Exploration of the conformational landscape in pregnane X receptor reveals a new binding pocket

Abstract

Ligand‐regulated pregnane X receptor (PXR), a member of the nuclear receptor superfamily, plays a central role in xenobiotic metabolism. Despite its critical role in drug metabolism, PXR activation can lead to adverse drug‐drug interactions and early stage metabolism of drugs. Activated PXR can induce cancer drug resistance and enhance the onset of malignancy. Since promiscuity in ligand binding makes it difficult to develop competitive inhibitors targeting PXR ligand binding pocket (LBP), it is essential to identify allosteric sites for effective PXR antagonism. Here, molecular dynamics (MD) simulation studies unravelled the existence of two different conformational states, namely “expanded” and “contracted”, in apo PXR ligand binding domain (LBD). Ligand binding events shifted this conformational equilibrium and locked the LBD in a single “ligand‐adaptable” conformational state. Ensemble‐based computational solvent mapping identified a transiently open potential small molecule binding pocket between α5 and α8 helices, named “α8 pocket”, whose opening‐closing mechanism directly correlated with the conformational shift in LBD. A virtual hit identified through structure‐based virtual screening against α8 pocket locks the pocket in its open conformation. MD simulations further revealed that the presence of small molecule at allosteric site disrupts the LBD dynamics and locks the LBD in a “tightly‐contracted” conformation. The molecular details provided here could guide new structural studies to understand PXR activation and antagonism.

Abbreviations

AF‐2

activation function‐2

FEL

free energy landscape

LBD

ligand binding domain

LBP

ligand binding pocket

MD

molecular dynamics

PCA

principle component analysis

PSNs

protein structure networks

PXR

pregnane X receptor

RXR

retinoid X receptor.

Introduction

Ligand‐regulated transcription factors of nuclear receptor (NR) superfamily control the expression of genes essential to a variety of biological processes including reproduction, development and metabolism.1, 2, 3 Pregnane X receptor (PXR), a member of NR superfamily, plays a key role in the detection and metabolism of different endogenous and exogenous compounds and helps in protecting the body from potentially toxic chemicals.4 Upon binding to different compounds, activated PXR upregulates the expression of a large array of genes encoding proteins of central role in xenobiotic metabolism. The growing list of genes include different alleles of cytochrome P450 (CYP 3A, 2C, 1A), glutathione‐S‐transferases, drug efflux pumps MDR1 (multi‐drug resistance gene‐1) and MDR2 etc.5

In contrast to other members of NR superfamily, PXR binds promiscuously to a large number of chemically and geometrically distinct compounds that include naturally occurring steroids, hormones, and bile acids, as well as exogenous chemicals like pesticides, herbal extracts, and pharmaceutical products.1 Like other NR family members, PXR possesses a DNA binding domain (DBD) and a ligand binding domain (LBD). Human PXR LBD is a three layer α‐helical sandwich (α1/α3, α4/α5/α8, and α7/α10–Fig. 1) with unique modification, unlike in several other NRs. The ligand binding pocket (LBP) of LBD features an extended five‐strand β‐turn‐β motif instead of the commonly observed two‐ to three‐strand motif. This unique 50–60 residue insert, which forms the novel PXR homodimer interface, provides an additional flexibility to the pocket that helps PXR to respond promiscuously to ligands of different size and shape.6, 7 It is shown that the homodimer interface is essential for PXR interaction with coactivators that bind at the activation function‐2 (AF‐2) surface, but does not impact PXR's association with activating ligands.5 Ligand‐activated PXR works in complex with retinoid X receptor alpha (RXRα). This PXR‐RXR‐coactivator complex binds to the promoter region of target genes and regulates the downstream transcription machinery.7

Figure 1.

Human PXR LBD in complex with colupulone bound to the LBP, highlighting different regions of interest. The α‐helices are shown in magenta and β‐strands are in cyan. PXR heterodimerizes with RXR through the interface formed at α5, α9 and α10 helices. The AF2‐surface, formed with αAF (known as activation function‐2 helix), that facilitates the interaction with transcriptional coactivators and corepressors, is also labeled. The flexible loop region between α1 and α3 helices is highlighted in green color.

Despite its protective role in detoxification, ligand‐activated PXR has been implicated in mediating a number of clinically significant adverse drug–drug interactions. Co‐administration of hyperforin, a potent PXR agonist, with irinotecan has shown to significantly reduce the anti‐neoplastic efficacy of the latter.8 PXR is shown to induce cancer drug resistance and tumour growth by upregulating the expression of enzymes that can affect chemotherapy metabolism. Recent studies have revealed that the chronic administration of PXR activators can also affect the metabolism of steroid hormones and fat‐soluble vitamins. Ligand activated PXR has also been implicated in the development of hypertriglyceridemia, nonalcoholic steatohepatitis and drug toxicity. Thus, due to their key regulatory role in different pathophysiological conditions, inhibitor designing for PXR has gained enormous pharmaceutical importance in the recent past.3, 4, 5

Even though studies have been carried out to design inhibitors based on structural information from PXR‐ligand complexes, promiscuous nature of the protein has made it difficult to develop an effective antagonist that targets the LBP.9 In recent years, research has been carried out looking for regions outside LBP as a promising pharmaceutical target, particularly AF‐2 surface and homodimerization interface.5 After it was first identified as a PXR antagonist, the anti‐fungal compound ketoconazole has been employed extensively in studies to understand the mechanism of PXR antagonism.10 In vitro and in silico studies suggested that ketoconazole potentially binds at the AF‐2 surface and disrupts the coactivator binding.11, 12 Pharmacophore modelling has also provided insightful information for designing less‐toxic and more‐specific PXR antagonists, which helped in developing more potent azole derivatives like SPB3255.13 Recently, Mani and group have employed a novel yeast two‐hybrid strategy to identify a new ketoconazole binding site in addition to the AF‐2 surface that is located in the vicinity of residue Ser208.14 This study suggested a new mechanism of PXR antagonism where ketoconazole potentially interferes with agonist binding and/or homodimerization. The steadily increasing list of PXR antagonists includes ET‐743, camptothecin, fluconazole, enilconazole, metformin, sesamin, fucoxanthin, coumestrol, resveratrol etc.5, 15 Although there is a large array of diverse chemical entities that exhibits PXR antagonist properties, it remains unclear where they bind and how they exert action on PXR. A dynamic view of ligand binding events in PXR will indeed enhance our understanding of the promiscuous nature of PXR and will be helpful in exploring new potential small molecule binding hot spots.

