X-ray Crystal Structure of the Liver X Receptor β Ligand Binding Domain

The x-ray crystal structures of the human liver X receptor β ligand binding domain complexed to sterol and nonsterol agonists revealed a perpendicular histidinetryptophan switch that holds the receptor in its active conformation. Hydrogen bonding interactions with the ligand act to position the His-435 imidazole ring against the Trp-457 indole ring, allowing an electrostatic interaction that holds the AF2 helix in the active position. The neutral oxysterol 24(S),25-epoxycholesterol accepts a hydrogen bond from His-435 that positions the imidazole ring of the histidine above the pyrrole ring of the tryptophan. In contrast, the acidic T0901317 hydroxyl group makes a shorter hydrogen bond with His-435 that pulls the imidazole over the electron-rich benzene ring of the tryptophan, possibly strengthening the electrostatic interaction. Point mutagenesis of Trp-457 supports the observation that the ligand-histidine-tryptophan coupling is different between the two ligands. The lipophilic liver X receptor ligand-binding pocket is larger than the corresponding steroid hormone receptors, which allows T0901317 to adopt two distinct conformations. These results provide a molecular basis for liver X receptor activation by a wide range of endogenous neutral and acidic ligands.

Nuclear receptors regulate transcription through the recruitment of coactivator proteins to the ligand binding domain (LBD) (11). Structural and biochemical studies reveal that the coactivator contains a short ␣-helical LXXLL sequence (where X ϭ any amino acid), known as an NR box, that binds the nuclear receptor LBD. The NR box is capped by a charge clamp on the surface of the LBD formed by a lysine on helix 3 and a glutamic acid on the C-terminal AF2 helix (12). Despite the availability of multiple co-crystal structures of ligand/receptor complexes, the mechanism by which small molecule ligands activate nuclear receptors is still poorly understood (13). We have shown previously that a residue in the AF2 helix of LXR␣, Trp-443, plays a role in the activation of the receptor by oxysterols (14). Based on these results, we proposed a model where the AF2 helix was stabilized in its active conformation by a hydrogen bond from the tryptophan indole NH to the sterol agonist (14). Interestingly, both the neutral sterol eCH and the acidic nonsterol T1317 are efficacious activators of LXR␣ and LXR␤ in biochemical and cell-based assays (Fig. 1, b and c). To probe the molecular basis by which both sterol and nonsterol agonists regulate LXR activity, we initiated crystallographic investigations of both LXR␣ and LXR␤ LBD. Diffracting crystals of the LXR␤ LBD complexed to eCH and T1317 were readily obtained, allowing us to determine those structures.

EXPERIMENTAL PROCEDURES
LXR␤ Expression and Purification-A construct of human LXR␤ LBD (ligand binding domain) containing amino acids 214 -461 and an N-terminal thrombin cleavable His tag was subcloned into the Escherichia coli expression vector pRSETa (Invitrogen) and expressed in BL21(DE3). All subsequent steps are at 4°C. Cells were resuspended in NiA (25 mM imidazole, pH 8.0, 150 mM NaCl, 3% 1,2-propanediol) and then lysed using a cell homogenizer (Rannie) followed by clarification by centrifugation.
Protein was further purified using anion exchange. His-tagged LXR was digested overnight with thrombin at a mass ratio of 1:500. The digested protein was diluted to 25 mM salt with Q0 (10 mM Tris, pH 8.0, 0.1 mM EDTA, 5% glycerol, 5 mM dithiothreitol), loaded onto the preequilibrated anion exchange column, and eluted by a 20-column 0 -250 mM NaCl gradient. LXR eluted as two peaks at ϳ150 mM NaCl, which were kept separate for crystallization trials. The eluted protein was dialyzed versus 10 mM Tris, pH 8.0, 0.1 mM EDTA, 5% glycerol, 5 mM * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The dithiothreitol, 150 mM NaCl, eCH, and SRC1, or T1317 was added, and the protein was concentrated to 12-14 mg/ml for crystallization trials.
