Computational binding affinity and molecular dynamic characterization of annonaceous acetogenins at nucleotide binding domain (NBD) of multi-drug resistance ATP-binding cassette sub-family B member 1 (ABCB1)

Abstract Multi drug resistance (MDR) in tumor might be caused leading to the overexpression of transporters, such as ATP-binding cassette sub-family B member 1 (ABCB1). A combination of non-toxic and potent ABC inhibitors along with conventional anti-cancer drugs is needed to reverse MDR in tumors. A variety of phytochemicals have been previously shown to reverse MDR. Annonaceous acetogenins (AAs) with C35/C37 long-chain fatty acids were reported for their anti-tumor activity, however, their effect on reversing MDR is not yet investigated. We aimed to investigate some selective AAs against the B1 subtype of ABC transporter using computational studies. Various modules of Maestro software were utilized for our in-silico analysis. Few well-characterized AAs were screened for their drug-likeness properties and tested for binding affinity at ATP and drug binding sites of ABCB1 through molecular docking. The stability of the ligand-protein complex (lowest docking score) was then determined by a molecular dynamic (MD) simulation study. Out of 24 AAs, Annonacin A (−8.10 kcal/mol) and Annohexocin (−10.49 kcal/mol) docked with a greater binding affinity at the ATP binding site than the first-generation inhibitor of ABCB1 (Verapamil: −3.86 kcal/mol). MD simulation of Annonacin A: ABCB1 complex for 100 ns also indicated that Annonacin A would stably bind to the ATP binding site. We report that Annonacin A binds at a greater affinity with ABCB1 and might act as a potential drug lead to reverse MDR in tumor cells. Communicated by Ramaswamy H. Sarma


