AcrB homologues can also be found in other Gramnegative pathogens, that also replace MexB efflux pump from Pseudomonas aeruginosa making an active drug complex MexB's natural partners MexA and OprM form an efflux complex (Dreier & Ruggerone, 2015). As a result, targeting AcrB to minimise extent of drug resistance and selectively inhibit the bacterial pathogenicity is considered as a novel strategy in the management of antimicrobial resistance. Based on the importance of this membrane protein for targeting drugs, we have selected this protein as target one.
In the identification of the rational design of drugs, molecular docking is the most widely used approaches. It has immense degree of accuracy for small molecule in the form of ligands, predicting the conclusive evidence within the required target binding site (Pal et al., 2019). The Molecular Dynamic simulation method, in combination through molecular docking and other in silico approaches, has been used to successfully screen and predict interaction between possible EPs and their inhibitors at molecular level (Jamshidi et al., 2016). Many other recent publications have described the effective use of these in silico approaches for screening and highlighting potent EPIs derived from plant sources (Jamshidi et al., 2016;Verma et al., 2021). However, despite of having great diversity of bioactive phytocompounds, and lack of concerted efforts, systematic screening, and evaluation of efflux pump inhibitors for therapeutic application is still required.

Experimental approach
The selected phytocompounds were assessed Lipinski rule, drug-likeness and toxicity prediction. The phytocompounds were screened through virtual screening and then shortlisted based on their binding energies for molecular dynamics simulation. Finally, E. coli TG1 used to test best docked compounds and experimentally validate as potent efflux pump inhibitor.

Bioinformatic tools
The bioinformatics tools that were used in the current study includes AutoDock Vina (Trott & Olson, 2009), Chimera 1.8.1 (Pettersen et al., 2004), Biovia discovery studio 4.2, Ligplot þ (Laskowski & Swindells, 2011) for docking analysis while Lipinski's rule of five determined by SwissADME (http://www. swissadme.ch/). The Lipinski's rule of five or Pfizer's rule of five, is a commonly used filter for predicting and evaluating the drug likeness of molecules. The admetSAR (http://lmmd. ecust.edu.cn/admetsar1/predict) for ADMET, possibly the first step to determine the suitability of novel compounds as drugs and to avoid wasting resources and PROTOX-II (https:// tox-new.charite.de/protox_II/) was used for prediction of toxicity of drug like molecule.

Molecular docking studies
The AcrB protein E. coli (Protein Data Bank: 4DX5, close monomer) was used to dock with selected phytocompounds using AutoDock 4.2 (http://autodock.scripps.edu) (Morris et al., 2009). The 3D crystal structures of protein AcrB was downloaded in .pdb format from Protein data bank (https:// www.rcsb.org/structure/4DX5). The bound ligand (minocycline) was removed from AcrB protein before docking, and the adequacy of AutoDock was validated using blind cognate docking (Het enyi & van der Spoel, 2011). The tertiary structures of phytocompounds i.e. Benzoic acid ( gov) in .sdf format and converted into pdb file using Chimera 1.10.2. The pdbqt files of both receptor protein and ligand were created by using MGL Tools-1.5.6 (Morris et al., 2009). The receptor protein and ligand residues were prepared, and docking analyses were run using virtual screening software, AutoDock Vina (Trott & Olson, 2009). From the binding energies of ligand interaction (kcal mol À1 ), inhibition constants (Ki) were calculated (Musa et al., 2011). LigPlot þ was employed to observe hydrophobic interactions and the number of hydrogen bonds between the protein-ligand complexes (Laskowski & Swindells, 2011). Figures were created using the graphical visualizer Biovia Discovery Studio.

Molecular dynamics simulations
Chlorogenic acid exhibited highest binding affinity towards multidrug transporter Acriflavine resistance protein B (AcrB). Therefore, this complex was selected for further computational studies using molecular dynamics (MD) simulations. The MD simulations were performed using Gromacs-2018.1 packages with amber99sb-ILDN force field (Berendsen et al., 1995;Maier et al., 2015). The topology chlorogenic acid was generated using antechamber packages in AmberTools21 (Sousa Da Silva & Vranken, 2012). AcrB alone and AcrBchlorogenic acid were solvated using TIP3P water model in triclinic boxes separately. Both structures were neutralised by adding equal number of counter ions. The systems were minimised using the steepest descent minimisation for 5000 steps to eliminate weak Van der Waals contacts. Both systems were subjected to NVT and NPT euqolibarion for 1 ns each using V-rescale thermostat at 300 K and Parrinello-Rahman barostat at 1.0 b, respectively (Bussi et al., 2007;Parrinello & Rahman, 1981). The MD simulation was run for 100 ns trajectories were saved at 10 ps intervals. Periodic boundary conditions (PBS) corrections were made before the analysis trajectories and all calculations were performed using Gromacs utilities. The binding energies between chlorogenic acid and AcrB were calculated using MM-PBSA calculation (Kumari et al., 2014).