The primary focus of this paper is to capture the dynamic behaviour of PXR LBD in its apo and ligand‐bound state. Here, we carried out extensive all‐atom molecular dynamics (MD) simulations of LBD in its ligand‐bound and ligand‐free states. Analysis of essential dynamics and pocket volume has revealed that the PXR LBD exists in dynamic equilibrium with two different conformational states–“expanded” and “contracted”. Ligand binding restricts the LBD dynamics and shifts this equilibrium towards one of the “ligand‐adaptable” states depending on the size and shape of the bound ligand. Computational fragment‐based mapping has identified a potential small molecule binding site, whose opening‐closing mechanism directly correlates with the conformational switch between the expanded and contracted states. Interestingly, presence of small molecule at the newly identified pocket has impaired the PXR LBD dynamics. To the best of our knowledge, this is the first report that has unveiled the dynamic picture of promiscuity in PXR and has explored a new potential small molecule binding pocket.

Results and Discussion

In order to understand the dynamics associated with ligand binding to PXR, we performed all atom MD simulations of PXR LBD in its apo and liganded states. Simulations were carried out for 100 ns of production run following 10 ns equilibration at 300 K. Different systems simulated in this study are presented in Table 1 and detailed description of system preparation is provided in the Methodology section.

Table 1.

Details of the PXR LBD Systems Simulated.

System	Orthosteric liganda	Allosteric ligandb	PDB IDc
1	–	–	1ILG [A]6
2	Colupulone	–	2QNV [A]23
3	Hyperforin	–	1M13 [A]24
4	Rifampicin	–	1SKX [A]16
5	SR12813	–	1NRL [A]17
6	T1317	–	2O9I [A]9
7	–	NSC1014

^aFor system 1‐6, missing loop regions were modelled with Modeller package.39

^bFor system 7, conformation of the PXR LBD with open α8 pocket was picked up from apo simulation trajectories (system 1, see the methodology section for details). The compound identifier “NSC1014” in the NCI Diversity Set IV represents the molecule 4‐methyl‐2‐pyridyl aminobenzyl‐8‐quinolinol.31

^cChain ID of the structure used in simulations is shown in squared brackets.

The root‐mean‐square deviation (RMSD) of LBD in system 1 with respect to the minimized starting structure as a function of simulation time indicates that the long loop region (residues 178‐197, Fig. 1) is comparatively more flexible than the rest of the protein (Fig. S1a‐b), as expected and observed from earlier studies.6, 16, 17 This is further supported by residue‐wise root‐mean‐square fluctuations (RMSF) (Fig. S2). Similar observations are also made from the simulations of liganded systems (systems 2–6, Table 1), suggesting that the ligands do not cause any major structural changes. Although the RMSD does not reflect any notable conformational change in the PXR LBD, subtle changes do occur, distinguishing the conformational dynamics of the apo and the liganded forms. Such differences captured by analysing some of the features from the simulations of systems 1–6 are discussed below.

Conformational flexibility in PXR LBD

Apo LBD exhibits conformational equilibrium with multiple states

A major part of the dynamics is represented by the top essential modes (principal components), which are identified through principal component analysis (PCA) of the simulation trajectories. The top two principal components (PCs) have captured significant motions of PXR LBD from each simulation ensemble (Fig. S3). The motions in apo LBD along the direction of PC1 and PC2 are presented as porcupine plots18 in Figure 2(A) and 2(B), respectively. It can be seen that an expanding/contracting motion of the LBP is the dominant movement along these modes. The secondary structural elements, β1 and β1“ strands, β1/β1” loop, α1/α2 loop, α2 and part of α10 helices, that forms the lower part of LBP (refer Fig. 1) are involved in this movement. The β1 and β1“ strand, β1/β1” loop and C‐terminal portion of α1/α2 loop move as a single entity in a direction opposite to the movement of α2, C‐terminal of α10, and N‐terminal portion of α1/α2 loop (Movie S1). It is to be noted that most of these secondary structural elements (region between α1 and α3, Fig. 1) are part of the unique ∼60 residues insert, one of the distinct features of PXR.6 Structural studies in the past have shown that this unique insert provides an additional flexibility to the PXR LBP, which helps PXR to bind promiscuously to a wide variety of ligands. The observed fluctuations of the unique insert between α1 and α3 in contrast to the interior of LBP were also noticed in an earlier study.17

Figure 2.

Dynamics of PXR LBD in apo state. (a) and (b) show the porcupine plots representing the principal motions along the direction of PC1 and PC2 respectively. Secondary structural elements which are involved in expanding/contracting motions are highlighted: α1/α2 loop (orange), α2 helix (magenta), β1‐β1' sheet and β1/β1' loop (blue) and C‐terminal of α10 helices (yellow). The α7, α8 helices and α7/α8 loop that form the part of “α8 pocket” are shown in red (discussed in section 2 of Results and Discussion). (c) A plot of varying LBP volume as a function of simulation time. The yellow solid line shows the LBP volume in apo crystal structure (PDB ID: 1ILG). (d) PCA based free energy (in kJ/mol) landscape using PC1 and PC2 as reaction coordinates. The low energy basins are marked with “E#”. Average structure from each basin with corresponding average LBP volume, with standard deviation, is shown alongside. The cliques due to side‐chain interactions that are specific to a given population are also projected on the average structure.

The expanding/contracting motion is further supported by the varying volume of LBP during simulation of the apo LBD (system 1). The pocket volume analysed along the MD trajectories of system 1 is plotted in Figure 2(C). Notably, LBP visits a wide range of volume during the course of simulation. The pocket takes a volume as low as ∼900 ų and also as high as ∼2000 ų, accessing conformations of contracted and expanded states respectively. This clearly indicates that LBP volume in the apo state explores both the larger (about 2000 ų) and smaller (about 900 ų) values than the volume of 1176 ų in apo crystal structure (PDB ID: 1ILG), with an average volume of about 1300 ų. It is interesting to note that the simulation has identified several conformational states of apo PXR LBD with distinct LBP volume.

The conformations accessed by PXR LBD during simulation have been further characterized by the free energy landscape (FEL) constructed along the first two PCs (PC1 and PC2). Although such a landscape is not sufficiently accurate to describe the metastable states and free energy barriers, it can be used to describe the conformations accessed by the molecule from the MD ensemble.19, 20 As can be seen from Figure 2(D), several low energy basins are accessed during the simulation of system 1. Representative average structures, from each energy basin are also depicted alongside in the figure. It is indeed remarkable to observe that the conformations in each basin correspond to distinct LBP volume. Notably, six minima depicted in Figure 2(D) (E1–E6) clearly shows that the landscape indeed has sampled the expanding/contracting movements in apo LBD. Specifically, E1 depicts the conformation with highest LBP volume (1885 ± 173 ų) while the basin E3 shows the lowest volume (928 ± 132 ų). Other basins show the conformations with LBP volume ranging from 1000 ų to 1300 ų.