Introduction
Chemotherapy is one of the widely used strategies to treat metastatic cancer (Lippert et al., 2008). The emergence of drug resistance in tumor cells leads to the failure of chemotherapy (Volm & Efferth, 2015). Reduction of intercellular concentration of drugs and frequent mutations in the drug targets are the major factors contributing to drug resistance in tumors (Bunting, 2002). The membrane transporters, such as ATP-binding cassettes (ABC) are primarily responsible for the efflux of chemo drugs from the tumor cells. The drug efflux mechanism of ABC transporters is dependent on the energy released during ATP hydrolysis (Giacomini et al., 2010).
One of the well-characterized ABC transporters is human permeability glycoprotein (P-gp)/Multidrug resistance (MDR) protein 1 or also known as ATP-binding cassette sub-family B member 1 (ABCB1). It has been recognized to play a key role in the development of MDR in cancer cells (Mishra et al., 2013). ABCB1 is encoded by the ABCB1 gene, possessing a molecular weight of 170 kDa (Sharom, 2006). High levels of ABCB1 are expressed in colorectal cancer cells and colorectal mucosa (Linn & Giaccone, 1995). The functional unit of ABCB1 comprises two nucleotide-binding domains (NBD) and two transmembrane domains (TMD) (Gutmann et al., 2010). The NBD is highly homologous with full-fledged Walker A, Walker B and signature motifs (Hrycyna et al., 1998). Transmembrane (TM) 10/TM 11 crosses over with TM4/TM5 and forms a cavity to form a drug-binding pocket in the inward (cellular) facing state of human ABCB1 (Katayama et al., 2014). It has been previously reported that as the binding affinity of ATP toward NBD1 is higher, NBD1 acts as the most preferred site for ATP binding. The ABCB1 transport cycle begins with the substrate (Oncodrugs) binding at the protein's TMD. ATP binds to the two NBD alternately, forming a closed dimer. As a result, the ATPase hydrolysis is favored and the energy thus produced facilitates a conformational change in the structure of the protein (Urbatsch et al., 1998;Yang et al., 2003). This structural conformation continuously pushes the bound drugs out of the cell and hence the tumor cell develops drug resistance (Gottesman et al., 2009;Jones & George, 2014). ABCB1 has been shown to have multiple substrates binding sites, including the H site, which is responsible for the binding of compounds, such as Hoechst-33342 and colchicine, the R site, which is capable of binding to rhodamine-123 and anthracyclines, and the P site, which is responsible for the binding of prazosin and progesterone (Martin et al., 2000;Shapiro & Ling, 1997;Shapiro et al., 1999). The development of ABCB1 inhibitors or modulators promises a number of therapeutic benefits in tumor management. Combinatorial chemistry approaches resulted in the discovery of ABCB1 inhibitors of the first, second and third generations (Palmeira et al., 2012;Srivalli & Lakshmi, 2012). The first-generation inhibitors are substrates by themselves and therefore compete with the drugs and act as competitive inhibitors. A well-known example is a verapamil, a calcium channel blocker, which had failed in the clinical trials due to its extreme toxicity (Dantzig et al., 2003). Slight structural modifications in the first-generation inhibitors led to the rise of the second-generation inhibitors, which were also ABCB1 substrates with a lower binding affinity toward the protein. Dexverapamil, the R-isomer of verapamil is an excellent example of second-generation inhibitors. The second-generation inhibitors were found to be cytochrome P450 3A4 substrates and hence majority of them hindered the activity of anticancer drugs as well. This caused issues with chemo drug clearance, making it difficult to determine appropriate drug dosage in cancer patients. Hence, these molecules also failed in their clinical trials (Thomas & Coley, 2003). High-throughput technologies and combinational chemistry were involved in the generation of the third generation of inhibitors. They were reported to be highly specific and also lacked interaction with cytochrome p450. Tariquidar, an anthranilamide derivative belonging to this category was experimentally proven to inhibit ATPase. These inhibitors were also rejected in several phase III trials (Ozben, 2006). Recently, several non-toxic plant-based molecules have been attempted to inhibit ABCB1. Gomisin A, a dibenzocyclooctadiene compound extracted from Schisandra chinensis has been proved to alter the interaction of ABCB1 substrate competitively and thus was shown to reverse MDR (Wan et al., 2006).
In line with these compounds, Annonaceous Acetogenins (AAs) of Annona muricata are a class of C-35 secondary metabolites derived from long-chain fatty acids of the polyketide pathway (Alali et al., 2010). AAs are reported to possess extensive anti-tumor activity (Carmen Zafra-Polo et al., 1998) and are powerful inhibitors of NADH: Ubiquinone Oxidoreductase (Abe et al., 2008;Zafra-Polo et al., 1996), mitochondrial complex I, and thereby capable of generating demand for adenosine triphosphate (ATP) in aggressive tumors (Degli Esposti et al., 1994;Wolvetang et al., 1994).
Few experimental studies (performed in BEL 740, HepG2 and HEK-293T cells) reported the usage of acetogenins in reversing MDR (Qian et al., 2015). Furthermore, our previous in-silico study has reported that AAs have a higher affinity for the inward facing conformation of ABCB1 (Jeevitha Priya et al., 2020). However, AAs were also reported for their neurotoxicity in mesencephalic cultures at higher concentrations (6 mg/ml) (Lannuzel et al., 2002). Moreover, chronic exposure to alkaloids of Annona muricata induced Parkinsonism in humans and animals (Caparros-Lefebvre & Elbaz, 1999). Hence AAs are not employed as anti-cancer drugs.
If AAs at lower concentrations can aid in reversing MDR in colon cancer cells, they could be proposed for its use as an adjuvant along with conventional anti-cancer drugs. Hence, we aimed to predict the binding affinity and molecular dynamic (MD) characterization of AAs at the NBD of multidrug resistance ABCB1.

System and software
All in-silico studies were carried out using the Schrodinger molecular modeling software suite installed in a DELL system with 64-bit LINUX OS. The ligands were prepared using LIGPREP; molecular docking and dynamics were carried out using Glide version 6.2 ( Schr€ odinger Release 2019-1: Maestro, Schr€ odinger, LLC, New York, NY, 2019 ). and Desmond version 3.6 (Schr€ odinger Release 2019-1: Desmond molecular dynamics system, D. E. Shaw Research, New York, NY, 2019), respectively.