Crystallization and Structure Determination-Crystallization trials were carried out using the hanging drop method by mixing 2 l of protein solution with 2 l of well buffer. LXR␤/T1317 complexes crystallized from 10 to 20% polyethylene glycol 3350 with 100 mM concentration of a number of salts, including NaF, KF, NaCl, KCl, sodium formate, sodium acetate, and potassium acetate at 4°C. LXR␤/eCH/ SRC1 complexes crystallized from 10 to 12% polyethylene glycol 3350 -8000 with 0.2 M NaCl at 4°C. Crystals took 4 -6 weeks to grow. Crystals were frozen in LN2 after transferring stepwise to a cryo buffer containing the well buffer with 30% polyethylene glycol 400.
All data were obtained at the Industrial Macromolecular Crystallography Association Collaborative Access Team (IMCA CAT) 17 ID beam line at the Advanced Photon Source at a wavelength of 1 Å on a MAR CCD detector. Crystals of the LXR␤/T1317 complex diffracted to 2.3 Å and were in the space group P2 1 2 1 2 1 with unit cell parameters a ϭ 60.25, b ϭ 82.454, c ϭ 123.175. Crystals of the LXR␤/eCH/SRC1 complex diffracted to 2.7 Å and were in the space group C222 1 with lattice constants of a ϭ 71.17 Å, b ϭ 120.01 Å, c ϭ 147.56 Å. All diffraction data were integrated and scaled using HKL2000 (15). The LXR␤/T1317 structure was solved by molecular replacement using the program AMoRe with a truncated monomeric form of human retinoic acid receptor-␥ as a search model. The LXR␤/eCH/SRC1 structure was solved by molecular replacement using a refined LXR subunit from the LXR␤/ T1317 structure as a search model. For both complexes, there were two molecules in the asymmetric unit related by a noncrystallographic dyad. Noncrystallographic averaging was utilized during refinement. Structures were subjected to multiple rounds of building using Quanta and refined using CNX and Refmac5.
Biological Assays-LXR cell-free ligand sensing assays and LXR cell-based transactivation assays were performed as described previously (14).
Chemical Compounds-T1317 and eCH were synthesized as described previously (10,16). The T-CH hybrid was synthesized from chlol-5-en-24-al by addition of (trifluoromethyl)trimethylsilane catalyzed by tetrabutylammonium fluoride, oxidation with Dess-Martin periodinane, and addition of a second equivalent of (trifluoromethyl)trimethylsilane catalyzed by tetrabutylammonium fluoride. Analytical data are as follows: 1

RESULTS
Structure Determination-The human LXR␤ LBD (residues 214 -461) was crystallized in complex with T1317 or with eCH and a peptide from the coactivator SRC1. Crystals of the LXR␤/ T1317 complex diffracted to 2.3 Å, whereas the LXR␤/eCH/ SRC1 complex diffracted to 2.7 Å (Table I). The LXR␤ LBD had the anticipated three-layered ␣-helical fold seen for other nuclear receptors (13). The most distinguishing features of the LXR␤ structure were a long helix 1 (ϳ18 residues) and a relatively large ligand binding pocket (830 Å 3 ) as compared with the classic steroid hormone receptors (420 -550 Å 3 ). In both structures, the asymmetric unit contained an LXR␤ homodimer, where each monomer was occupied by a single ligand (Fig. 2). The dimer interface was composed of residues in helices 7, 9, and 10 and is similar to other homo-and heterodimer interfaces (13). In the eCH structure, the C␣ atoms in the two subunits have a root mean squared deviation of 0.4 Å, whereas in the T1317 structure, the C␣ atoms have a root mean squared deviation of 0.6 Å.