Drug likeness, ADMET screening and toxicity prediction of phytocompounds
Drug scans were conducted to see whether any phytochemicals met the drug-likeness requirements or not. Lipinski's rule of five was used to determine drug-likeness characteristics such as the number of hydrogen acceptors (should not be more than 10), the number of hydrogen donors (should not be more than 5), molecular weight (mass should not be more than 500 Daltons) and partition coefficient log P (should be less than 5) using SwissADME (http://www.swissadme.ch/). The absorption, toxicity and drug-like properties of phytocompounds were determined through ADMET screening. 2.6. In vitro validation assay 2.6.1. Chemicals, bacterial strain and growth media The antibiotics tetracycline, nalidixic acid (purity ! 98%), chlorpromazine and 2-hydroxy 1,4 naphthoquinone were purchased from Sigma-Aldrich, India while the rest other phytocompounds chlorogenic acid (5-O-caffeoylquinic acid), piperine, reserpine, thymol and ethidium bromide was procured from Hi-Media. The bacterial strain E. coli TG1 wild type strain harbouring AcrAB-TolC efflux transporter, capable of constitutively expressing AcrB pump responsible for variety of substrate extrusion and later stages of biofilm development, was kindly provided as a gift by Prof. Kunihiko Nishino, Institute of Scientific and Industrial Research, Osaka University, Japan. The bacterial strain was maintained on Luria-Bertani (0.5% yeast extract, 15.0 g tryptone, 0.5% NaCl, 20 g/L Agar) medium at 37 C. The MIC and checkerboard assays were conducted in Cation Adjusted Mueller Hinton Broth (300 g/L beef infusion, 17.5 g/L casein acid hydrolysate, 1.5 g/L starch, 20 g/L Agar) procured from Hi-media.

Determination of minimum inhibitory concentration (MIC)
The microplate dilution assay was employed to assess the minimum inhibitory concentration (MIC) of the antibiotic tetracycline, nalidixic acid and selected phytocompounds as described by the Clinical and Laboratory Standards Institute guidelines (Clinical and Laboratory Standards Institute, 2017).
In the current study, ceftazidime antibiotic was used as positive control for MIC determination. Briefly, the bacterial culture was diluted to a turbidity level of 0.5 McFarland standard in PBS buffer. In each well of the 96-well microtiter plate, 0.15 mL of compound at concentrations prepared at 2fold serial dilutions in MH broth medium was supplied, and aliquots of 0.05 mL were transferred to each well. After 18 hours of incubation at 37 C, the MIC results were recorded. The MIC was identified as the lowest concentration of a drug at which no visible growth of the bacteria was observed.

Fluorescence-based ethidium bromide (EtBr) accumulation assay
The propensity of the bacterial strains to absorb ethidium bromide was investigated with and without the test phytochemicals (Coldham et al., 2010) with slight modifications. The overnight grown bacterial culture was inoculated in LB broth and incubated at 37 C for about 4 hours. The bacterial cells were centrifuged for 10 minutes at 10,000 rpm, then resuspended in phosphate-buffered saline (PBS) with an OD of 0.1 at 600 nm. Each tube received 2 mL of bacterial culture, chlorpromazine was used as a positive control, and 10 ml of ethidium bromide was applied below its sub-MIC concentration. Before the fluorescence was measured, the test compound was added at their respective sub-inhibitory concentration. The tubes were incubated at 37 C in dark for 30 minutes with a 5 min interval for each cycle and the fluorescence of the solution was recorded at excitation and emission wavelength of 530 and 590 nm, respectively using RF-5301PC spectrofluorometer (Shimadzu, Japan). The loss of ethidium bromide from the cells was suggested by a decrease in fluorescence as compared to the control tube without any test compound. The percentage increase in fluorescence was used to determine the ability of the compound to accumulate ethidium bromide.