Retrieval of ligands and ADME prediction
A total of 24 AAs (possessing anticancer activity) (Dilipkumar & Agliandeshwari, 2017) and the reference compounds were retrieved in structured data format (SDF) from the ChemSpider (http://www.chemspider.com) and Drug bank database (https://go.drugbank.com/), respectively. The 2D structures of the ligands (AAs and reference compounds) were shown in Supporting information S.1. Before docking, the ADME properties of AAs were evaluated using the QikProp module of the Schrodinger suite (Schrodinger Release 2019-1a, 2019). Out of all AAs, compounds abide by the drug likeness properties as per Lipinski rule of five (Lipinski, 2004) were only selected for docking studies.

Molecular docking
Molecular docking of protein and ligand was performed in the extra precision (XP) mode of Glide version 7.7 in the Maestro workspace (Schr€ odinger Release 2019-1b, 2019). The methodology to determine the binding affinity of protein and ligand involves steps, such as protein preparation, ligand preparation, grid generation and molecular docking.

Preparation of protein (refinement of protein)
The outward-facing crystal structure of human ABCB1 bound with ATP was retrieved from the RCSB PDB database (PDB ID: 6C0V) (Kim & Chen, 2018). By using the protein preparation wizard, the protein was refined by editing missing atoms in the selected chains, adjusting the bond orders, removal of bound water molecules and generation of sulfide bonds. Finally, the protein structure was minimized using the optimized potentials for liquid simulations-2005simulations- (OPLS-2005 force field to the default root mean square deviation (RMSD) value of 0.30 (Schr€ odinger Release 2019-1c, 2019).

Ligand preparation
The ligands were prepared using the LIGPREP module of Schrodinger Maestro (Schrodinger Release 2019-1d, 2019). The specified chiralities of the ligands were retained while optimizing their geometries. Bond orders were fixed and the charged groups were neutralized. The ligands were then energy minimized by the OPLS-2005 force field using the Epik module and the missing hydrogen bonds were added.

Grid generation
The grids were generated for both the ATP binding regions (NBD1 and NBD2) in the target protein individually using the Glide grid generation module. The grid dimension was adjusted to 10 Â 10 Â 10 of the crystal structure of the protein at the regions of NBD and the drug binding site (Aller, 2010;Kim et al., 2006;Sauna & Ambudkar, 2007). The van der Waals radius scaling factor was adjusted to 1.0 with no constraints. The partial charge cutoff was maintained around 0.25 and finally, the grids were generated.

MM-GBSA
The protein-ligand complex with the highest docking score was selected and their end-point free energy of binding was predicted by molecular mechanics-generalized born and surface area (MM-GBSA). MM-GBSA is a free energy binding calculation, which is intermediate in both accuracy and computational effort between empirical scoring and strict alchemical perturbation methods (Genheden & Ryde, 2015). MM-GBSA was calculated by using the Prime-MM-GBSA module. OPLS-AA force field was used for computing the relative binding energies. The binding energy was calculated using the formula (Equation (1)) mentioned below (Schr€ odinger Release 2019-1e, 2019).

MD simulation
The MD simulation was carried out to predict the stability of the protein-ligand complexes using the Desmond version 3.6 package (Schr€ odinger Release 2019-1f, 2019). The MDs simulation analysis was performed using the default parameters of the system builder. As ABCB1 is a TM protein, the predefined dipalmitoyl phosphatidylcholine (DPPC) membrane was chosen and the position of the membrane was adjusted according to the binding site residues. The temperature of the membrane patch was pre-equilibrated at 325 K. The OPLS3e force field was selected and SPC solvent model was used for mimicking the water molecules. The shape and size of the repeating units at 10 Å distance were limited by enabling the orthorhombic periodic boundary option. The entire system was neutralized by adding Clions before the simulation. The salt concentration was set to 0.15 M to simulate the background salts. The MD was performed in isothermal-isobaric (NPT) ensemble at 1.01325 bars over 100 ns with the recording interval of 4.8 ps. The entire system was relaxed before the run, and the simulation interaction diagrams were analyzed thereafter.

Results
This study aimed to identify an inhibitor of ABCB1 which either interrupts the binding of ATP at the NBD or the substrates (chemotherapeutic drugs) at the TMD. Initially, a total of 24 AAs were tested for their ADME properties. The compounds which passed these properties were docked at the NBD and TMD sites of the crystal structure of ABCB1 in an ATP-bound, outward-facing conformation (PDB ID: 6C0V). AAs with the lowest docking score (at NBD and TMD) in comparison to reference and co-crystallized ligand (ATP) was selected and tested for its MD stability.

