Structural Mapping of Inhibitors Binding Sites on P-glycoprotein: Mechanism of Inhibition of P-Glycoprotein by Herbal Isoflavones

Aims: Understanding the pattern of inhibitors binding to p-glycoprotein (Pgp). Study Design: Pgp is an ATP dependent transporter protein, responsible for multi-drug resistance in metastatic tumors. It removes toxins by exporting a variety of structurally unrelated compounds outside the cells, which make Pgp a promising target for designing anti cancer supplementary therapeutic molecules. Isoflavones are present in soyabean and other herbal extracts. The idea was to explore inhibitor binding sites on Pgp to find hotspots which eventually may prove useful in designing compounds with higher specificity and affinity. Place and of Study: School of Biotechnology, Gautam Buddha University, Greater Noida, between February 2012 and December 2012. Methodology: The biochemical nature of binding of isoflavones to Pgp has been extensively studied, but the atomic details of their interactions were not understood. Therefore, we have used in silico methods to study binding eight isoflavones bind to site 2 by forming van der Waals and H-bonded interactions. Both the sites are highly hydrophobic in nature and are contributed mainly by side chain of non polar residues present on twelve transmembrane  -helices. Site 1 has minimum dimension of 7.5Å and maximum as 22Å whereas, site 2 is wider and deeper than site1. One sidewall of the site 2 is formed by polar amino acid residues of helix H12, which makes several hydrogen bonds with ligands. Conclusion: Structure analysis revealed that addition of polar group to hydrophobic ligand may enhance its binding affinity for Pgp, which may be used for designing potent inhibitors to find lead compounds for drug design.


INTRODUCTION
Toxic chemotherapy is used for treatment of cancer. Multi drug resistance (MDR) is the most common phenomena occurring in metastatic cancerous cells. Therefore, MDR has been major thrust area for cancer biologist since long time. MDR is generally conferred by efflux of drug compounds by a group of transmembrane proteins in an ATP dependent manner [1]. These proteins get overexpressed in cancerous cells. A variety of such ATP dependent transporters have been identified, such as P-gycoprotein (ABCB1) [2] MRP1 (multi drug resistance associated protein 1, ABCC1) [3], MRP2 (ABCC2), MRP3 (ABCC3) [4] and BCRP/MXR1 (ABCG2) [5,6]. These are expressed in various tissues of the human body including intestinal lumen, bile canaliculus in liver hepatocytes and capillaries in the blood brain barrier [5]. Along with other members, Pgp has also been reported to be involved in the exotransport of anticancer drugs [7,8]. It has a very broad specificity to compounds ranging from 330Da to 4000Da molecular weights [9,10]. It binds mainly to hydrophobic compounds and expels them outside the cell [11].
Effects of a variety of compounds on Pgp, such as oroxyin A [12], fluorescent compounds [13], selective serotonin reuptake inhibitors [14], mollugin [15], antipsychotic drugs [16] etc. have been reported. Structures of several bacterial ABC transporters have been determined [17][18][19][20][21][22], although only low resolution structure of mouse Pgp is available [23][24][25]. Mouse Pgp serves as a good model as it has a significant similarity with human Pgp [26]. It is a 170 kDa transmembrane protein, expressed in various resistant tumour cell membranes. As it is involved in MDR in tumour cells, designing of inhibitors to Pgp may have a great therapeutic potential.
Pgp has similar topology as other members of the ABC transporter family including MRP1, MRP2, MRP3, BCRP proteins and plays similar role in xenobiotic metabolism. In vivo and in vitro studies have indicated that flavones and isoflavones inhibit Pgp and other ATP transporters mediated exotransport of anticancer drugs and other xenobiotic compounds. These transporters were over expressed and effects of flavones and isoflavones were investigated by monitoring exotransport of various known substrates and compounds [48][49][50][51]. Although interactions of a large number of natural compounds have been extensively studied with Pgp and MRP1 [52][53][54][55][56], the details of the binding pattern of isoflavones to Pgp was not understood so far. Since the three-dimensional crystal structure is available only for mouse Pgp and the sequence of mouse Pgp share 87% identity with the human form, and almost 100% identity with residues present in binding cavity [26], therefore we performed in silico studies to identify interactions between isoflavones and mouse Pgp.
Such knowledge can provide further scope of designing potent and specific inhibitors of Pgp, which may have a good therapeutic potential. Isoflavones binding studies may play a vital role in finding hotspots in Pgp and eventually may be proved useful in designing compounds with high affinity and specificity to the protein, which can be used along with anti cancer chemotherapeutic drugs. This manuscript provides details of interactions between eleven isoflavones varying in their chemical nature and affinity with Pgp, which has helped in understanding the precise sites and the functional groups involved in inhibitor recognition.