Earlier studies on other protein structures have shown that the network parameters such as cliques derived from the network of side chain interactions analysed through protein structure networks (PSN) can effectively capture subtle differences in conformations.21 In the present study, cliques have been mapped on the conformations from different energy basins in Figure 2(D). Further, a comparison of the similarities and differences in terms of commonality in cliques are presented in Table S1. For example, we can see that 17 cliques are common between E1 and E2, whereas 10 and 2 cliques are unique to E1 and E2, respectively. Thus, the LBP volume, FEL and PSN analyses have revealed that PXR LBD in its apo form exhibits different conformational populations, which are easily inter‐convertible during simulation. These conformations are not drastically different in the overall topology as seen from RMSD values, however, are distinctly different in their structural properties due to differences in their side chain interactions. Here, we designate the conformations with highest and lowest LBP volumes as “expanded” and “contracted” states respectively. The effect of ligand binding on conformational flexibility is discussed in the next section.

Ligand locks the conformational flexibility

PXR has evolved to bind promiscuously to a wide variety of compounds with different chemical and geometrical nature. In the absence of gross structural changes as noted above, it is likely that ligand binding brings out subtle changes in conformations and their equilibrium populations.21, 22 Here, we have carried out extensive MD simulations of ligand‐bound PXR LBD (systems 2–6 in Table 1) to bring out the ligand induced subtle changes in LBD dynamics. The LBD structures co‐crystalized with ligand of varying geometry and chemical natures, picked from protein data bank were used in this study. Systems 2‐6 correspond to the LBD bound to different ligands; colupulone,23 hyperforin,24 rifampicin,16 SR12813,17 and T1317⁹ (chemical structures given in Fig. S4). The correlated motions associated with ligand binding were explored from the MD ensembles of these simulations, through similar analyses as described above for the apo system. The porcupine plots presented in Figure 3(a,b) show the motions along PC1 and PC2 in colupulone‐bound PXR LBD. Similar porcupine plots for other ligand‐bound LBD are shown in Figure S5‐S8 a‐b. The figures clearly demonstrate that the entire LBP moves as a single entity in all the liganded systems, in contrast to the expanding/contracting motions exhibited in system 1. The characteristic movements of the structural segments–β1 and β1“ strands, β1/β1” and α1/α2 loops, and α2 and α10 helices–observed in the apo state have significantly reduced in presence of ligands (Movie S2). Furthermore, in all the ligand‐bound systems overall motion of the protein is also restricted. Thus, the PCA indicates that ligand binding results in reduced dynamics of PXR LBD, as opposed to multiple states accessed in the apo form.

Figure 3.

Dynamics of ligand‐bound PXR LBD. (a) and (b) show the porcupine plots representing the principal motions in colupulone‐bound PXR LBD, along the direction of PC1 and PC2 respectively. Secondary structural elements highlighted are the same as in Figure 2a‐b. (c) A plot of varying LBP volume as a function of simulation time in different ligand‐bound systems (black–apo, red–colupulone, green–hyperforin, blue–rifampicin, cyan–SR12813, and magenta–T1317). The yellow solid line shows the LBP volume in apo crystal structure (PDB ID: 1ILG). (d) PCA based free energy (in kJ/mol) landscape using PC1 and PC2 as reaction coordinates in colupulone‐bound PXR LBD. The low energy basins are marked with “E#”. Average structure from each basin with corresponding average LBP volume, with standard deviation, is shown alongside. The cliques due to side‐chain interactions that are specific to a given population are also projected on the average structure. Results from PCA analyses and FEL for other liganded systems are shown in supporting Figures S5‐S8.

The calculated volume of LBP further provides additional information about the restricted dynamics in ligand‐bound PXR LBD. The fluctuations in LBP volume in all the ligand‐bound systems have significantly decreased as compared to the apo system [Fig. 3(C)]. Markedly, the average LBP volume is correlated with the geometrical nature of the bound ligand. For instance, the average LBP volume in rifampicin (largest ligand used in this study, ∼800 Da) and T1317 (smallest ligand) bound systems are 1800 ų and 910 ų, respectively. The FEL analyses of the ligand‐bound systems also revealed the confined dynamics in PXR LBD. The landscape of colupulone bound LBD [Fig. 3(D)] indicates that the low‐energy conformations are limited in comparison with that of the apo system [Fig. 2(D)]. Similar observations were made with other liganded systems (Fig. S5‐S8c) as well. The restricted dynamics in liganded systems is clearly evident from the LBP volume accessed by conformations from low energy basins. For instance, the LBP volume is in a narrow range of 1200 ų−1500 ų in the conformations present in all the low‐energy basins of colupulone bound LBD [Fig. 3(D)], in contrast to a wide range (900 ų −1800 ų) observed in the apo state. Similarly, the conformations from low energy basins of rifampicin bound LBD takes up volume of 1800 ų to 2100 ų (Fig. S6c).

The effect of ligand binding to a protein can result in major or minor structural changes. It is often difficult to capture the subtle changes by conventional methods of analysis. However, as mentioned above, PSN captures such subtle changes through network parameters.21 The characterization of conformations by the clique architecture has been carried out on ligand‐bound PXR systems (systems 2‐6). The identified cliques mapped on the conformations from different low‐energy basins of system 2 are depicted in Figure 3(D) and the corresponding ones for systems 3‐6 are presented in Figure S5‐S8c. Additionally, the clique analysis showing the extent of structural overlap/difference between conformations from various basins within and across the simulated systems are presented in Table S1. The entries in the table show that the conformations within a system is closer to each other with more common cliques, whereas they differ significantly across the systems with reduced commonality. Furthermore, it is seen that the commonality in cliques from different basins within liganded systems is much more than that seen in apo system. These comparisons indicate that the conformations at the network level are more uniform in liganded systems, whereas they are diverse in the apo state, confirming the results obtained from PCA and pocket volume analysis. The size of the largest community in a structure is also an indicator of the protein rigidity/flexibility. Communities of non‐covalent interactions have been shown to stabilize the protein structures and also to modulate the ligand binding regions of the protein.25 Dynamically stable communities in apo and ligand‐bound systems are shown in Figure 4. In contrast to the apo system, larger communities are formed in ligand‐bound systems where ligand becomes an integral part of the community. Strikingly, the ligands impart rigidity to LBD and lock the protein in a specific conformation in all the cases, although different regions of LBP get highly connected through different ligands.