Checkerboard synergy assay
In a 96-well microtiter plate, the checkerboard synergy titration assay of antibiotics in combination with the best-docked compounds was performed (Maheshwari et al., 2016). All the phytocompounds were dissolved in DMSO and diluted at sub-inhibitory concentration (MIC/4) in MHB medium. The concentration of phytocompound tested was remained the same throughout the experiment, whereas the antibiotic was serially diluted to assess a decrease in MIC in presence of the test compound. The bacterial inoculum was adjusted to a cell density of 10 6 CFU/mL, and 50 ml of inoculum was added to each well. The following formula was used to measure the fractional inhibitory concentration (FIC) index of the compound: The FICI value less than 0.5 was considered synergistic, FICI between 0.5 and 4 was additive/indifferent, and FICI greater than 4 was considered antagonistic. Further, Combenefit HSA plot was used to predict the synergy using OD 600 values of most active combination (https://sourceforge.net/projects/combenefit/).

Drug likeness, ADMET screening and toxicity prediction
The early preclinical analysis is supported by drug-likeness filters, which help to prevent expensive late-stage preclinical and clinical failure. The Lipinski rule of 5 was used to analyse the drug-likeness properties of molecules. Except for reserpine, all of the selected phytocompounds followed Lipinski's rule of five (Supplementary material, Table S1). The PROTOX-II server was used to determine the toxicity of phytocompounds. The effects of the admetSAR study and toxicity prediction were in Table S2 (Supplementary material). ADMET profiles of all the phytochemicals tested were within a reasonable range, indicating their efficacy as potent drug candidates. Benzoic acid, chlorogenic acid, and cinnamic acid had a certain level of hepatotoxicity. Based on ADMET properties, all the phytocompounds screened showed good human intestinal absorption as well as penetration to blood-brain barrier (BBB). All the compounds except the toxicity issue of reserpine fulfil the enlisted criteria and could be considered therapeutically safe for the development of effective drug candidate. Therefore, the reserpine is classified as non-drug like compound because of one threshold of toxicity.

In silico screening of phytocompounds as AcrB inhibitor
The results of docking interaction between targeted receptor protein AcrB and phytocompounds have shown in Table 1.
The binding free energy of each phytocompound was used to predict the Ki value for the interaction. Based on the binding affinity, phytocompounds tested could be arranged into an order of putative inhibition efficiencies as Chlorogenic acid > piperine > reserpine > capsaicin > curcumin > 2hydroxy 1,4 naphthoquinone > chlorpromazine > and thymol. According to docking studies, 5-methoxyfurfural would act more like the substrate rather than the putative inhibitor and binding affinity was also the lowest (À4.7) of the phytocompounds screened in the assay. All the compounds were stabilised by hydrogen bonding and hydrophobic interactions involving phenylalanine residues except the chlorpromazine and 2-hydroxy 1,4 naphthoquinone as depicted in Table S3 (Supplementary material). The docked out poses and binding of chlorogenic acid, piperine, curcumin and capsaicin is sown in Figure 1.

Molecular dynamics simulation
Further analysis of the dynamics of AcrB-chlorogenic acid complex and their interaction were studied using MD simulations. For MD simulations, the complex with lowest binding energy was used as initial conformation.

Analysis of RMSD and RMSF
For preliminary analysis of MD simulation, root-mean square deviations (RMSD) of both systems were calculated with respect to their backbone of initial structures. RMSD was calculated to assess the stability of both systems (AcrB and AcrB-chlorogenic acid complex) under physiological conditions and data obtain is shown in Figure 2A. The RMSD of AcrB showed some fluctuation during initial duration while it became quite stable after 50 ns. A similar RMSD pattern was obtained for AcrB-chlorogenic acid complex. This shows that both systems reached equilibrium after 50 ns. The average RMSD of AcrB and AcrB-chlorogenic acid complex found to be 0.604 ± 0.083 and 0.614 ± 0.069 nm, respectively. The stability of chlorogenic acid was also assessed and it was found to be stable during entire simulation period. The fluctuation in uncomplexed AcrB and complexed with chlorogenic acid was analysed by calculating root mean square fluctuation (RMSF) of C a atoms of the protein. The RMSF of majority of amino acids of AcrB remained below 0.3 nm both in complexed and uncomplexed form ( Figure 2B). Resides Ala215 to Ile-235 showed highest fluctuation which is due to the random coiled conformation. The RMSF pattern of AcrB alone and complex was almost same indicating that both systems were equally stable. The RMSF of individual atoms of chlorogenic acid showed some variations ( Figure 2C) which is attributed to the dynamical shift of atoms from their initial position at the binding site.