Docking Isoflavones on Pgp
The atomic coordinates of 11 isoflavones (diadzein, daidzin, formononetin, ononin, genistein, genistin, biochanin A, sissotrin, coumestrol, coumestrin, glycitein) were generated using structures described earlier [51]. 3-D atomic coordinates for isoflavone structures were generated using LIGPREP [57] with the correct chiralities of atoms. The resulting geometries were optimized by molecular mechanics using IMPACT [57] in a dynamic environment using standard TIP4P water model. Energy minimization was done using Polak-Ribier conjugate gradient algorithm using Optimized Potentials for Liquid Simulations 2005 (OPLS 2005) force field. RMS gradient of 0.01 was used as the convergence threshold. Conformational models of all the ligands were generated. The ligands were prepared using LIGPREP [57]. The structures with correct bond length, bond angles, and conformation, stereochemistry and ionization states were selected for docking studies. All the isoflavones except coumestrol and coumestrin have identical basic structure, which have variations in groups substitute at C7, C8, C9 and C14 positions on the nucleus structure (Table 1). Docking studies were performed using GLIDE [58]. Atomic coordinates of the native receptor were extracted from protein data bank (PDB: 3G5U). The submitted coordinates were corresponding to a homo dimer of Pgp. Monomer (Chain A) of the Pgp was used for study. Each monomer contained 1271 amino acid residues and 18521 non hydrogen atoms. The structure of Pgp complex with a cyclic peptide is also available as PDB code 3G60 in protein databank. Pre-processing of the native transporter structure was done by assigning correct bond order, addition of hydrogen atoms, capping uncapped termini, adjusting bonds and formal charges for groups and correcting mislabeled elements. All docking studies were carried out using extra precision (XP) method of GLIDE (Grid-Based Ligand Docking with Energetic) which examines the complementarities of ligand-receptor interactions using a grid-based method based on the empirical Chem Score function for flexible ligand docking. The centroid of selected active site residues (using PDB: 3G60) were used for receptor grid generation for docking as well as blind docking. 1000 docked poses per compound were generated for the different conformations of the compounds. The affinity grid maps centered on the peptide binding site had dimensions of 80 × 80 × 80Å with 0.375Å spacing between grid points. The binding free energy of the receptor-ligand binding (ΔG bind ) is calculated by taking the difference between the free energies of the receptor-ligand complex (ΔG complex ) and the unbound receptor (ΔG receptor ) and ligand (ΔG ligand ).
The structures of docked complexes of Pgp with isoflavones were analyzed using "O" program [59] and COOT [60]. The surface potential of the Pgp was calculated using PyMol [61].

Structure of the Pgp Model
Pgp is a transmembrane pump which effluxes drugs and other xenobiotic compounds outside the cell and is responsible for MDR. A crystal structure was reported for mouse Pgp [25]. The PDB coordinates were available for native Pgp (PDB id: 3G5U) and complex with a cyclic hexapeptide inhibitor (PDB id: 3G61) in protein data bank (PDB). The native structure of Pgp was determined at 3.8Å resolution. Pgp has a helical bundle rich fold with 56% αhelical and 7% β-strand secondary structure content. The structure shows a crystallographic dimer having 1284 amino acid in each monomeric unit. Structure of Pgp can be divided into two highly homologous parts, each half of the structure contains six membrane spanning αhelices in form of transmembrane domain (TMD), followed by a nucleotide binding domain (NBD). So the complete molecule contains twelve long -helices in form of TMD and two NBD regions.