In the eCH/LXR␤ complex, the sterol bound with the A-ring oriented toward helix 1 and with the D-ring and epoxide tail oriented toward the C-terminal end of helix 10, Trp-457, and His-435 (Fig. 3a). This orientation was similar to that predicted in the model of Spencer et al. (14) and was similar to that of estradiol, progesterone, and dexamethasone in their complexes with the estrogen, progesterone, and glucocorticoid receptors. However, the steroid core of eCH was flipped 180°around its long axis so that the angular methyl groups pointed in the direction opposite that in the steroid hormones (Fig. 3a). Although the epoxide oxygen atom lay adjacent to Trp-457, it actually made its hydrogen bond with the imidazole ring of His-435. The A-ring hydroxyl formed a hydrogen bond with Glu-281 on helix 3 and was positioned close to Arg-319 on helix 5. This is reminiscent of the estradiol/estrogen receptor complex, where the phenolic A-ring hydroxyl makes strong hydrogen bonds with Glu-353 on helix 3 and Arg-394 on helix 5 (17). However, in LXR␤, the A-ring hydroxyl group lay out of the plane of the Arg-319 guanidinium group, and despite its favorable distance, cannot make a hydrogen bond with good geometry. In addition to these polar interactions, eCH also made extensive lipophilic interactions with the ligand binding pocket, and its conformation was essentially the same in both subunits. In the T1317/LXR␤ structure, the acidic carbinol group lay in approximately the same position as the epoxide of eCH and also formed a hydrogen bond with the histidine ring of His-435. However, the nonsterol ligand was observed in different conformations about the tertiary sulfonamide in the two LXR␤ subunits. In one subunit, the T1317 adopted a gauche conformation (Fig. 3a), whereas in the other subunit, it adopted an anti conformation (Fig. 3b). As a result, the nonsterol ligand fit into a position corresponding to the C-and D-rings of eCH and made weak hydrogen bonds with Ser-278 in each case but did not reach into the volume occupied by the A-ring of eCH and failed to make interactions with either Glu-281 or Arg-319 (Fig. 3, a and b). Despite the differences in the regions of the ligand binding pocket occupied by the eCH and T1317, all of the amino acids contacting both ligands are conserved between LXR␣ and LXR␤.
Mechanism of Ligand Activation-A tryptophan in the LXR␣ AF2 helix (residue 443) has been shown to be essential for oxysterol activation of the receptor (14). In LXR␤, the corresponding residue is Trp-457. In contrast to the published homology model (14), neither the sterol nor nonsterol ligand formed a direct interaction with the AF2 helix in the crystal structure (Figs. 3 and 4). The Trp-457 indole was oriented such that the nitrogen atom was pointed away from eCH, making it impossible to form a direct hydrogen bond with the epoxide oxygen. Instead, both ligands interacted with His-435, which made an edge to face interaction with the tryptophan on the inner surface of the AF2 helix (Fig. 4, a and b). The His-435 side chain had some freedom to rotate about its C␣-C␤ bond, allowing it to swing the edge of its imidazole ring across the face of the indole ring. This rotational freedom lets His-435 interact with either the weakly negative -electron cloud of the fivemembered pyrrole ring or the more strongly negative -electron cloud of the benzene ring (18). This weak electrostatic interaction can become a strongly favorable ''cation-'' interaction when the imidazole is positively charged and directed into the most negatively charged regions of the indole -electron cloud (19,20).
The hydrogen bonding oxygen atom of the agonist ligand was in approximately the same location in the eCH and T1317 structures, but the position of the His-435 imidazole was different depending on the strength of its interaction with the ligand. In the eCH complex, the N⑀2 atom of His-435 was positioned to donate a hydrogen bond to the epoxide of the sterol side chain at a distance of 3.45-3.50 Å (Fig. 4c). In this orientation, the imidazole directed its electropositive C⑀1 and N⑀2 hydrogens toward the C␥ atom and six-membered ring of the tryptophan side chain, corresponding to the electronegative -cloud of the indole (18), providing a mechanism for the sterol to hold the AF2 helix in its active conformation (Fig. 4c). The observation that other ligands with only hydrogen bond acceptors in their side chains, such as 24-ketocholesterol and dimethyl cholenamide (7,14), are also efficacious activators of LXR suggests that this electrostatic interaction is a viable mechanism for ligand activation of the receptor.