Analysis of R g and SASA, and energies
The mass-weighted root mean square distance of a collection of atoms from their common centre of mass is defined as the radius of gyration (R g ). R g is also considered am important parameter to study the stability of proteins and their complexes during MD simulation (Rath et al., 2021). In a broad notion, globular and compact proteins hace lesser variations in R g compared to expanded and open form of proteins. The R g of AcrB and AcrB-chlorogenic acid complex is presented in Figure 3A. There was a negligible change in R g of both the system over simulation time. The average R g of, respectively. Moreover, the R g of both the systems approximately overlapped each other further indicating the stability of AcrB and its complex is aqueous environment. Solvent accessible surface area (SASA) of biological macromolecules are considered as indicator of the stability during MD simulation studies (Rath et al., 2021). SASA of AcrB and AcrB-chlorogenic acid complex were found to be 439.434 ± 8.832 and 456.991 ± 10.222 nm 2 , respectively ( Figure 3B). SASA of the complex was slightly higher than the protein which is due to the attachment of ligands, making the surface are slightly larger. There was a negligible variation in SASA of both systems further validating the stability of both structures. Finally, the structures were validated by calculating their physicochemical parameters such as potential and total energies ( Figure 3C). Both these energies were stable during entire simulation duration showing that the system reached were in equilibrium and stable during MD simulation.

Analysis of secondary structure and hydrogen bonds
The binding of chlorogenic acid with AcrB studied by analysing the hydrogens bond profiles. The average number of hydrogen bonds were found to be 1.532 ( Figure 4A). Moreover, the hydrogen bond existence between chlorogenic acid and AcrB with >1% existence are plotted ( Figure  4B). As evident from the data, there was continuous hydrogen bond existence for entire simulation period. The   variation in hydrogen bond profile shows the dynamic nature of hydrogen bond. The effect of interaction of chlorogenic acid on AcrB was studied by calculating the average secondary structure from entire trajectory ( Figure 5). The a-helix, b-sheet and coils in AcrB was found to be 37.80, 16.56 and 17.09%, respectively. There were negligible changes in the secondary structure upon the binding of chlorogenic acid validating the structural stability of the complex in aqueous condition.

Principal component analysis
Principal component analysis (PCA) is standard statistical tool for analysis of large-scale motion which is performed by reducing the dimensionality of data set without losing important information, which is characterised by eigenvectors (Fouedjou et al., 2021;Siddiqui et al., 2021). PCA was performed to analyse the differences in flexibility parameters between the AcrB alone and complexed with chlorogenic acid. Using PCA analysis, set of eigenvectors and eigenvalues were projected ( Figure 6A). Both the structures occupied roughly same conformational space denoting similar levels of structural stability. The free energy landscapes of AcrB alone and AcrB-chlorogenic acid were plotted to decipher the protein folding patterns ( Figure 6B). Both the structures reached energy minima in the landscape. For further analysis, the lowest energy minima of both structures were extracted, and Ramachandran plots were made (Figure 7). The phi (u) and psi (w) angles for AcrB alone were found to be À77.15 and   26.42, respectively. The u and w angles for the complex was obtained as À79.30 and 28.64, respectively. There were only two disallowed residues in both AcrB alone AcrB-chlorogenic acid complex showing the stability of both the energy minima structures.

Calculation of binding energies and identification of key residues
The detailed analysis of the binding energies involved in interaction was performed using MM-PBSA calculation. Typically, in protein-ligand interactions, the non-covalent interactions are predominant. Such forces include hydrophobic, van der Waals forces, hydrogen bonds, electrostatic bonds, etc. All these forces either contribute negatively or positively to the overall binding energy (Siddiqui et al., 2019). A total of 100 snapshots of the entire trajectory was taken from 50 to 100 ns at equal intervals MM-PBSA calculations and the result are presented in Table 2. The interaction of chlorgenic acid was mostly favoured by van der Waals and electrostatic forces. There was also small contribution of SASA energy in the overall binding energy. On contrary, polar solvation energy impaired the binding of chlorogenic acid to AcrB. The overall binding energy for the interaction was found to be À9.732 ± 0.461 kcal/mol. The binding energy contribution of all residues was also calculated. The polar, apolar and total binding energies of key residues of AcrB for interaction is shown in Table 3. Phe358, Gln360, Arg418, Ala421, Ile500 and Gly506 were the major contributor to overall binding energy.