Prediction of ADME properties
Twenty-four AAs were assessed for Lipinski rule of five parameters and ADME properties using QikProp. Few critical physical descriptors and bio-pharmaceutically important parameters including octanol/water and water/gas log Ps, log S, log BB, overall CNS activity, were evaluated (Table 1). The acetogenins were predicted to be readily absorbed by the human small intestine (% intestinal absorption). Moreover, the negative values of QPlogBB and CNS indicated that all selected AAs were inactive toward CNS. As QlogPo/w of AAs is between À2 and 6.5, it can be simplified that the compounds are relatively hydrophilic and tend to have high water solubility. Finally, 23 AAs (excluding solamin) were found to abide by Lipinski rule of five and therefore, can be considered as 'drug-like'.

Molecular docking
Molecular docking of AAs was performed at the NBD and TMD domains of the protein (ABCB1) to determine the ligand's binding poses. Both the active sites in the NBD domain (NBD1 and NBD2) were explored for the docking studies. Due to the structural constraints of ABCB1's TM regions, docking of AAs with the TMD of the crystal structure of ABCB1 produced no valid poses; however, docking of ligands at the crystal structure of ATP bound ABCB1produced greater binding affinity. Among the AAs, Annohexocin had the lowest binding energy of À10.49 and À10.33 kcal/mol at the NBD1 and NBD2 sites, respectively, while Annonacin A had À8.10 and À8.31 kcal/mol at the NBD1 and NBD2 sites. The three-dimensional interactions of the AAs: ABCB1 are also shown ( Figure 1)

Molecular mechanics-generalized born and surface area (MM-GBSA)
The docked poses were also subjected to MM-GBSA calculation in order to predict the end-point binding free energy.MM-GBSA dG bind of complexes of annonacin A with ABCB1, annomuricin B with ABCB1 and annohexocin with ABCB1 were À105.23, À126.01 and À81.7, respectively ( Table  2). The MM-GBSA dG bind of acetogenins with NBD regions was found to be 4-5-fold higher than the co-crystallized ligand (ATP: À79.9). Acetogenins, namely annomuricin E, had the lowest docking score of À5.39 kcal/mol and a dG bind of À92.0, while annohexocin had the highest docking score of À10.49 kcal/mol and a dG bind of À81.7. Similarly, the cocrystallized ligand (ATP) was predicted to have a binding affinity of À20.04 kcal/mol and a lower dG bind of À79.9. It is observed that the predicted docking scores and the MM-GBSA dG bind of the ligands were found to be positively correlated (R ¼ 0.03), which was not statistically significant.

Molecular dynamic simulation
The dynamic interactions of the docked complexes were studied using Desmond. The complexes obtained from molecular docking: annonacin A: ABCB1, Annohexocin: ABCB1 and annomuricin B: ABCB1 and the crystal structure of the protein with its co-crystallized ligand were used for the MD simulation study.

Root mean square deviation (RMSD) analysis
Initially, the apo form of ABCB1 was tested for MD simulation for 50 ns, the simulation report indicated a stable and equilibrated state at 45 ns and it was productive up to 50 ns, with an RMSD of 2.1-2.7 Å. However, the MD simulation of the apo form of ABCB1 was extended up to100 ns and the RMSD of the apo form of ABCB1 was also within the same range (2.1-2.7 Å) ( Figure 2). Therefore, the MD simulation of the top-scoring AAs: ABCB1 complexes were subjected to 100 ns in our study. The RMSD of ABCB1: Annonacin A complex was observed between 2.1 and 2.4 Å from 40 ns. The maximum RMSD value of the ligand was observed to be 6.4 Å and there was a diffusion of ligand away from its binding pocket at 90 ns. However, the rigidity of the protein is increased upon ligand binding and therefore the complex maintained its stability till 90 ns (Figure 3(A)). On the other hand, the RMSD of ABCB1: Annohexocin complex was 2.8-3.0 Å and there were also limited fluctuations at time intervals 0-5 and 65-85 ns (Figure 3(B)). The RMSD of ABCB1: annomuricin B complex was between 3.0 and 4.6 Å with only little fluctuations during simulation (Figure 3(C)). It is well-known that the changes of the order of 1-3 Å are perfectly acceptable for globular proteins (Carugo & Pongor, 2001).