Figure 4.

Dynamically stable communities (k−1/k−2) formed in apo (a), and ligand‐bound PXR LBD–colupulone (b), hyperforin (c), rifampicin (d), SR12813 (e), and T1317 (f). The communities show the ligand‐mediated interactions across the LBP, which lock the LBD conformation. In apo LBD, lack of communities in the LBP region provides additional flexibility to the LBP. Parts of the α8 pocket–α7, α8 helices and α7/α8 loop–are shown in magenta (α8 pocket is discussed in section 2 of Results and Discussion). The nodes (amino acids) that form the communities are shown as beads. In (b) – (f), centre of mass of the corresponding ligand molecule is represented as cyan color bead.

To summarise, the analyses of apo and liganded systems clearly indicate that PXR LBD accesses multiple conformations in the apo state and ligand binding shifts the equilibrium towards a single “ligand‐adaptable” conformation. These results are in agreement with the hypothesis that ligand binding stabilizes PXR and increases half‐life of the protein.17 The present study sheds light on the molecular mechanism of promiscuous binding of structurally diverse compounds to PXR. In apo state, LBP shows expanding/contracting motions where the pocket volume fluctuates between ∼900 ų–∼2000 ų. It is observed that the α1‐α3 region, a ∼60 residue insert unique to PXR LBD,6 and α10 helix are involved in this expanding/contracting motion. The α1‐α3 stretch has a high percentage of loop regions in PXR, whereas it is short and well‐structured in several members of NR family (Fig. S9). Notably, the α2 helix in peroxisome proliferator‐activated receptor alpha (PPARα), which corresponds to the α1‐α3 region in PXR, has been proposed as the ligand access site in the binding pocket.26 Our study has shown that the flexible α1‐α3 stretch is the major contributor to the contracting/expanding motion in PXR LBD and hence possibly the cause of its promiscuity, allowing ligands of varying size and shape to LBP. Once the ligand enters the LBP, it freezes the LBD in a “ligand‐adaptable” conformation, which is appropriate for the size and shape of the bound ligand. Remarkably, the varying LBP volume reported for apo and ligand‐bound PXR LBD crystal structures well supports the finding of dynamic behaviour of LBD observed from the present study (Table S2).9, 16, 17, 23, 24 To the best of our knowledge, this is the first study to indicate that dynamical properties such as the conformational flexibility of the apo form and the rigidity attained by ligand binding, are the key factors in promiscuous ligand‐bound activation of the PXR.

Potential small molecule binding hot spots in PXR LBD

Recent studies have shown that a significant number of pathophysiological conditions have been mediated by PXR due to its promiscuity in ligand binding. Antagonizing PXR action has important clinical implications in preventing adverse drug‐drug interactions and improving therapeutic efficacy.3 Inhibition of PXR activity faces a greater challenge due to its promiscuous nature. Based on the structural details of PXR‐ligand interactions, studies have been carried out to design competitive antagonist for PXR. However, most of these designed antagonists have failed to inhibit the activity, instead several of them have ended up as PXR activators.9 In recent years, the possibilities of antagonist binding to sites that are distinct from the LBP are also being explored.5, 14 In this study, a large number of equilibrium conformations generated from molecular simulations have been used to explore possible small molecule binding sites in PXR LBD. Indeed, this approach has yielded promising results and the details of investigations are presented below.

Binding site mapping with MD ensemble representatives

The importance of dynamical properties such as alternate conformations, equilibrium populations, and energy landscapes of proteins for identifying druggable targets, is being explored in recent times.27, 28, 29 As we have seen from the above section, simulations have yielded several conformations of PXR LBD. Here, we explored these conformations for potential small molecule binding pockets. The pockets were first identified on the apo and the liganded crystal structures and then compared with that of the structures extracted from simulation trajectories.

Binding pocket mapping has been carried out for identifying small molecule binding hot spots using a fragment‐based FTMap algorithm.30 Recent studies have shown the potential of FTMap in exploring novel druggable targets.27, 28 In all the ligand‐bound structures, mapping was carried out after removing ligands from the pocket. As expected, mapping with apo and ligand‐bound crystal structures detected LBP as the top ranked consensus site (CS1) (Fig. S10). Moreover, most of the mapped CSs are present within the LBP. FTMap also mapped AF‐2 surface as one of the potential binding sites. These results show that FTMap is able to successfully identify the known low‐energy binding sites in experimentally determined LBD structures. It is noteworthy to mention that mapping with crystal structures do not exhibit any new promising binding hot spots.

Selected snapshots have been used to perform mapping exercise on conformations generated from simulations. Specifically, the reduced ensemble structures have been obtained from conformational clustering of MD trajectories (Fig. S11). Here, the centroids of the top fifteen clusters (described in Methodology) from each simulated system have been selected for mapping. The five highest ranked CSs obtained from the mapping of the reduced ensemble structures of apo and ligand‐bound PXR LBD are shown in Figure 5. Like in crystal structures (Fig. S10), conformations from MD ensembles also show LBP as one of the top ranked CSs. However, in simulated apo system [Fig. 5(A)], it is highly impressive to note that FTMap has identified new binding hot spots (CS1 and CS4) that exist outside the LBP, while the traditional LBP has become the second and third ranked sites (CS2 and CS3). From the chosen fifteen clusters, eight of them possess this hot spot. This newly identified small molecule binding pocket, which resides above the LBP, is lined by residues from α5, α7 and α8 helices, β2 strand and, α7/α8 and α5/β2 loops (Fig. 1). Interestingly, mapping of the ensemble representatives from ligand‐bound systems failed to locate the newly identified pocket as one of the potential binding hot spots [Fig. 5(B‐F)]. Thus, the mapping result from the apo simulation trajectories is in contrast to the results from crystal structures and the simulated liganded systems where LBP was identified as the prominent low‐energy CS.

Figure 5.