The Ligand Binding Cavity
Pgp is a tandem protein, comprising of two identical homologous halves, containing total 1271 amino acids. . Each consecutive α-helix is connected through flexible loop regions which are capable of making the helices move apart while performing its functions. The wall of the ligand binding site is formed by the contribution of residues from these 12 α-helices. Furthermore, hydrophobic stacking of these α-helices makes them more amenable to opening and closing of the interstitial space. The ligand binding site is present in form of a wide and long hydrophobic cavity. The cavity is predominantly formed by hydrophobic amino acids. Aromatic residues are most abundant in the cavity and form a layer of aromatic amino acids. Such a layer has also been reported in modeled structure of MRP1 [62]. Indeed, some compounds have also been found to be common inhibitors of both Pgp and MRP proteins [63].  . 1). The remaining part of sequence forms two ATP binding domains which are α-β bundle. ATP binding domains are located from 378-626 and 1018-1271 amino acid residues. The cavity formed is very big in size and has a total volume of approximately 6000Å 3 [25].

Binding of Isoflavones
The binding affinity of isoflavones and Pgp was measured in terms of docking score and the energy of the complex formation. Docking scores distinguish between molecules that bind strongly in their optimal placement from those that bind weakly to the transporter. The 426 ligand binding site is present in form of a wide and long hydrophobic cavity. The cavity is predominantly formed by hydrophobic amino acids. Aromatic residues are most abundant in the cavity and form a layer of aromatic amino acids. Such a layer has also been reported in modeled structure of MRP1 [62]. Indeed, some compounds have also been found to be common inhibitors of both Pgp and MRP proteins [63].  (Fig. 1). The remaining part of sequence forms two ATP binding domains which are α-β bundle. ATP binding domains are located from 378-626 and 1018-1271 amino acid residues. The cavity formed is very big in size and has a total volume of approximately 6000Å 3 [25].

Binding of Isoflavones
The binding affinity of isoflavones and Pgp was measured in terms of docking score and the energy of the complex formation. Docking scores distinguish between molecules that bind strongly in their optimal placement from those that bind weakly to the transporter. The 426 ligand binding site is present in form of a wide and long hydrophobic cavity. The cavity is predominantly formed by hydrophobic amino acids. Aromatic residues are most abundant in the cavity and form a layer of aromatic amino acids. Such a layer has also been reported in modeled structure of MRP1 [62]. Indeed, some compounds have also been found to be common inhibitors of both Pgp and MRP proteins [63].  (Fig. 1). The remaining part of sequence forms two ATP binding domains which are α-β bundle. ATP binding domains are located from 378-626 and 1018-1271 amino acid residues. The cavity formed is very big in size and has a total volume of approximately 6000Å 3 [25].

Binding of Isoflavones
The binding affinity of isoflavones and Pgp was measured in terms of docking score and the energy of the complex formation. Docking scores distinguish between molecules that bind strongly in their optimal placement from those that bind weakly to the transporter. The extensive ligand binding cavity of Pgp is responsible for binding of a wide variety of hydrophobic ligands. H3 and H9 helices are present on the peripheral face of the molecule and seem to be playing no direct role in ligand binding, whereas remaining ten helices have protruding-in side chains of amino acids, which are responsible for formation of hydrophobic cavity. In present study, due to structural and chemical variations isoflavones occupy two different but overlapping sites in the hydrophobic cavity, which are formed by six α-helices, H4, H5, H6, H7, H8 and H12. These sites are designated as site 1 and site 2 for the ease of description in the manuscript. Sissotrin, coumestrin and formononetin bind to site 1 whereas biochanin A, genistein, genistin, diadzein, diadzin, glycitein, coumestrol and ononin bind to site 2 in the wide hydrophobic cavity. Glycosides isoflavones have higher binding affinity to Pgp than the aglycones, as indicated by their docking scores, which can be attributed to their polar nature of glycosides by presence of glucose moiety, as indicated by relative energy, docking scores and buried surface area of the ligands ( Table 1). The glycosides isoflavones have docking score and energy of the complex ranging from -7.9 to -7.6 and -64.9 to -55.8 kcal/mol respectively, whereas the docking score and energy of complex in case of aglycones ranged from -7.3 to -4.8 and -48.0 to -36.4 kcal/mol respectively.