Root mean square fluctuation (RMSF)
The conformational changes of the amino acid residues of ABCB1 upon ligand binding were monitored by root mean square fluctuation (RMSF) of protein. During the simulation, the amino acid residues of the protein showed some fluctuations and it is evident from the simulation interaction diagram (Supporting information S.4). Furthermore, the RMSF of the protein's apo form was also examined (Supporting  information S.4.D), indicating that the fluctuations observed in the protein fitted with the ligand (Supporting information S.4. A-C) were caused by the binding of the ligands. Moreover, the RMSF profiles (Supporting information S.4) revealed that the interaction of annonacin A with Val596, Ile620, annohexocin with Gly250, Val596, Ile620, Ile900 and annomuricin B with Met50, Val596 and Ile620 of ABCB1 might contribute to the primary changes in the apo form of the protein. In general, the crystal structure of protein constitutes 57.92% of helices and 8.99% of beta-strands (Supporting information S.5). Besides, ligand RMSF showed the nature of fluctuations of internal atoms of the ligand upon protein binding. The atom of the Annonacin A from C 1 to C 7 showed a greater fluctuation and this might be due to exposure of atoms to the solvent molecules. However, the  atom of Annonacin A from C 8 to C 42 showed less or no fluctuations due to high forces of attraction at the vicinity of the binding pocket of the protein (Supporting information S.6). Similarly, atoms of annohexocin and annomuricin B showed a higher fluctuation from C 1 to C 9 and C 1 to C 7 , respectively (Supporting informations S.7 and S.8).

Protein-ligand contacts
The protein: ligand (ABCB1: Annonacin A, ABCB1: Annohexocin and ABCB1: annomuricin B) interactions were stabilized by the formation of water bridges, hydrogen, ionic and hydrophobic interactions. The amino acid residue of ABCB1, such as Tyr401, Val 407, Ile409, Leu443Val478, Ile1154, Phe1157, Leu1161, Leu1176 and Ile1184 were involved in hydrophobic interactions with annonacin A. Multiple water bridge formation was also found between annonacin A and amino acid residue, such as Asp164, Lys408, Ser434, Thr435, Gln441, etc. (Figure 4(A)). No intramolecular hydrogen bonding was observed within the Annonacin A. The interaction of ABCB1-annonacin A has occurred over 30% of the entire simulation time ( Figure  4(B)). During the simulation, 58% of interactions were observed with the residue Gln1175 (Figure 4(B)). Furthermore, protein residues with multiple contacts were also recognized (Supporting information S.9). A similar interaction profile was observed between the amino acid residues of ABCB1 and the ligands (annohexocin and annomuricin B). Annohexocin interacted with the ABCB1 residue Lys1181 approximately 94% of the time during the simulation. Whereas, Asp555 of ABCB1 interacted with annomuricin B (100%) throughout the entire duration of simulation (Supporting information S.10 and S.11).

