The Human Nuclear Xenobiotic Receptor PXR: Structural Determinants of Directed Promiscuity

See allHide authors and affiliations

Science  22 Jun 2001:
Vol. 292, Issue 5525, pp. 2329-2333
DOI: 10.1126/science.1060762


The human nuclear pregnane X receptor (hPXR) activates cytochrome P450-3A expression in response to a wide variety of xenobiotics and plays a critical role in mediating dangerous drug-drug interactions. We present the crystal structures of the ligand-binding domain of hPXR both alone and in complex with the cholesterol-lowering drug SR12813 at resolutions of 2.5 and 2.75 angstroms, respectively. The hydrophobic ligand-binding cavity of hPXR contains a small number of polar residues, permitting SR12813 to bind in three distinct orientations. The position and nature of these polar residues were found to be critical for establishing the precise pharmacologic activation profile of PXR. Our findings provide important insights into how hPXR detects xenobiotics and may prove useful in predicting and avoiding drug-drug interactions.

The pregnane X receptor (PXR; also known as NR1I2), a member of the nuclear receptor family of ligand-activated transcription factors, is a key regulator of cytochrome P450-3A (CYP3A) gene expression in mammalian liver and small intestine (1–5). The CYP3A gene products are heme-containing proteins that metabolize a wide variety of chemicals, including >50% of all prescription drugs (6). PXR is activated by most of the xenobiotics (exogenous chemicals) that are known to induce CYP3A gene expression, including the commonly used antibiotic rifampicin, the glucocorticoid dexamethasone, and the herbal antidepressant St. John's wort (1–4, 7, 8). Like other nuclear receptors, PXR contains both a DNA-binding domain and a ligand-binding domain. PXR binds to the xenobiotic DNA response elements in the regulatory regions of CYP3A genes as a heterodimer with the 9-cis retinoic acid receptor, also known as the retinoid X receptor (RXR) (1, 2, 4).

PXR can mediate dangerous drug-drug interactions. For example, hyperforin, a constituent of St. John's wort, activates PXR and up-regulates CYP3A expression, which leads to the metabolism of vital drugs including the antiretroviral drug indinavir and the immunosuppressant compound cyclosporin (8–11). Unlike the steroid, retinoid, and thyroid hormone receptors, which are highly selective for their cognate hormone, PXR has evolved to detect structurally diverse compounds. These include exogenous drugs and toxins as well as endogenous compounds, such as lithocholic acid (a toxic bile acid) and certain C21 steroids (pregnanes) (1,3, 7). Although these diverse interactions imply promiscuity, PXR also exhibits specificity, as evidenced by the differences in the pharmacologic activation profile of PXR across species. For instance, human PXR is activated by rifampicin and the cholesterol-lowering drug SR12813 (12, 13), whereas mouse PXR is not (7); mouse PXR is activated by the synthetic steroid 5-pregnen-3β-ol-20-one-16α-carbonitrile (PCN), whereas the human receptor is not. Thus, by binding diverse but precise arrays of compounds, PXR exhibits directed promiscuity.

Unraveling the structural basis of how PXR recognizes an array of different endogenous and exogenous compounds is critical to our understanding of how harmful compounds are cleared from the body and may also improve our ability to predict and avoid dangerous drug-drug interactions. Accordingly, two structures of the ligand-binding domain of hPXR (hPXR-LBD) were determined by molecular replacement (14) and were refined with the program CNS (15): an apo structure at 2.5 Å resolution, and a complex with the high-affinity ligand SR12813 (dissociation constantK d = 41 nM) at 2.75 Å resolution (Table 1).