Site 1: Binding of sissotrin, coumestrin and formononetin
These three compounds bind to Pgp with almost identical manner at a common site. In case of sissotrin and coumestrin 30 poses were obtained (docking score ranging from -7.35 to -6.34 for different poses in case of both molecules). The analysis of complexes revealed that their binding site is formed mainly by side chains of residues Leu 300, Tyr 303, Phe 332, Leu 335, Phe 339, Phe 724, Phe 766, Phe 974, Leu 978 and Phe 990. Sissotrin, coumestrin and formononetin are stacked in this hydrophobic cavity though van der Waals interactions. Along with these hydrophobic forces, sissotrin and coumestrin shared common hydrogen bonds with the amide side group of Asn 717 and hydroxyl side group of Ser 989 whereas, formononetin made a hydrogen bond with the side chain of Tyr 303. In addition coumestrin also makes one hydrogen bond with the side amide group of Asn 721 (Fig. 2a).

Site 2: Binding of biochanin A, genistein, daidzein, daidzin, genistin, glycitein, coumestrol and ononin
Only one optimal pose was obtained in this category of isoflavones except for genistin (50 poses) and ononin (30 poses). The average docking scores for these are shown in Table 1.
The docking score of different poses for genistin ranged from -7.5 to -6.5 and for ononin, it ranged from -7.7 to -6.7. Site 2 is located slightly overlapping and almost perpendicular to the site 1. It recognizes the above mentioned eight isoflavones in the hydrophobic cavity.  (Fig. 2b).

Comparison with the Binding of Cyclic Peptide Inhibitor
Pgp complex with a cyclic peptide has also been determined and deposited to the PDB database with accession code 3G60 [25]. The inside of hydrophobic cavity has a gradient of polarity and hydrophobicity. The cytosolic face (towards NBD) has higher occurrence of charge and polar residues, whereas the relative number of hydrophobic residues increases as we go deeper in the cavity towards extracellular side (Fig. 3). Isoflavones occupies two distinct but overlapping sites on Pgp structure as discussed in previous section. The cyclic hexa peptide (AV) 3 binds at site 2 of Pgp and therefore it share binding space with biochanin A, genistein, daidzein, daidzin, genistin, glycitein, coumestrol and ononin isoflavones. The peptide has bound slightly deeper in the site 2 of the hydrophobic cavity which may be attributed to its higher hydrophobic nature and easier diffusion in the cavity, as the cavity has relatively more hydrophobic residues towards extracellular side, to prefer similar compounds (Fig. 3).

Comparison with the Binding of Cyclic Peptide Inhibitor
Pgp complex with a cyclic peptide has also been determined and deposited to the PDB database with accession code 3G60 [25]. The inside of hydrophobic cavity has a gradient of polarity and hydrophobicity. The cytosolic face (towards NBD) has higher occurrence of charge and polar residues, whereas the relative number of hydrophobic residues increases as we go deeper in the cavity towards extracellular side (Fig. 3). Isoflavones occupies two distinct but overlapping sites on Pgp structure as discussed in previous section. The cyclic hexa peptide (AV) 3 binds at site 2 of Pgp and therefore it share binding space with biochanin A, genistein, daidzein, daidzin, genistin, glycitein, coumestrol and ononin isoflavones. The peptide has bound slightly deeper in the site 2 of the hydrophobic cavity which may be attributed to its higher hydrophobic nature and easier diffusion in the cavity, as the cavity has relatively more hydrophobic residues towards extracellular side, to prefer similar compounds (Fig. 3).