Minimum inhibitory concentration (MIC)
The phytocompounds having efflux pump inhibitory activity based on the in-silico analysis was subjected to MIC determination. The MIC value of chlorogenic acid, piperine, reserpine and curcumin against E. coli was 512 lg/mL, chlorpromazine, 2-hydroxy 1,4 naphthoquinone, capsaicin was 128 lg/mL and thymol were 100 lg/mL. Therefore, the highest antimicrobial effect was observed by the thymol. The positive control in this study for MIC determination, ceftazidime, was given a value of 125 g/mL.

Efflux pump inhibition by ethidium bromide accumulation assay
The efflux pump inhibitory potential of active phytocompounds was also confirmed using ethidium bromide accumulation assay. Figure 8 displays fluorescent intensities of EtBr obtained in presence of phytocompounds (chlorogenic acid, piperine, reserpine, 2-hydroxy 1,4 naphthoquinone, capsaicin and thymol) and chlorpromazine against a time period. Fluorescent intensities were the least for the EtBr alone as compared to the added phytocompounds. Among the tested compounds the presence of chlorogenic acid had a better increase in fluorescence by the accumulation of 33.6% followed by piperine 24.2%. Capsaicin and curcumin also showed an increase in fluorescence by 18.3% and 15.3%, respectively.

Synergistic interaction between phytocompounds and antibiotics
The interaction between the seven shortlisted phytocompounds and chlorpromazine with antibiotics tetracycline and nalidixic acid are classified as synergistic, indifferent/additive as given in Table 4. All the phytocompounds showed a varying level of antibiotic potentiating effect. FIC index was ranged in interval from 0.25 to 1.5 for tetracycline and 0.28 to 0.75 for nalidixic acid. The phytocompound 2-hydroxy 1,4 naphthoquinones do not show any potentiating effect both with tetracycline and nalidixic acid (FIC index 1.5 and 0.75, respectively). The interaction of chlorogenic acid with tetracycline and nalidixic acid against E. coli was with FIC index of 0.25 and 0.28, respectively indicating a good synergistic effect. Capsaicin and curcumin showed a similar synergistic effect with nalidixic acid having FIC index 0.5. The data on the effectiveness of the most active combination i.e. chlorogenic acid and tetracycline based on OD 600 after 24-hour treatment is represented as Combenefit HSA plot (Figure 9).