Table 1

Crystallographic data and refinement statistics. Data from crystals containing I-SR12813 were used only to guide and confirm the positioning of SR12813 ligands; thus, a refined structure was not generated. Crystallization: The purified hPXR-LBD/SRC-1 complex (30) was concentrated in 250 mM NaCl, 20 mM tris-HCl (pH 7.8), 5% glycerol (v/v), 5 mM dithiothreitol, and 2.5 mM EDTA in the presence of 10-fold molar excesses of the SR12813 or I-SR12813 compounds to final concentrations of 4 and 5 mg/ml, respectively. SR12813 and I-SR12813 were synthesized in-house. The apo complex was concentrated in the same buffer to 5 mg/ml. Crystallization of hPXR-LBD was achieved by hanging-drop vapor diffusion at 22°C against the following condition: 50 mM imidazole at pH 7.2, 10% 2-propanol. The purified hPXR-LBD protein used in the crystallography (comprising residues 130 to 434) was in a 1:1 complex with amino acids 623 to 710 of the human transcriptional coactivator protein SRC-1 (31). However, this 88–amino acid coactivator peptide was not observed bound to PXR in the structure, and examination of extensively washed and dissolved crystals by SDS-PAGE revealed that this coactivator fragment was not present. Structure determination: Structures were determined by molecular replacement with AMoRe (14) using the crystal structure of the VDR as a search model (45% sequence identity over 225 equivalent Cα positions). Clear solutions were obtained in the proper enantiomorphic space group,P43212. The asymmetric unit contains one hPXR-LBD monomer. The structures were refined using torsion angle dynamics in CNS with the maximum likelihood function target, and included an overall anisotropic B factor and a bulk solvent correction (15). Residues 142 to 177 and 198 to 431 of hPXR-LBD were traced in the structures reported here. Ordered electron density for the remaining amino acids was not observed at any time during refinement. For both structures, 10% of the observed data were set aside for cross-validation using the R freestatistic before any structural refinement (32). Manual adjustments and rebuilding of the model were performed using the program O (33) and sigmaA-weighted electron density maps (34). At the later stages of refinement, 281 and 166 solvent sites were added to the apo and SR12813 complexes, respectively. Structures exhibit good geometry with no Ramachandran outliers. Molecular graphics figures were created with MOLSCRIPT (35), Raster-3D (36), and Grasp (37).

View this table:

The hPXR-LBD is an “α-helical sandwich” composed of three layers: α1/α3, α4/α5/α8, and α7/α10. This region of the molecule is similar to nuclear receptor LBDs of known structure (16,17) (Fig. 1A). The standard three-stranded β sheet is expanded to a five-stranded antiparallel β sheet (Fig. 1B), and the ligand-binding cavity is localized at the bottom of the molecule. The structures of the apo and ligand-bound forms of the hPXR-LBD are similar, exhibiting a root mean square deviation (RMSD) of 0.89 Å over all atoms. In both the apo and SR12813-bound structures, the activation function 2 helix (αAF), which plays a critical role in transcriptional activation, is packed against the body of the receptor in a position that appears permissive for coactivator interactions. The hPXR-LBD is most closely related in structure to the vitamin D receptor (VDR) (18), sharing 45% sequence identity and 1.8 Å RMSD over 225 equivalent Cα positions. More limited structural similarity (2.4 to 2.9 Å RMSD) is shared with RXR, peroxisome proliferator-activated receptor γ (PPARγ), and the estrogen and progesterone receptors (19–24) [Web table 1 and Web fig. 1 (25)].

Figure 1

(A) Structure of the LBD of the human xenobiotic receptor PXR. Residues 142 to 177 and 198 to 431 of hPXR in complex with three orientations of SR12813 are shown; α helices are in red and β strands are in green, including the two novel β strands, β1 and β1′ that complete the five-stranded antiparallel β sheet observed in this structure. Secondary structural elements are numbered according to the RXR structure (19). See also Web fig. 1 (25). (B) hPXR-LDB electron density: unbiased (F obsF calc) electron density into which the novel β1/β1′ strands (residues 210 to 228) of hPXR were traced (2.75 Å resolution, contoured at 1.2σ). (C and D) The ligand-binding cavity of hPXR. A cutaway view of the binding cavity, including electrostatic surface potentials (positive in blue, negative in red), reveals a relatively smooth, uncharged surface. The cavity is enclosed by portions of five α helices (α3, α5, α7, α10, and αAF), three β strands (β1, β3, and β4), and three loops (the α10-αAF region and the two mobile regions between β4 and α7 and from residue 198 to β1). Select residues lining the cavity are indicated. In particular, the positions of the following polar residues that contact SR12813 are indicated in red: Ser208, Ser247, Gln285, His407, and Arg410. Regions of the surface 309–321 loop, which may facilitate the expansion of the ligand-binding pocket, are also shown, including the conserved hydrophobic residues Phe315, Leu318, Leu319, and Leu320.

Although the hPXR-LBD is similar to known nuclear receptor LBDs, it contains several distinct features that appear critical to its function as a promiscuous xenobiotic receptor. First, the variable region between α1 and α3 is a four-residue turn in the hPXR-LBD (Fig. 1A). In the PPARs, this region contains α2 and is the proposed ligand access site for the binding pocket (21, 26). Second, α6 is replaced in the hPXR-LBD by a conserved, flexible loop (residues 309 to 321) that lies adjacent in space to the ligand-binding cavity and may be involved in accommodation of both small and large ligands in the binding pocket. Third, the hPXR-LBD has two additional β strands not observed previously in a nuclear receptor LBD (Fig. 1A) (17). These form the fourth (β1, residues 210 to 217) and fifth (β1′, residues 221 to 226) strands of a five-stranded antiparallel β sheet (Fig. 1B). An “insertion domain” containing roughly the same number of residues (but only 12% sequence identity) was engineered out of the VDR before crystallization and structure determination (18) [Web fig. 1 (25)].