Ligand torsion profile
The conformational changes of each rotatable bond of the Annonacin A that occurs throughout the simulation are represented via the torsion profile. A total of 30 rotatable bonds were observed in Annonacin A. The rotatable bond between C 7 -C 30 and C 4 -C 14 of annonacin A (Supporting information S.12.A) was spinning around throughout the simulation but the other rotatable bonds tend to possess a defined spread of the torsion angles. Moreover, the majority of the rotatable bonds of the ligand annonacin A were found at À1800 and 180 . Annohexocin and annomuricin possessed 32 and 31 rotatable bonds, respectively (Supporting information S.12.B and C). The torsion of each rotatable bond is given by a bar plot and a dial plot. The dial plot represents the conformation of the torsion: initially, during the course of the simulation, it was present at the center, later it appeared radially outwards (Supporting information S.12).
The residue-specific grid generation in molecular docking of the acetogenins with ABCB1 indicated the selectivity of ABCB1 over other ATP binding cassette family (ABCG2 and ABCC1) as provided in Supporting information S.13. Molecular docking analysis demonstrated the binding capacity of apigenin (lowest binding energy, À5.43 kcal/mol) with the ATP-binding site of P-gp through hydrogen bond formation with residues Lys408 and therefore hindering ATP binding which is essential for the function of ABC transporters (Saeed et al., 2015). Natural compounds, such as Biochanin A and silymarin were also reported to be directly involved in modulating the ATPase activity of human ABCB1 (Zhang & Morris, 2003). Moreover, flavonoids were also reported to occupy the ATP binding site of ABCB1 and hence act as specific chemosensitizers of MDR1 (Conseil et al., 1998). However, the above-reported findings require an experimental study to explore the mechanism of ABCB1 inhibition.
Further, the free energy of binding of AAs at NBD of ABCB1 was calculated through Prime MM-GBSA. The endpoint binding free energy (MM-GBSA dG bind) of complexes annonacin A: ABCB1 (À105.23 kcal/mol), annomuricin B: ABCB1 (À126.01 kcal/mol) and annohexocin: ABCB1 (À81.7 kcal/mol) was greater in comparison to ATP: ABCB1 (À79.96 kcal/mol) and the first-generation inhibitor verapamil (À67.46 kcal/mol). Naturally occurring compounds were reported as dual inhibitors of transport activity of ABCB1 and breast cancer resistance protein (BCRP). The binding free energy (dG bind) of these compounds was in the range of À24.55 to À50.25 kcal/mol. The effective binding energies of the identified (natural) compounds were reported to be 7.44 ± 3.45 kcal mol À1 and were experimentally proven to inhibit MDR in A549 cells (Sachs et al., 2019). From our study, it was evident that AAs (annonacin A, annomuricin B and annohexocin) binds with the NBD region of ABCB1 with greater binding energy than these reported natural compounds. In addition, it has been observed that the binding affinity of AAs (annonacin A, annomuricin B and annohexocin) toward the NBD region of ABCB1 was found to be higher than Tariquidar, a third-generation inhibitor (À5.46 kcal/mol). Therefore, our compounds of interest (annonacin A, annomuricin B and annohexocin) might bind stronger with ABCB1 than the co-crystallized ATP ligand and the reported P-gp inhibitors (especially first-generation inhibitor) as well. This in turn will help in inhibiting the efflux function of P-gp. However, this needs to be experimentally validated.
Afore mentioned, the stability of ABCB1: AAs complexes was analyzed by MD simulation for 100 ns. The MD of the TMD region of ABCB1 was reported to be more active in comparison with the NBD regions (Ferreira et al., 2012;Liu et al., 2013;O'Mara & Mark, 2012). In addition, it was also reported that the ATPase activity observed at the NBD of ABCB1 is assisted by the flexibility of the protein (Wen et al., 2013). Since ABCB1: annohexocin and ABCB1: annomuricin B complexes were unstable during MD simulation, the results of ABCB1: annonacin A has been discussed further. Analysis of MD trajectory of annonacin A: ABCB1 complex indicated greater stability of annonacin A with NBD1 of human ABCB1 during the simulation of 100 ns.
It is highly evident that the protein (ABCB1) was found to be compactly packed, and it was noticed that the binding of Annonacin A did not change the rigidity of the protein. Surprisingly, the RMSD of the complex experienced a few variations (> RMSD of the free protein) at 90 ns. This indicated that the ligand was buried at the binding pocket of the protein stably and later diffused away at the end of the simulation. It was also seen that the RMSF of C-a atoms of residues at NBD sites of the protein also indicated the stable nature of the secondary structures of the protein upon binding with Annonacin A. It is also clear that atoms of annonacin A occupied most of the binding cavity of NBD of the protein. Annonacin A interacted strongly through hydrogen bond formation and water bridges with the residue GLN1175 of ABCB1 which constituted 58% of the entire simulation time. Virtual screening of compounds from ZINC database against ABCB1 reported that compounds, such as methyl 4-[bis(2-hydroxy-4-oxochromen-3-yl) methyl] benzoate (ZINC 09973259, CID 4694077) and ethyl 1-(1,3-benzodioxole-5-carbonyl) À3-(3-phenylpropyl) piperidine-3-carboxylate (ZINC 15078148, ZINC 15078146, CID 26410703 and CID 45252040) inhibited ABCB1 at its NBD by interacting with residue Gln1175 with a high docking score (Brewer et al., 2014). In addition, a protease inhibitor nelfinavir exhibited a docking score of 9.52 (PDB ID: 6C0V) at the NBD of ABCB1 and inhibited P-gp non-competitively by blocking the activity of ATP synthase (in vitro studies) (Thomas & Coley, 2003). Similarly, chemically synthesized derivatives of D-a-Tocopheryl polyethylene glycol 1000 succinate (TPGS) were noted to play an important role in P-gp inhibition by blocking the ATP binding sites of the transporter (Binding affinity ranged from À5 to À11 kcal/mol) (Liu et al., 2018). Since the docking scores of these reported compounds and annonacin A toward the NBD of ABCB1 are closer, it can be expected that annonacin A could also potentially block ATP synthesis and increase the intracellular drug concentration in tumors.
Moreover, the IC50 values of annomuricin E (Annona muricata) in HT-29 colon cancer cell and CCD841 normal colon cells were observed to be 1.62 ± 0.24 and 32.51 ± 1.18 lg/ml after 48 h, respectively. Since annomuricin E and annomuricin B are isomers of annomuricin we could expect a similar cytotoxic profile for annomuricin B (Zorofchian Moghadamtousi et al., 2015). In addition, AA 89-2 from Atemoya plant was found to possess cytotoxic effects against KBv200 cells (multidrug-resistant cells) and KB cells (parental drug-sensitive cells). Also, acetogenins (bullatacin, 89-2) were reported to be involved in intracellular accumulation and reversing the mechanism of MDR by decreasing the P-gp function (Fu et al., 2003(Fu et al., , 1999. Further, the drug score and drug likeness of the AAs predicted by Osiris property explorer (https:// www.organic-chemistry.org/prog/peo/) revealed that the acetogenins are likely to be used as drugs (Supporting information S.14) (Ayati et al., 2012). Besides, the predicted significant pharmacokinetic parameters also showed that the AAs are druggable candidates. The MD trajectories of the Protein: ligand complexes (annonacin A: ABCB1, annohexocin: ABCB1 and annomuricin B: ABCB1) showed relatively similar and were consistently stable throughout the entire simulation. Significant descriptors, such as RMSD, RMSF and Protein: ligand contacts confirmed the structural integrity of the docked complexes. Besides, the MD trajectories of the three complexes, annonacin A: ABCB1 complex-maintained stability with the lowest RMSD throughout the atomistic simulation of 100 ns. However, crystallization conditions differ from the simulation conditions, and hence experimental studies are needed to warrant this information.