Comparison with the Binding of Cyclic Peptide Inhibitor
Pgp complex with a cyclic peptide has also been determined and deposited to the PDB database with accession code 3G60 [25]. The inside of hydrophobic cavity has a gradient of polarity and hydrophobicity. The cytosolic face (towards NBD) has higher occurrence of charge and polar residues, whereas the relative number of hydrophobic residues increases as we go deeper in the cavity towards extracellular side (Fig. 3). Isoflavones occupies two distinct but overlapping sites on Pgp structure as discussed in previous section. The cyclic hexa peptide (AV) 3 binds at site 2 of Pgp and therefore it share binding space with biochanin A, genistein, daidzein, daidzin, genistin, glycitein, coumestrol and ononin isoflavones. The peptide has bound slightly deeper in the site 2 of the hydrophobic cavity which may be attributed to its higher hydrophobic nature and easier diffusion in the cavity, as the cavity has relatively more hydrophobic residues towards extracellular side, to prefer similar compounds (Fig. 3).

Comparison with the Binding of Flavones and other Ligands
It was observed that phenothiazines interact with drug binding sites in TM region, whereas flavones bind to NBD sites of Pgp and MRP1 proteins [54]. Recently, docking studies of flavones on Pgp has also been reported [64]. In contrast to isoflavones, flavones have been found to be indeed interacting to NBD of Pgp. In the current study, isoflavones bind to the hydrophobic cavity in TM region whereas, flavones could be modulating Pgp mediated exotransport of drugs by inhibiting the nucleotide binding. One more study has also shown the substrate and inhibitor binding sites in Pgp [65] by docking of berberine (BBR) as substrate and HIV protease inhibitors (HIVPI) as inhibitors of Pgp. BBR binds to a deeper site in comparison to HIVPIs to Pgp, so that binding of BBR get competitively inhibited by HIVPIs. When compared, it was observed that isoflavones also bind to upper part of the binding cavity like HIVPIs, which is close to the cyclic peptide inhibitor binding site as well [25]. Cytotoxic drugs and other inhibitor molecules also have been found to be interacting at two distinct sites although for modulating ATPase activity [66] whereas isoflavones inhibit ligands binding activity of Pgp.

CONCLUSIONS
Pgp contains a wide cavity for binding to relatively hydrophobic and aromatic ligand molecules. The modeled human Pgp structure has shown very good similarity with the mouse Pgp [67]. The hydrophobic cavity is formed by contribution of side chains of mainly Phe, Leu and Ile residues from six α-helices in the Pgp structure. Isoflavones recognizes two slightly distinct but overlapping sites on Pgp called site 1 and site 2. Sissotrin, coumestrin and formononetin recognize interacting groups in site 1 whereas biochanin A, genistein, daidzein, daidzin, genistin glycitein, coumestrol and ononin bind to site 2 in the wide hydrophobic cavity present in Pgp. Site 1 and site 2 are located differently in the hydrophobic cavity and also contain a overlapping space as well (Fig. 4). Both the sites have hydrophobic amino acids to form van der Waals interactions with the ligands. In addition to hydrophobic residues, site 2 also contains one polar face from helix H2, which is responsible for forming several hydrogen bonded interactions with the ligands. A gradient of polar and hydrophobic amino acid residues is present inside the ligand binding cavity of Pgp. The cytosolic side of the cavity is very polar and charged whereas the hydrophobic environment starts increasing towards extracellular side of the cavity (Fig. 3). Site 1 has minimum dimension as small as 7.5Å and maximum as 22Å whereas site 2 is wider and deeper than site 1. Therefore, Pgp is capable of binding a variety of substrates, although hydrophobic compounds bind deeper in the cavity whereas relatively polar compounds bind towards cytosolic side, which is also evident from deeper binding of BBR to Pgp [65]. Hence, the cyclic (AV) 3 peptide being highly hydrophobic binds in the deeper region of site 2 in comparison to the isoflavones. These features may be utilized to construct compounds which have hydrophobic and polar groups and can occupy both the sites to form stronger attractive interactions. As the movement of helices are responsible for transporting drugs outside the cell, bigger compounds which can occupy most of the area of the cavity including site 1 and 2 may form an effective inhibitor of Pgp. In addition, presence of polar groups in the aromatic/hydrophobic ligand may enhance its binding affinity for Pgp.