The ligand-binding cavity of the hPXR-LBD is largely hydrophobic and is lined by 28 amino acid residues that are desolvated when SR12813 binds. The structures of the apo and ligand-bound cavities are similar, exhibiting a RMSD of 1.12 Å over all atoms in the 28 residues. The binding cavity volume of 1150 Å3 is substantially larger than that of many other nuclear receptors, including the progesterone, estrogen, retinoid, and thyroid hormone receptors. Twenty cavity-lining residues are hydrophobic, four are polar (Ser208, Ser247, Cys284, and Gln285), and four are charged or potentially charged (Glu321, His327, His407, and Arg410) [Web table 2 (25)]. A salt bridge between Glu321 and Arg410 effectively neutralizes their charged character, so that the inner surface of this ligand-binding cavity is relatively uncharged and hydrophobic (Fig. 1, C and D). Five critical polar residues spaced evenly throughout the upper portion of the ligand-binding cavity form key binding interactions with SR12813 (Ser208, Ser247, Gln285, His407, and Arg410) (Fig. 1, C and D).

hPXR is able to bind both small and large ligands. The flexible loop involving residues 309 to 321 spans the space between the COOH-terminus of β4 and the NH2-terminus of α7, and exhibits a mean thermal displacement parameter of 82.3 Å2over main-chain atoms despite persistent electron density. Nine of the 13 residues in this loop are completely conserved in the mammalian PXR molecules of known sequence, including four solvent-exposed hydrophobic residues (Phe315, Leu318, Leu319, and Leu320). The 309–321 loop is linked to the ligand-binding cavity of hPXR by a non–solvent-accessible pore (Fig. 1, C and D). The binding of a large compound (e.g., rifampicin) may force this pore to open, lining an enlarged ligand-binding pocket with additional hydrophobic residues. Thus, structural flexibility may allow PXR to bind both to small ligands and to the much larger compound rifampicin.

Three distinct binding modes of the high-affinity ligand SR12813 (positions 1, 2, and 3) were observed in the ligand-binding cavity of hPXR (Fig. 2). These orientations were identified during structural refinement and were rigorously confirmed using difference maps involving data obtained from crystals containing an iodinated form of SR12813 (I-SR12813) (27). Each orientation forms distinct interactions with residues that line the ligand-binding cavity of PXR. Although ligand position 3 forms the most hydrophilic interactions, positions 1 and 2 were clearly indicated in the detailed examination of difference maps and in refinement. Of the 19 residues involved in contacting these orientations of SR12813, only Phe288 interacts with all three ligand conformations; a phenylalanine residue is conserved at position 288 in the known mammalian PXR sequences. The remainder of the hydrophobic residues contact either one or two ligand orientations (Fig. 2). Two polar side chains (Ser247 and His407) interact with two orientations of the ligand; the remaining four (Ser208, Cys284, Gln285, and Arg410) interact with only one orientation [Web table 2 and Web fig. 2 (25)].

Figure 2

Three experimentally observed positions of SR12813 in the ligand-binding pocket of hPXR. Intermolecular interactions are shown with amino acid side chains in blue and the Cα atom as a sphere. The positions 1, 2, and 3 of SR12813 are rendered in cyan, purple, and orange, respectively. Equivalent side chains from the apo structure are shown in white. A small number of residues undergo rotamer shifts (Met243, Cys284, and His407) or small shifts in position (Ser208 and Leu209) upon SR12813 binding. Residues mutated to examine the specificity of mouse PXR are underlined. (A) Position 1 makes van der Waals contacts with eight side chains, and forms one 3.0 Å hydrogen bond with Ser247. (B) Position 2 makes van der Waals contacts with seven side chains, and forms one 2.8 Å hydrogen bond with His407. (C) Position 3 makes van der Waals contacts with six side chains, and forms three hydrogen bonds with Ser247, Gln285, and Ser208, which forms a water-mediated hydrogen bond. See also Web table 2 and Web fig. 2 (25).

Two salt bridges occur across the region of the ligand-binding cavity that is closest to the surface of hPXR: Arg410-Glu321 and Arg413-Asp205 (Fig. 3A). To examine the functional relevance of these salt bridges, we individually mutated each of these residues to alanine. Using a luciferase reporter gene assay (28), we treated CV-1 cells with increasing concentrations of either SR12813 or rifampicin (Fig. 3B). Mutation of Asp205 resulted in a marked decrease in the basal (ligand-independent) transcriptional activity of PXR, whereas mutation of either Glu321 or Arg413 modestly reduced PXR basal activity. In contrast, changing Arg410 resulted in a marked increase in the basal activity of PXR. These data demonstrate the importance of these residues in determining the basal activity of the PXR. In addition, rifampicin was found to be a more potent activator of Asp205 → Ala PXR than of wild-type PXR. This effect was not seen with SR12813, which suggests that SR12813 and rifampicin bind to PXR in different ways.