In the T1317 structure, the imidazole swung 1.3 Å toward the acidic bis-trifluoromethyl carbinol, bringing the His-435 N⑀2 atom to a distance of only 2.59 -2.75 Å from the carbinol oxygen (Fig. 4d). In this orientation, the imidazole directed its electropositive C⑀1 hydrogen into the -cloud of the indole benzene ring, again holding Trp-457 in the active position. The imidazole rotation also opened space for a water molecule, observed in both subunits, that made a hydrogen bond to the backside N␦1, further stabilizing the complex. Hydrogen bond lengths tend to correlate with their energy (21), suggesting that the acidic carbinol group makes a stronger hydrogen bonding interaction than the epoxide. This is consistent with quantum mechanics calculations, which indicate that the epoxide oxygen has a relatively weak electrostatic charge as compared with that on the bis-trifluoromethyl carbinol or even as compared with an ordinary hydroxyl group (data not shown). The T1317 hydrogen bond may be further strengthened by partial proton transfer, as can occur when the pK a values of the partners are suitably matched (21). In this case, the bis-trifluoromethyl carbinol group has a pK a of 8.4 in water, relatively close to that of histidine, with a pK a of 6.5-7.0. Thus, the His-Trp interface induced by T1317 may be an example of a cation-electrostatic interaction (19,20). Similar cation-interactions have been observed to regulate the activity of ion channels (22,23) and enzymes (19,20). Although the degree of AF2 stabilization by the His-Trp interactions is believed to be different for the sterol and nonsterol ligands, the position of the C-terminal helix does not deviate significantly in the crystallized complexes.
Mutagenesis of the His-Trp Activation Switch-To explore the functional differences in the electrostatics of the His-Trp activation switch, we utilized mutants of LXR that might allow activation by T1317 but not eCH. Although tryptophan is the optimal pairing for histidine in a cation-interaction, phenylalanine or tyrosine can also substitute (18). LXR-GAL4 chime-  2. Structure of the LXR␤ LBD. LXR␤/eCH/SRC1 crystallized as a dimer with an orientation and dimer interface similar to that seen with other nuclear receptors. Helices in the two LXR monomers are shown in red and pink, whereas ␤-strands are yellow, loops are cyan, and the SRC1 helix is magenta. LXR helices 1-10 and AF2 are labeled H1-H10 and HAF2. Nitrogen, oxygen, and hydrogen atoms are colored blue, red, and white, respectively, whereas carbon atoms are colored green, yellow, cyan, cyan, and yellow in eCH, Glu-281, Arg-319, His-435, and Trp-457, respectively. The same dimer orientation was obtained in the T1317 complex (not shown). ras were generated where Trp-443 in LXR␣ and Trp-457 in LXR␤ were mutated to phenylalanine or tyrosine. The transactivation capability of the LXR-GAL4 chimeras was measured by their ability to transcribe a secreted placental alkaline phosphatase reporter on a UAS-tk promoter in transient transfection experiments (Fig. 5). As demonstrated previously (14), neither tryptophan point mutant could be activated by eCH. However, T1317 activated both the W443F LXR␣ and the W457F LXR␤ mutants. We interpret these data as evidence that phenylalanine can substitute for tryptophan in the electrostatic interaction induced by T1317 but not in the interaction induced by eCH. Neither compound activated the W443Y LXR␣ or the W457Y LXR␤ mutants (data not shown).