Mapping results of cluster centroids showing the spatial distribution of hot spots in apo (a) and ligand‐bound PXR LBD–colupulone (b), hyperforin (c), rifampicin (d), SR12813 (e) and T1317 (d). Top five consensus sites (CS) are shown in different colors–CS1, blue; CS2, red; CS3, green; CS4, orange and CS5, yellow. It is to be noted that in all the ligand‐bound PXR LBD, LBP is detected as the top ranked CS. In colupulone‐bound LBD (b), CS3 is located at the outer surface of LBP, between α1 helix and five‐strand β sheet. Moreover, CS2 and CS4 are found in the vicinity of AF‐2 surface. Excitingly, unlike in liganded systems, CS1 in apo system (a) is formed in a pocket located above LBP and the ranks of the pockets in the conventional LBP reduce to CS2 and CS3. The CS5 is found proximal to AF‐2 surface. For clarity, CSs are projected on the top five cluster centroid structures.

Correlated motions associated with α8 pocket

In order to delineate the appearance of newly identified pocket, we meticulously examined the dynamic behaviour of PXR LBD in apo and ligand states. As discussed in the earlier section, we have observed that the top two PC (PC1 and PC2) from apo PXR LBD simulation capture the expanding/contracting motions of LBP [Fig. 2(A‐B)] that involves the movement of α2 and α10 helices, β1 and β1' strands, β1/β1' loop and α1/α2 loop. Further, the expanding/contracting motion is also associated with the dynamics of α8 helix, α7‐α8 loop and C‐terminal portion of α7, where the newly mapped binding pocket is located. It is interesting to note from PCA that the upward/downward motion of α8 helix leads to the opening/closure of the identified binding pocket [Fig. 2(A‐B), Movie S1]. We referred this pocket as “α8 pocket” in subsequent sections. We have further confirmed the characteristic movement of α8 helix by following the time evolution of the angle between Cα atoms of residues Thr290 (α5), Lys331 (α7) and Glu337 (α8) [hereafter abbreviated as “angle TKE”, Fig. 6(A)]. This angle is used as a metric to capture the motion of α8 helix, which determines the opening‐closure of the α8 pocket. As depicted in Figure 6(B), in apo system, the angle visits its lowest value of 41^° and the highest value of 62^° during simulation. In accordance with the results from PCA, the angle calculation also indicates the upward/downward movement of α8 helix that leads to the opening/closure of the α8 pocket. Interestingly, the measured angle TKE value of 45^° in the apo crystal structure [PDB ID: 1ILG, yellow solid line in Fig. 6(B)] implies that the pocket is in its closed state. Hence, the FTMap algorithm failed to identify the α8 pocket as a potential binding site in apo crystal structure and the pocket appeared in the open conformation during the course of simulation.

Figure 6.

Time evolution of the angle between Cα atoms of Thr290 (α5), Lys331 (α7) and Glu337 (α8) (denoted as “angle TKE”) is used as a metric to capture the motion of α8 helix, which determines the opening‐closure of the α8 pocket. (a) The location of angle TKE residues in PXR LBD. (b) Time evolution of the angle TKE in different simulated systems. Color scheme is same as in Figure 3c. The yellow solid line represents the angle TKE value in apo crystal structure (PDB ID: 1ILG).

The above results suggest that the appearance of the α8 pocket is correlated with the expanding/contracting motion in apo PXR LBD. Notably, mapping with liganded crystal structures and ensembles of ligand‐bound simulations failed to identify the α8 pocket as one of potential CS [Fig. 5(B‐F), Fig. S10b‐f). This is primarily due to the restricted dynamics of the α8 pocket, as evident from PCA of ligand‐bound PXR LBD, where the motion of α8 helix, α7‐α8 loop and C‐terminal portion of α7 being constrained upon ligand binding [Fig. 3(A‐B), S5‐S8 a‐b, Movie S2). This clearly indicates that ligand binding regulates the opening/closure mechanism of α8 pocket.

The dynamics of α8 pocket is further examined by monitoring the angle TKE in different liganded systems. The time evolution of angle TKE in systems 2–6, is presented in Figure 6(B). Remarkably, the angle is smaller for all the liganded systems as compared to that of apo crystal structure and its simulated conformations (system 1). The average angle for all the ligand‐bound systems is found to be much smaller than that of the apo system. Together, this indicates that α8 helix moved downward to such an extent that α8 pocket goes to a tightly closed conformation upon ligand binding. This is due to the fact that the ligands interact with four residues–Glu321, Met323, Leu324 and His327–from α7 helix, which in effect restricts the upward movements of α7 and α8 helices (Fig. S12a). The strength of interaction of ligands with these residues however depends on the chemical and geometrical nature of the bound ligand (Fig. S12b).

In earlier sections, we had shown that the LBP volume reflects the expanding/contracting motion exhibited by PXR LBD (Figs. 2 and 3). In addition, the correlated dynamics of LBP and α8 pocket is also evident from the variations in pocket volume. A two dimensional plot of volume of LBP and α8 pocket from systems 1‐6 are shown in Figure 7. As pointed out earlier, the apo form has accessed a wide range of LBP volume (∼900 ų–∼2000 ų) and this fluctuation has reduced in presence of ligands. The volume occupied, however, depends on the nature of the ligand, for instance, it is about 910 ų and 1800 ų respectively for the complexes with T1317 and rifampicin. Strikingly, in apo system, the α8 pocket volume is generally greater than ∼250 ų and reaches about ∼500 ų, whereas in all the liganded systems this pocket volume has drastically reduced to values well below 200 ų. Thus, the presence of ligand in LBP restricts LBD to access the conformations with larger volumes of α8 pocket, which is in agreement with the mapping results of ligand‐bound LBD that failed to recognize this pocket. To sum up, the fragment‐based mapping has identified a new site, α8 pocket, in PXR LBD and the ligand binding at LBP leads to the closure of this pocket, whereas in apo system the open and closed states of the pocket exist in dynamic equilibrium.

Figure 7.

Correlated changes in LBP and α8 pocket volume in apo and liganded PXR LBD. Color scheme is same as in Figure 3c. It is to be noted from the figure that the α8 pocket volume is as high as 500 ų in apo LBD, whereas the ligand binding at LBP reduces the volume to values even lower than 100 ų, during the simulations. Also, the figure clearly shows the varying LBP volume depending on the size and shape of the bound ligand. Volume calculations were carried out with snapshots saved every 10 ps of the last 50 ns simulations.

Effect of small molecule binding at α8 pocket on PXR LBD dynamics

After establishing the presence of the new hot spot– “α8 pocket”, we re‐evaluated the conformational landscape of PXR LBD by (a) identifying drug‐like small molecules for this pocket by structure‐based virtual screening and then (b) performing simulation of PXR LBD with a ligand bound to the α8 pocket. The results from these studies are detailed below.