Figure 3

(A) Two salt bridges adjacent to the ligand-binding cavity of hPXR. Arg410 forms a 2.8 Å salt bridge with Glu321 while Arg413 forms a 2.6 Å electrostatic interaction with Asp205. Arg410 and Asp205 are also in van der Waals contact at 3.4 Å apart. The three orientations of the SR12813 molecule in the ligand-binding cavity are indicated, rendered in the same colors as in Fig. 2. The LBD is rotated ∼180° about the vertical axis relative to that shown in Fig. 1A. (B) Mutagenesis of the electrostatic residues. CV-1 cells were transfected with expression plasmids for hPXR and for each of the four residues mutated to alanine (D205A, R413A, E321A, or R410A, as indicated), and with the XREM-CYP3A4-luciferase reporter. Cells were treated with the indicated concentrations of either SR12813 or rifampicin for 48 hours, and cell extracts were assayed for luciferase and alkaline phosphatase activity. Luciferase values were normalized to alkaline phosphatase activity. Data points represent the mean of assays performed in quadruplicate. Mutants were generated using the QuikChange mutagenesis kit (Stratagene) according to the manufacturer's instructions. All mutants were confirmed by sequence analysis. Transient cotransfection experiments were performed using CV-1 cells and the XREM-CYP3A4-luciferase reporter (28), containing the enhancer and promoter of CYP3A4 driving luciferase expression, as described (38). Rifampicin and PCN were purchased from Sigma; SR12813 was synthesized in-house. (C) Four point mutations alter mouse PXR's sensitivity to ligands. CV-1 cells were transfected with expression plasmids for hPXR, mouse PXR, or R203L-P205S-Q404H-Q407R (mouse-human hybrid, using the mouse PXR numbering scheme; the corresponding human residues are 206, 208, 407, and 410, respectively) as indicated and the XREM-CYP3A4-luciferase reporter. Cells were treated with vehicle alone or with 2.5 μM PCN or SR12813 for 48 hours, and cell extracts were assayed for luciferase and alkaline phosphatase activity. Luciferase values were normalized to alkaline phosphatase activity. Data points represent the mean of assays performed in quadruplicate ± SE and are plotted as activation relative to vehicle treatment. Transient cotransfections and mutations were performed as described in (B).

Marked differences in the pharmacological activation profiles of PXR occur across species. For example, human and rabbit PXR are activated efficiently by SR12813, whereas mouse PXR is not. Conversely, mouse PXR is activated more efficiently by the synthetic steroid PCN than by the human or rabbit orthologs (7, 29). We next investigated the importance of the polar residues in the pocket in establishing these species differences. Four residues that interact with SR12813 in the hPXR crystal structure and differ between human and mouse PXR were chosen for mutagenesis (Fig. 2) [Web table 2 (25)]. Each of these residues was mutated in the context of a mouse PXR expression plasmid to the corresponding hPXR amino acid: Arg203 → Leu, Pro205 → Ser, Gln404 → His, and Gln407 → Arg (R203L-P205S-Q404H-Q407R), using the mouse numbering scheme. As expected, wild-type mouse PXR responded to PCN and was only weakly activated by SR12813 in reporter assays (Fig. 3C). However, the R203L-P205S-Q404H-Q407R mutant (mouse-human hybrid PXR) was no longer activated by PCN but was activated efficiently by SR12813. Thus, by using the crystal structure to design targeted mutants, we were able to confer a human-like response to the mouse PXR. This finding indicates that PXR selectivity can be affected by changes in only a few polar residues within the molecule's ligand-binding cavity.

In summary, the hPXR-LBD has evolved several structural features that permit it to function as a broad chemical “sensor.” A small number of polar residues are spaced throughout the smooth, hydrophobic ligand-binding pocket of hPXR. Changes in only a few of these polar residues can have marked effects on the responsiveness of hPXR to various xenobiotics. The character of the ligand pocket mirrors the character of most of the known PXR ligands, which are generally hydrophobic and contain a small number of polar groups capable of hydrogen bonding. The unique composition of the ligand pocket not only allows hPXR to bind a diverse set of chemicals, but also (as seen with SR12813) permits a single ligand to dock in multiple orientations. This binding mode stands in sharp contrast to other nuclear receptor– ligand interactions, which have evolved to be highly specific. Because hPXR activation is responsible for an important class of drug-drug interactions, these structures may be useful for in silico screening of drug candidates to predict and avoid dangerous side effects.

  • * To whom correspondence should be addressed. E-mail: redinbo{at}


View Abstract

Navigate This Article