Models of the LXR point mutants were built based on the LXR␤ crystal structures to explore the structural effects of these mutations. Modeling indicated that the W457F phenylalanine was held tightly in a conformation similar to that of the tryptophan. When bound to T1317, the His-435 C⑀1 hydrogen was directed toward the -electron cloud of the W457F phenylalanine. However, the weaker hydrogen bond with eCH swung the His-435 C⑀1 hydrogen to the edge of the W457F -electron cloud, weakening the cation-interaction. This rotation also brought the electropositive N⑀2 hydrogen near the electropositive C␦1 and C⑀1 hydrogens of the phenylalanine, further opposing the interaction. Modeling suggested that the W457Y mutant would behave similarly with respect to the cationinteraction but that the polar hydroxyl of the tyrosine would occupy a lipophilic pocket of LXR, destabilizing the AF2 helix and leading to the inability of either ligand to activate this point mutant.
To rule out the potential influence of the hydrophobic portion of the LXR agonists in these results, a hybrid molecule was synthesized in which the epoxide of eCH was replaced by the bis-trifluoromethyl carbinol of T1317 (Fig. 1a). The T-CH hybrid molecule was assayed for activation of the point mutated LXR-GAL4 chimeras (Fig. 5). Activation was only observed for the W443F LXR␣ mutant and W457F LXR␤ mutant, confirming that the acidic carbinol and not the tertiary sulfonamide in T1317 was responsible for activation of the point mutants.
LXR-GAL4 chimeras were also generated in which His-421 in LXR␣ or His-435 in LXR␤ were mutated to alanine. Neither T1317, eCH, or the T-CH hybrid could activate the H421A LXR␣ or H435A LXR␤ point mutants, confirming the essential role of the histidine in activation of the receptor by both the sterol and nonsterol agonists (Fig. 5). DISCUSSION The LXRs are important drug targets for the treatment of cardiovascular diseases (24). Nonsterol LXR activators have been shown to increase reverse cholesterol transport (4), decrease local inflammatory markers (25), and prevent atherosclerosis in mice (26). However, the progression of compounds to human clinical trials has been hampered by the effects of LXR agonists on hepatic lipogenesis, which result from increased expression of SREBP-1 (10) and FAS (27). A clear understanding of the molecular basis of small molecule activation of LXR may aid the development of modulator ligands that lack the side effects of potent agonists like T1317.
The structures of the eCH/LXR␤ and T1317/LXR␤ complexes demonstrate that His-435 is the critical residue in the LBD that mediates ligand activation of the receptor. All of the amino acids that line the ligand binding pocket, including the histidine trigger and the AF2 tryptophan, are conserved in LXR␣, so the mechanism of ligand activation is almost certainly identical. Histidine is unique among the naturally occurring amino acids in that it is able to function as either a hydrogen bond donor or acceptor by changing tautomers (28). When bound to eCH, His-435 donated a hydrogen bond to the agonist ligand, whereas it may act as an acceptor when the acidic T1317 is bound (Fig. 4, b and d). Remarkably, the receptor was able to accommodate both electrostatic states through a 1.3-Å shift in the histidine, which permitted the perpendicular His-Trp interaction to stabilize the AF2 helix in its active conformation in both cases. The ability of LXR to be activated by both proton donors and acceptors suggests that the receptor may be able to detect a range of natural ligands. It is interesting to note that although eCH is a neutral hydrogen bond acceptor, other endogenous LXR activators (e.g. 24(S)-hydroxycholesterol and 22(R)-hydroxycholesterol (1, 7)) can function as both donors and acceptors, whereas the acidic circulating cholesterol metabolite cholestenoic acid has also been shown to activate LXR␣ (29). Thus, LXR shares the property of other retinoid X receptor heterodimers such as the peroxisome proliferator-activated receptors (30) and pregnane X receptor (31) that have evolved a molecular mechanism to sense multiple lipid metabolites rather than a single endocrine hormone.