Structure‐based virtual screening against α8 pocket

We used structure‐based virtual screening to identify best‐fit drug‐like small molecules for the α8 pocket. The virtual screening was carried out with 1593 compounds from the NCI Diversity Set IV database31 using Autodock Vina.32 For this purpose, a representative PXR LBD structure with open α8 pocket conformation, with an angle TKE value of 55^°, was picked up from the apo simulation trajectory [Fig. 6(B)]. The angle TKE of 55^° was chosen since about 45% of the simulation trajectory possesses an angle value ≥55^°. The calculated volume of α8 pocket in the selected structure was found to be 360 ų. As shown in Table S3, we have identified a handful of virtual hits with relatively high binding affinity to α8 pocket. Furthermore, virtual screening was carried out with the RMSD‐based cluster centroids to examine the effective binding of small molecules with different conformational states of the α8 pocket sampled during simulation. The centroids with α8 pocket volume higher than 350 ų (Table S4) have shown comparatively similar results as in Table S3. Notably, the centroids with α8 pocket volume lesser than 240 ų failed to retain any virtual hit in the pocket. A quinolinol derivative compound (identifier “NSC1014” in NCIDS IV database31) with comparatively high binding affinity (−8.93 kcal/mol) was considered for further study. This drug‐like compound forms hydrogen bonds with residues Leu334 and Glu336 from the α8 helix (Fig. S13). The PXR LBD‐NSC1014 complex was subjected to 100 ns simulation in order to characterize the newly identified binding hot spot and its relevance in the context of LBD dynamics. It should however be noted that we encountered many other virtual hits with binding affinity comparable to NSC1014 (Table S3). Since all the compounds were docked to α8 pocket in the open conformation of PXR LBD, it is likely that they exhibit similar dynamical response as NSC1014.

Effect of open α8 pocket on LBD dynamics

To understand the effect of open α8 pocket on dynamical behaviour of PXR LBD, we carried out MD simulations of LBD with NSC1014 docked at the α8 pocket (system 7 in Table 1). The angle TKE as a function of simulation time is presented in Figure S14. A comparison of this angle with the one accessed by the apo system clearly shows the constraint enforced by NSC1014. The presence of NSC1014 keeps the α8 pocket in the open state throughout the course of 100 ns simulation, with an average angle TKE of 57^°. On the other hand, as noted earlier, this angle fluctuates considerably in system 1, taking up values ranging from ∼40^° to ∼60^° with an average of 51^°. Porcupine plots for system 7 along the direction of PC1 and PC2 [Fig. 8(A‐B)] clearly indicate that the presence of NSC1014 in α8 pocket has restricted the movement of α8 helix, freezing the pocket in its open state. Notably, in contrast to the apo system (Fig. 2), the presence of NSC1014 at α8 pocket also restricts the expanding/contracting movement of LBP as seen from PCA. A more detailed view of such a restricted dynamics can be seen in Movie S3. The pocket volume analysis also supports the finding from PCA. As shown in Figure 8(C), compared to the apo system, α8 pocket volume has increased in the presence of NSC1014. More interestingly, this has also led to a significant decrease in LBP volume. In fact, LBP has shrunk to a volume (average of ∼750 ų) lower than the minimum volume accessed by apo LBD (∼900 ų) during simulation. Such variations in the pocket volumes can be associated with subtle variations in the conformational states of LBD. The FEL has also been drastically altered by the binding of the compound NSC1014 at α8 pocket. As shown in Figure 8d, LBD with compound NSC1014 in the α8 pocket visited minimal number of low‐energy basins whereas the apo LBD visited about five low‐energy basins. The LBP volume taken up by the conformations extracted from different energy basins show that the volume has significantly reduced. This clearly indicates that the presence of small molecule at α8 pocket has influenced the conformational equilibrium between expanded and contracted states in PXR LBD.

Figure 8.

Dynamic behaviour of PXR LBD in presence of virtual hit “NSC1014” bound to allosteric α8 pocket. (a) and (b) show the porcupine plots representing the principal motions in NSC1014‐bound PXR LBD, along the directions of PC1 and PC2 respectively. Secondary structural elements highlighted are the same as in Figure 2a‐b. (c) LBP and α8 pocket volumes accessed during the simulation of LBD‐NSC1014 complex (shown in violet). These volumes from the apo LBD simulations are shown in black for comparison. Volume calculations were carried out with snapshots saved every 10 ps of the last 50 ns simulations. (d) PCA based free energy (in kJ/mol) landscape using PC1 and PC2 as reaction coordinates. The low energy basins are marked with “E#”. Average structure from each basin with corresponding average LBP volume, with standard deviation, is shown alongside. The cliques due to side‐chain interactions that are specific to a given population are also projected on the average structure.

Together, simulations of systems 1–7 have shown that the PXR LBD in different forms (apo/orthosteric ligand‐bound/allosteric ligand‐bound) occupy conformations with distinct preferences for pocket volumes. The apo form spans a conformational space with a wide range of LBP volume (∼900 ų to ∼2000 ų) and reasonably significant α8 pocket volume (∼200 ų to ∼400 ų). The conformational state of liganded systems (system 2–6, Table 1) are limited to the geometrical nature of the bound ligand at LBP with insignificant volume (<200 ų) of α8 pocket. From the sampling of conformational states, it is clear that the ligand binding at LBP results in locking the α8 pocket in a closed state. Further, all ligands bound to LBP have the same effect of rigidifying the regions around LBP and consequently of LBD, providing an explanation for the promiscuous nature of PXR. The presence of small molecule NSC1014 at α8 pocket constrains the pocket in its open conformation with a wider volume (∼400 ų–∼800 ų). Strikingly, the presence of a chemical entity at α8 pocket drastically reduces the accessible LBP pocket volume to less than 900 ų. Thus, the wide conformational space (measured in terms of pocket volume) visited by the apo form gets shifted and locked to specific states, as a consequence of binding of ligands at LBP or at the newly identified α8 pocket. It can be speculated that such a conformation of PXR LBD in the presence of NSC1014 at α8 pocket can probably reduce the ligand accessibility to LBP, thus may influence the ligand mediated PXR activation.