Discussion
Many Gram-negative bacteria use multidrug efflux proteins as a resistance mechanism against a variety of antibiotics. E. coli is one of the most widely encountered resistant bacteria in clinical settings and inhibiting drug efflux mechanisms to reverse resistance is the area of interest (Lamut et al., 2019). Computational strategies have gained a lot of momentum in pharmaceutical research because of their ability to recognise and generate novel promising compounds especially using the molecular docking (Schaduangrat et al., 2020). In the current study, computational studies were used to determine the interaction of selected phytocompounds with AcrB efflux protein of E. coli TG1. A typical RND (resistance-nodulationdivision) pump, which is a tripartite structure comprises of an integral membrane efflux transporter with limited  substrate specificity (AcrB), an outer membrane channel (TolC) and a periplasmic protein adapter (AcrA). Antibiotics bypass through periplasmic space via a porin or lipid bilayer diffusion, where they interact with substrate-binding pocket of AcrB (Opperman et al., 2014). The findings of study revealed that the compounds; chlorogenic acid, piperine, reserpine, curcumin, capsaicin, chlorpromazine, 2-hydroxy 1,4 naphthoquinone and thymol were effective based on their binding affinity and pharmacokinetic properties. The most active phytocompound as efflux pump inhibitor was chlorogenic acid in terms of their predicted binding affinity (À9.1) among all the compound tested. Though chlorogenic acid (5-O-caffeoylquinic acid) have very weak antimicrobial activity it showed strong efflux pump inhibitory potential against MFS transporters of the gram-positive pathogen (Fiamegos et al., 2011). In our modelling studies, all the best-docked compounds showed interaction with phenylalanine residues except the compound chlorpromazine and 2-hydroxy 1,4 naphthoquinone. This is not unusual in considering that phenylalanine residues play a key role in stabilisation and substrate recognition of RND-type drug efflux pumps (Ohene-Agyei et al., 2012). Identification of drug-like features serves as an additional criterion that medicinal chemists employ as a selection factor to choose more promising compounds as leads in order to avoid the huge cost of research in the pre-clinical or clinical phases. Further, the ligands that showed the best binding affinity were also verified by Lipinski's rule of five and ADMET. The Lipinski's rule of 5 is one of the most extensively utilised selection factors. The Lipinski's rule of five is used to screen drugs for good oral bioavailability and permeability (Lipinski et al., 2001), as well as enhanced pharmaceutical properties. The MW of the chlorogenic acid molecule, the number of H-bond Ds, the number of acceptors and the number of rotations are all quite close to the Lipinski's rule's range of parameters. The molecular weight is a measurement of the size of a molecule. Because the transit across biological membranes is unfavourable, large molecules will have difficulty being absorbed (Navia & Chaturvedi, 1996). Among the all screened compounds (19), all followed Lipinski's rule of five except the reserpine. Reserpine extracted from the roots of Rauwolfia vomitoria is a plant-based EPI. The efflux pump inhibitory potential of reserpine is well known against the ABC and MFS transporters of gram-positive pathogens (Garvey & Piddock, 2008;Parai et al., 2020). Reserpine enhances the activity of antibiotics by specifically interacting with the amino acid residues of the Bmr efflux pump protein, which that is involved in tetracycline efflux in B. subtilis (Nikaido, 1998). However, it was found that reserpine could not be used with antibiotics since it showed nephrotoxicity (Pfeifer et al., 1976). Our ADMET server-based analysis also indicated toxicity prediction of reserpine. On the other hand, chlorogenic acid fulfils the criteria of drug-likeness prediction. Based on the toxicity prediction through the PROTOX II  The FIC index 0.5 was considered synergistic, FIC index between 0.5 and 4 was additive/indifferent, and FIC index 4 was considered antagonistic. All the phytochemicals and chlorpromazine were taken at their respective sub-MIC concentration (MIC/4). MIC of Tetracycline -32 lg/mL. MIC of Nalidixic acid -512 lg/mL. server, the acute toxicity values of chlorogenic acid were LD50 > 1190 mg/kg. These levels are significantly higher than the MIC for this phytocompound alone and in combination with the tested antibiotic (FIC), implying its therapeutic potential. Next, we identified the fluorescence of ethidium bromide which acts as a substrate for AcrB-TolC efflux pump of E. coli TG1 by selected phytocompound using a fluorometric method. Chlorogenic acid, piperine, reserpine, 2-hydroxy 1,4 naphthoquinone, capsaicin and to some extent thymol clearly inhibited EtBr efflux in a transient manner at subinhibitory concentration (1/4MIC or less). Chlorogenic acid showed potent inhibitory effect in the efflux pump at a concentration much below their MIC. The addition of chlorogenic acid, followed by piperine and capsaicin leads to the EtBr accumulation in E. coli TG1 by 33.6%, 24.2 and 18.3%, respectively. As regard of efflux pump inhibition, chlorogenic acid generating the highest accumulation level in E.coil TG1 when directly compared with reference chlorpromazine. In contrast thymol showed weak accumulation of EtBr in the E.coil TG1 besides showing interaction with AcrB protein in in silico analysis. It might be suggested that thymol inhibited the intracellular accumulation of EtBr in E.coil TG1 by affecting certain protein mechanisms, it could have interfered with processes like porin expression and cell wall permeability, or it could have interfered with EtBr itself. However, the per cent increase in fluorescence by reserpine was lower in E. coli TG1. The decreased activity of reserpine may be due to acting as pump substrates or competitive inhibitors that bind to the same or an overlapping site in the binding pocket (Metzger et al., 2002). The effect of curcumin and capsaicin uncovering their potential to be act as efflux pump inhibitor in addition P-gp (P-glycoprotein) inhibitor (Nabekura et al., 2005;Negi, 2014;Shriram et al., 2018). Further in vitro validation of potent efflux pump inhibitor was determined by interaction with antibiotics using checkerboard synergy assay. The MICs of tetracycline (1-0.125 lg/mL) was decreased from 32 to 128-fold by all the selected phytochemicals except the 2-hydroxy 1,4 naphthoquinones. Synergism was not obtained in any combination of 2-hydroxy 1,4 naphthoquinones with tetracycline. The indifferent effect of 2-hydroxy 1,4 naphthoquinones, Chlorpromazine, and thymol with nalidixic acid was also observed against E. coli TG1. The best synergistic interaction among all the combination tested was found between chlorogenic acid and tetracycline with FIC index (0.25). To bolster these findings, and based on the fact that the RND efflux pumps operate by utilising the electrochemical gradient generated by the proton motive force (PMF) (Anes et al., 2015). Because of its negative surface charge, chlorogenic acid may bind to the outer membrane (OM) via electrostatic interactions and chelate Mg 2þ , disrupting OM integrity. These antibiotics were unable to penetrate the intact OM of Gram-negative bacteria, but they were able to penetrate the damaged OM. The findings based on decrease in MIC of tetracycline by chlorogenic genic may be consistent with our hypothesis on alternative mode of action that at low concentrations it inhibits efflux system and acts on E coli membranes. Previously, it was found that half MIC of chlorogenic acid causes increase in outer membrane permeabilization of Shigella dysenteriae, making the bacteria sensitive to antibiotics erythromycin and rifampicin below their respective MIC concentration (Lou et al., 2011). The loss of membrane integrity is linked to the loss of the cell's ability to synthesise ATP (Brogden, 2005). Furthermore, at low concentrations, chlorogenic acid may inhibit E. coli efflux systems by interfering with the energy required to keep the pumps functional, as an energetic inhibitor like CPZ inhibiting efflux activity through transient membrane potential dissipation. Chlorpromazine and thymol exhibited synergistic interaction towards a single antibiotic tetracycline. These results revealed that chlorogenic acid altered cell membrane permeability, causing the bacterial cell membrane depolarisation. Figure 9. Combenefit mapped Surface HAS plot. The data is based on OD 600 values obtained on chlorogenic acid and tetracycline combination from the 24-hour incubation used to generate this plot. Cream to light brawn shows increased reduction (synergy) while dark brawn indicates no reduction (non-synergy).
Previously, in vitro synergistic interaction of chlorogenic acid with quinolone antibiotics (ciprofloxacin, Gatifloxacin and Clinafloxacin) have been reported with FIC index 1 (Chai et al., 2019). Literature suggest that, chlorogenic acid isolated from the green coffee been extract found to be safe for the patient having mild hypertension symptoms (Watanabe et al., 2006). Additionally, since probiotic bacteria are not susceptible to chlorogenic acid, they can be used to replace synthetic antibiotics as a preventive safeguard food additive, cutting the cost of medicine. Though, the antimicrobial and antibiofilm activity of chlorogenic acid was previously reported against gram-positive strain (Karunanidhi et al., 2013;Zhang et al., 2019). To the best of our understanding, this is the first report of efflux pump inhibitory potential and synergistic interaction between chlorogenic acid and tetracycline against efflux overexpressing E. coli TG1.