Although we were unable to obtain diffracting apo-LXR crystals, the structures of LXR␤ with structurally distinct agonists suggests how the receptor conformation may change in the absence of ligand. The hydrogen bonding oxygen atoms of the agonists were in approximately the same location in both structures, but His-435 shifted, depending on the strength of the hydrogen bond. The shift of His-435 toward T1317 opened space for a water molecule that made a hydrogen bond to the backside ND1, further stabilizing the complex. The outward shift of His-435 in the eCH structure displaced the water molecule and brought the backside ND1 atom closer to the carbonyl oxygen of Ser-432, to a distance of 3.9 -4.0 Å. In the unliganded state, Ser-432 might form a stronger hydrogen bond with His-435, sliding the partial negative charge of the NE2 nitrogen closer to the indole ring. This would effectively break the cation-interaction with Trp-457, leaving the AF2 helix-free to assume conformations that fail to stabilize coactivator recruitment.
The eCH and T1317 ligands made contacts with residues in helices 3, 4, 5, the ␤-turn, helix 10, and indirectly, the AF2 helix and sat in the same common pocket observed for other nuclear receptors. In general, the ligand-receptor interface was dominated by hydrophobic contacts. Although a polar interaction was observed between Glu-281 and the A-ring hydroxyl of the sterol, no corresponding interaction was seen in the T1317 complex, suggesting that it is not required for high affinity binding to LXR. The hydrogen bond to the sterol A-ring hydroxyl by Glu-281, adjacent to Arg-319, is remarkably similar to Arg-Glu pair in the estrogen receptor that binds the phenolic OH of estradiol (17). In each case, the hydroxyl group is bound between an arginine from helix 5 and a glutamate from helix 3. Another important feature of both structures is that the agonist ligands do not fill the LXR␤ pocket. Indeed, there was enough room for T1317 to adopt distinct anti and gauche conformations about the tertiary sulfonamide (Fig. 3), suggesting that the LXR pocket can potentially accommodate a wide range of sterol and nonsterol ligands.
In conclusion, the structure of LXR␤ in complex with eCH and T1317 identified a His-Trp switch that mediates activation of the nuclear receptor. Sequence alignment indicates that the nuclear bile acid receptor farnesoid X receptor has histidine and tryptophan in corresponding positions, and x-ray crystallography suggests that it also uses a cation-mechanism of ligand activation (32). No other human nuclear receptors contain tryptophan in the AF2 helix; however, several receptors have a suitably positioned phenylalanine that could function as a -donor. Reanalysis of the x-ray crystal structures of the vitamin D receptor (33), thyroid hormone receptor (34), and retinoid-related orphan receptor ␣ (35) shows cation-stabilization of the AF2 helix through a His-Phe complex. Vitamin D and thyroid hormone directly contact the histidine residue, as seen in the LXR complexes. Since the position and electrostatic state of the histidine depends on the hydrogen bonding character of the ligand (Fig. 4, c and d), it is possible that different classes of ligands could recruit different subsets of coregulators within cells. Although we did not detect differences in the recruitment of SRC1 peptides to LXR␣ or LXR␤ by eCH and T1317 (Fig. 1b), the functional differences in the electrostatic mechanism of AF2 stabilization, combined with the large ligand binding pocket, suggest that LXR and related receptors will be a good targets for the development of modulator ligands with improved therapeutic windows. FIG. 5. Activation of LXR point mutants. The effect of sterol and nonsterol ligands on the transcriptional activity of LXR histidine and tryptophan mutants as measured by the induction of secreted placental alkaline phosphatase reporter activity on wild type (WT) and point mutant LXR-GAL4 chimeras. Data are expressed as relative reporter activity with the corresponding fold activation as compared with vehicle indicated above the bars. All compounds were tested at 10 M. All data are n ϭ 3 Ϯ S.E. a, comparison of wild type LXR␣ point mutants. b, comparison of wild type LXR␤ point mutants.