In the recent past, handful of research has been carried out to identify effective PXR antagonists.5 In vitro and in silico studies have identified different azole derivatives including ketoconazole, K2 analogue (analogue of ketoconazole),33 SPB3255,13 etc., as potent lead compound against PXR activation. Although alternative sites of interaction have been proposed for PXR antagonist,5, 14 approachability of known antagonists for the α8 pocket will be helpful in characterising the druggability nature of the newly identified pocket. Table S5 shows the binding affinity of some of the known PXR antagonist for the α8 pocket, calculated through molecular docking. Notably, the binding energies are significantly high when the antagonists are docked at the α8 pocket as compared to the AF‐2 surface. The results are indeed encouraging and thus, should stimulate in silico, structural and mutational studies in further exploring the α8 pocket for allosteric PXR antagonism. On a similar line, it is interesting to note the recent work by Wassman et al., where they successfully reactivated mutant p53 by targeting a transiently open L1/S3 pocket identified through MD simulations.27

At a conceptual level, the present study addresses the phenomenon of allostery. Our results show that α5 and α7 helices, which contribute interacting residues (285, 289 and 330) to α8 pocket also contribute residues (284, 285, 288, 291, 292, 321, 323, 324, 327) to interact with ligands bound to the well‐established LBP.6 This suggests that α8 pocket is indeed an allosteric site, as the α5 and α7 helices orientations are modulated by ligands binding to either of the two sites (Movies S2 and S3). Reading along similar line, recent thoughts on allostery also go beyond the static picture of protein conformations.34, 35 The dynamic view such as the equilibrium conformational populations and the effect of ligands on conformational redistribution and, at a more fundamental level, on the free energy landscape, are some of the ideas being put forward to explain allostery. Here, we have touched upon some of these concepts by analysing the simulation trajectories of PXR LBD in apo and different ligand‐bound states.

Summary and Conclusions

MD simulations are carried out to understand the correlation of protein dynamics with the promiscuous nature of ligand binding in PXR LBD. The simulations unveil different conformational states exhibited by apo PXR LBD– “expanded” and 'contracted'. The study further reveals that the ligand binding events restrict the LBD dynamics and lock the protein in a conformation feasible to the size and shape of the incoming ligand, that explain the promiscuous nature of ligand binding exhibited by PXR. Computational solvent mapping with MD ensembles has identified a potential small molecule binding hot spot located outside LBP, denoted as α8 pocket. The simulations further suggest a direct correlation between opening/closure of the α8 pocket and expanding/contracting motions of LBP. MD simulation of PXR LBD complexed with a virtual hit NSC1014, identified through structure‐based virtual screening, shows that the open conformation of α8 pocket has allosterically arrested the LBD in a contracted state.

The present study has clearly brought out the importance of alternate conformations of PXR needed for its function. It is to be noted that members of the NR superfamily has been an important pharmaceutical target for wide range clinical conditions and PXR antagonism has gained considerable significance in recent years. Although alternative sites for antagonists have been proposed in the literature,5, 14 the one identified here as α8 pocket seems to be promising in terms of parameters such as allosteric pocket, alternative conformations and consistency of protein‐ligand interactions. In vitro and in vivo exploration of targeting α8 pocket for effective PXR antagonism is likely to be a focus of future research.

Materials and Methods

Data set and molecular simulations

A series of all‐atom MD simulations of PXR LBD in its apo and liganded states (Table 1) were performed. The initial structure of apo6 (system 1) and five different ligand‐bound PXR LBD (system 2‐6) were obtained from protein data bank. Different ligand‐bound structures used in this study include PXR LBD co‐crystalized with colupulone (a constituent of Hops extract),23 hyperforin (a constituent of St. John's wort),24 rifampicin (antibiotic),16 SR12813 (a cholesterol‐lowering compound),17 and T1317 (a synthetic oxysterol).9 The chemical structure of all the ligands are depicted in Figure S4. The mutated Ser284 was replaced with wild‐type cysteine residue in T1317‐bound crystal structure. A set of partial atomic charges for all the ligands were obtained via quantum calculations. A B3LYP geometry optimization procedure was performed using Gaussian 0936 with the 6‐31 + G* basis set. The atom‐centered RESP charges37 were calculated via fits to the electrostatic potentials obtained from the calculated wave functions. The missing interaction parameters for all the ligands were generated using antechamber tools in Amber 11.0.38 The 4‐methyl‐1‐piperazinyl ring of rifampicin, for which crystal data is not available, was added using antechamber tools in Amber 11. A ligand library created using geometry optimized rifampicin was used for this purpose. The coordinates of missing residues in α1‐β1, β1′‐α3, and β4‐α7 regions were modelled using the Modeller package.39

The hydrogen atoms were added on heavy atoms by leap program in Amber 11.0 package38 and energy minimized for 1000 steps by steepest descent algorithm, followed by another 1000 steps by conjugate gradient method. The protonation states of histidine residues–HID or HIE–were determined using WHATIF program.40 The systems were solvated in a cubic periodic box of explicit water with water molecules extending 10 Å outside the protein complex on all sides. The 3‐site TIP3P water model was chosen to define the water molecules. The simulation box contained ∼15,000 TIP3P water molecules, and the box size was around 75 × 82 × 80 ų. Charge neutrality was maintained by adding three Na+ ions in all the systems. An extensive set of minimization and thermalization were performed by maintaining harmonic restraints on protein heavy atoms and the temperature was gradually raised to 300 K in canonical ensembles. The harmonic restraints were slowly reduced to zero and solvent density was adjusted under isobaric and isothermal conditions at 1 atm and 300 K. The systems were equilibrated for 10 ns in NPT ensemble, with 2 fs simulation time step. The long‐range electrostatic interactions were calculated using Particle Mesh Ewald sum with a cut off of 10 Å applied to Lennard‐Jones interactions. The SHAKE algorithm was used to constrain all bonds involving hydrogen atoms. Sander module of Amber 11.038 MD simulation software package with Amber ff99SB force field was used for all simulations to generate 100 ns production data.