Conclusion
To conclude, out of the 24 AAs, we observed a greater binding affinity of annonacin A with ABCB1 through molecular docking and dynamic simulation studies. We hypothesize that annonacin A may compete with ATP for binding at the NBD site of ABCB1 and thereby reduce overall energy generated during ATP hydrolysis. As a result, the efflux of chemo drugs by ABCB-1 (P-gp) may also be either decreased or prevented in colon cancer cells. We have predicted and demonstrated the molecular interactions of AAs with NBDs of ABCB1 at its outward-facing conformation through in-silico simulation. However, in-vitro and in-vivo studies are needed in an order to explain the pharmaco-kinetic and -dynamic behavior of Annonacin A in biological systems. study. The computational study was carried out at molecular modeling lab facility (Schr€ odinger Suite), PSG College of Pharmacy, Coimbatore.

Disclosure statement
The authors report no conflicts of interest.

Ethics approval
This article does not contain any work pertaining to the involvement of human participants or animal studies.

Consent to participate
No human studies were conducted and hence not applicable.

Consent for publication
No human studies were conducted and hence not applicable.

Availability of data and material
Not applicable.

Code availability
Not applicable.

Funding
No funds, grants or other support was received for conducting this research.

Authors Contribution
All authors contributed equally to the design and implementation of the research. Computational studies were performed by Jeevitha Priya Manoharan, Kavinkumar Nirmala Karunakaran, Subramanian Vidyalakshmi and Karthik Dhananjayan. The first draft of the manuscript was written by Jeevitha Priya Manoharan and modified by Karthik Dhananjayan and Subramanian Vidyalakshmi. All the authors proof read and approved the final manuscript.