Conclusion
The above findings may form the basis of in silico screening and evaluation of phytocompounds against AcrB efflux pump responsible for MDR in Gram negative bacteria. In silico screening, study revealed that many of the selected phytocompounds showed a varying level of efflux pump inhibitory activity (based on binding affinity) against AcrB transporter of E. coli TG1. The drug-likeness, ADMET prediction, and toxicity analysis of the compounds that possess EPI activity may aid in the development of non-toxic and efficient phytocompound and antibiotic formulations. Both ethidium bromide accumulation and checkerboard synergy assay results demonstrated that chlorogenic acid act as effective efflux pump inhibitor potentiating the efficacy of tetracycline and nalidixic acid. The other phytochemicals such as piperine, capsaicin and curcumin in combination with antibiotics were also able to provide stable therapeutic outcomes with higher efficacy. Our in vitro work received direct support from the in-silico validation, where the chlorogenic acid act synergistically in combination with tetracycline and reduced antibiotic dosage by 128-fold. As the prevalence of drug-resistant infections continues to rise alarmingly, particularly among nosocomial patients, combination therapy of chlorogenic acid and tetracycline would be a more effective formulation for combating multidrug resistance. Therefore, the synergistic therapeutic combinations may be helpful in combating MDR infections in community or hospital settings. Further research is needed for the genotypic validation, therapeutic efficacy, and safety of the above active combinations in vivo infection model.