The simulations trajectories saved at an interval of 2 ps were used for analyses. Fpocket program suite41 was used for the pocket volume analysis. The volume analysed with Fpocket showed good agreement with the previously reported values for different PXR LBD crystal structures (Table S2) Ligand–protein pairwise interaction energies were calculated using MMPBSA.py program in CPPTRAJ trajectory analysis tool in Amber 11.0.38 Ligand interaction diagrams were prepared using LigPlot+ software.42 The visual analysis of protein structures were carried out using VMD43 and PyMOL.44

Principal component analysis (PCA) and free energy landscape (FEL)

Principal component analysis (PCA), also termed essential dynamics analysis, is utilized in this work to study the large concerted motions in PXR LBD in its apo and liganded state. PCA is a dimension‐reduction tool, which transforms the original set of correlated (possibly) variables into a reduced set of uncorrelated (independent) variables called the principal components (PCs).45 These PCs thus help in locating the essential motions in complex systems with high dimensionality, like in proteins. The analysis involves the calculation of positional covariance matrix C of the atomic coordinates and its eigenvectors. The eq (1) define the elements of the 3N dimensional positional covariance matrix C, which is calculated with the ensemble of protein structures,

$$C_{ij}=〈\left(x_i-〈x_i〉\right)\cdot \left(x_j-〈x_j〉\right)〉\hspace{0.17em}\left(i,j=1,2,3,\hspace{0.17em}\dots ,\hspace{0.17em}3N\right),$$ (1)

where x_i and x_j are the atomic coordinates (here backbone atoms), N is the number of atoms considered and ⟨x⟩ represents the ensemble average. The eigenvectors of covariance matrix, V, obtained by solving V ^T CV =ë, are the principal modes in the configurational space along which the fluctuations observed in the simulation are uncoupled with respect to each other. Associated eigenvalues, ë, define the mean square fluctuation of the motion along these vectors. The principal modes are ranked (in ascending order) according to their contribution to the total mean square fluctuation. We removed the overall rotational and translational motions prior to generating the covariance matrix. PCA analysis was carried out using PCAZIP program package.46 The porcupine plots were used to visualize the eigenvectors and the length of the porcupine represents the magnitude of the motion.18 The porcupine plots were visualized using VMD.43

The free‐energy landscape (FEL) of a protein can be obtained using conformational sampling methods, like MD simulations, that allow exploring the near‐native conformational states. To obtain a two‐dimensional representation of the energy landscape, motion of top two principle components, PC1 and PC2, were selected as reaction coordinates. The energy landscape along these two reaction coordinates can be obtained by eq (2),

$$G\left(q1,q2\right)= -k_{\mathrm{B}}T\text{ln}P\left(q1,q2\right),$$ (2)

where k _B is the Boltzmann constant, T is the temperature of simulation, and P(q1,q2) is the normalized joint probability distribution.

Protein structure networks

In this study, protein structure networks (PSNs) were generated at the level of non‐covalent side chain interactions using the technique defined in refs 47 and 48. Briefly, the networks were constructed by considering amino acid residues as nodes and connections (edges) were made between a pair of residues, based on the existence of non‐covalent interaction between them. For each pair of residues, interactions were considered within a cut off distance of 4.5Å and the strength of the interactions was calculated using eq (3),

$${\mathrm{I}}_{\text{ij}}=\frac{{\mathrm{n}}_{\text{ij}}}{\sqrt{{\mathrm{N}}_{\mathrm{i}}\times {\mathrm{N}}_{\mathrm{j}}}}\times 100,$$ (3)

where I_ij is the interaction energy between residue i and j, n_ij is the number of atom pairs within the cut off distance and N_i/j is the normalization value of residue i and j.

The PSNs have found a number of applications in elucidating protein structure‐function relationships. This includes tracking of subtle conformational changes either in the equilibrium populations or due to perturbations like ligand binding, which is characterized in terms of higher order network connectivity through cliques/communities.21, 49 Clique in a network is a set of k‐nodes with each node connected to all others in the set. Community is defined as a set of connected cliques (sharing either k−1 or k−2 nodes). Specific application of cliques/communities in PSN, relevant to the present work, is to identify changes in subtle conformational populations from simulation trajectories and also to identify changes due to different liganded states of PXR LBD. Here, PSNs were generated using an automated approach (PSN Ensemble), which was developed to calculate dynamically stable network parameters along the simulation trajectories.50 In this study, the network parameters were generated at I_min of 3% from MD snapshots saved at every 10 ps and considered to be dynamically stable if they occur in at least 50% of the snapshots.

Conformational clustering and binding site mapping

An RMSD‐based conformational clustering algorithm was used to generate a reduced dataset of the 100 ns long simulation trajectories. Prior to clustering, all the snapshots were superimposed on each other using Cα atoms of key secondary structures–helices α3, α5, α7‐α10 and αAF, and five strand β sheet, in order to remove rotational and translational motions. Since our aim is to do a global search for binding sites, RMSD‐based clustering was carried out with the same large set of Cα atoms. The cluster algorithm in CPPTRAJ module of Amber 1138 was used for the calculation, where the RMSD threshold was set to 2 Å. Representative centroid structures from top 15 clusters were used for binding site mapping (Fig. S11).

Each cluster centroid conformation was mapped for potential small molecule binding sites using the fragment‐based FTMap algorithm through its online server (http://ftmap.bu.edu). The detailed description of the algorithm and methodology can be found elsewhere.30 Briefly, using a fast Fourier transform approach, FTMap scans the global surface of the protein with a diverse library of 16 small molecule probes. Energetically favourable protein‐probe complexes from different probe types were clustered into consensus sites (CSs), which define the hot spots where multiple probes can bind with high affinity. The CSs were ranked based on the number of different probe clusters they incorporate, with the top‐ranked CS representing the largest and most important site of interest. It should be noted that a hot spot can be formed by binding of multiple CSs that represents different functional moieties of a ligand.

Virtual screening

Screening of a small molecule library of NCI Diversity Set IV (NCIDS)31 against the open α8 pocket was performed using Autodock Vina.32 One of the LBD conformations with open state of α8 pocket was extracted from the trajectories of apo simulations (system 1) using an angle metric, “angle TKE” (see section “Structure‐based virtual screening against α8 pocket” in Results and Discussion for further details). Screening was carried out within a search space 20x20x20 Å centred on the α8 pocket with an exhaustiveness value of 100. The compound with highest binding affinity was selected for further investigation. We generated system 7 (Table 1) by docking one of the top‐ranked virtual hits NSC1014, identified through screening, to the α8 pocket of LBD. MD simulations of system 7 were carried out following the similar protocol as in system 1–6, to explore the effect of bound compound on LBD dynamics.

Protein‐ligand docking

Molecular docking of known PXR antagonists‐ ketoconazole, K2 analogue (analogue of ketoconazole), and SPB03255, to the α8 pocket and AF‐2 surface were performed with AutoDock 4.2.51 An affinity grid map of 20x20x20 Å with 0.375 Å spacing centred on respective binding sites were generated using Autogrid. Ligand conformational search was carried out using Lamarckian genetic algorithm with a population of 150 individuals, a maximum number of 1,000,000 energy evaluations and a maximum number of 50,000 generations. The final docked conformations lying within 1 Å RMSD were clustered and the best docked conformation was used for analysis.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Supporting Information