Agonist Ligands Mediate the Transcriptional Response of Nuclear Receptor Heterodimers through Distinct Stoichiometric Assemblies with Coactivators*

Background: Correct assembly of coactivators with nuclear receptor (NR) heterodimers is critical for transactivation. Results: The stoichiometry of SRC1 on the CAR:RXR heterodimer varies with the liganded state. Conclusion: When both subunits of the heterodimer are in agonist-bound conformation, each subunit independently binds a coactivator molecule. Significance: A novel mechanism whereby distinct stoichiometric NR·coactivator complexes affect transcriptional levels. The constitutive androstane (CAR) and retinoid X receptors (RXR) are ligand-mediated transcription factors of the nuclear receptor protein superfamily. Functional CAR:RXR heterodimers recruit coactivator proteins, such as the steroid receptor coactivator-1 (SRC1). Here, we show that agonist ligands can potentiate transactivation through both coactivator binding sites on CAR:RXR, which distinctly bind two SRC1 molecules. We also observe that SRC1 transitions from a structurally plastic to a compact form upon binding CAR:RXR. Using small angle x-ray scattering (SAXS) we show that the CAR(tcp):RXR(9c)·SRC1 complex can encompass two SRC1 molecules compared with the CAR(tcp):RXR·SRC1, which binds only a single SRC1. Moreover, sedimentation coefficients and molecular weights determined by analytical ultracentrifugation confirm the SAXS model. Cell-based transcription assays show that disrupting the SRC1 binding site on RXR alters the transactivation by CAR:RXR. These data suggest a broader role for RXR within heterodimers, whereas offering multiple strategies for the assembly of the transcription complex.

Nuclear hormone receptors (NR) 3 relay cellular signals through distinct multiprotein assemblies (1). At the basic level, small molecule signals produce structural changes within NRs and these changes determine the composition of the interacting protein complex. These changes are essential for transcriptional activity and appear to be conserved among all ligandactivated receptors that have been studied to date. NRs are characteristically modular proteins with distinct functional domains (2). At the N terminus is the DNA binding domain (DBD), which determines target gene selectivity. The ligand binding domain (LBD) is a multifunctional module that contains the ligand binding pocket, a dimerization interface that associates with the retinoid X receptor (RXR) and a C-terminal ligand-dependent transactivation domain (AF2). Multiple biochemical and structural studies on nuclear receptors have demonstrated that ligand binding results in the specific conformational changes that are associated with a transcriptionally active state (3). In this active state, the conformation of the AF2 domains typically rearrange along the receptor surface, thereby creating a new docking site for transcriptional coactivator proteins (4).
Because both receptors within NR heterodimers can bind small molecule agonist ligands, in the simplest model for transactivation, agonist binding to either receptor can generate comparable transcriptional levels of downstream genes. Furthermore, this model would predicate that the presence of agonists to both receptors at once would yield proportionately higher levels of transcription. Such model systems are exemplified by the CAR:RXR (Fig. 1A), PPAR␣:RXR, and LXR:RXR heterodimers (5). Yet, there are NR heterodimers that exhibit transcriptional responses that are distinct from this model (5,6). For instance, transactivation by RAR:RXR, VDR:RXR, and TR:RXR only occurs in the presence of the RAR, VDR, and TR agonists, and when used in combination with the RXR agonist, transactivation levels are enhanced, unaffected, or are repressed, respectively (5,7). Through structural and biophysical studies, the mechanism of transactivation has been recognized to occur through conformational changes that restrict recruitment to a single coactivator protein to RAR:RXR (8) or decrease T3 agonist binding affinity to TR:RXR (7).
CAR is most abundantly expressed in the liver and intestine and has been directly linked to the transcription of genes involved in the clearance of both xenobiotics (9 -11), and endogenous toxins such as bilirubin (12). These target genes include select P450 family monooxygenases, phase II conjugating enzymes, and xenobiotic transporters. Therefore, CAR serves as a master regulator of xenobiotic clearance and its activation can be considered a form of chemical immunity. Within the nucleus, CAR binds to RXR and forms a functional heterodimer that recognizes its specific target genes. Additionally, the transcriptional activity of CAR is induced simply by association with RXR and with no apparent need for a CAR ligand (13)(14)(15) (Fig. 1A). Although ligand is not required for activation, constitutive CAR activity is mediated through the same conserved functional domains as those utilized by ligandactivated receptors; thus the CAR:RXR heterodimer recruits coactivator proteins through the AF2 transactivation domain (16,17). Transactivation levels mediated by CAR:RXR can be augmented by agonist ligands such as 1,4-bis[2-(3,5-dichloropyridyloxy)]benzene (tcp) (18) and 6-(4-cholorophenyl)imidazo [2,1-b] [1,3]thiazole-5-carbaldehyde O-(3,4-dichlorobenzyl)oxime (CITCO) (19), which are selective for mouse and human CAR, respectively, whereas 9-cis-retinoic acid (9c) can function as an RXR agonist (20). In both cases, these agonists enhance constitutive activity by stabilizing the constitutive AF2-coactivator interaction. Additionally, transactivation by the CAR:RXR heterodimer can be potentiated by the RXR agonist, 9c (Fig. 1A). 9c binds the ligand binding pocket of RXR and evokes the canonical NR conformational changes that result in direct interactions with coactivators. The SRC coactivator proteins function to recruit the cellular transcriptional machinery to activated NRs (21,22). SRCs also play an essential role as histone acetyltransferases to acetylate histone proteins and consequently enhance transcriptional activity (23). Thus, there is a direct link between agonist ligand binding, coactivator recruitment, and the transcription of downstream genes.
The activity of permissive NR heterodimers such as CAR: RXR that is potentiated by ligands to both CAR and RXR raises interesting questions about the precise structural assembly of nuclear factors that promote such transactivation. In this study, we propose that with permissive NR heterodimers represented by CAR:RXR, the levels of coactivator recruitment are proportional to the liganded state of the heterodimer. Thus, CAR(tcp): RXR⅐SRC1 is a 1:1 NR:coactivator complex, whereas CAR(tcp): RXR(9c)⅐SRC1 exists in 1:2 binding stoichiometry. Moreover, these levels of coactivator recruitment are proportional to transcriptional activity (Fig. 1A). These data further suggest that in this subset of NR heterodimers RXR performs a substantial role in regulating transcriptional responses within the cell. Because of the polarity of the liver X receptor response element DNA⅐CAR:RXR complex used in the transactivation assays (see "Experimental Procedures"), CAR occupies the 3Ј half-site, directly upstream from the luciferase gene. Thus, a major role of the coactivator molecule bound to CAR is to recruit the transcriptional machinery needed for luciferase production.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-CAR:RXR was purified as described earlier (24,25). Briefly, the murine CAR LBD (residues 117-358) was subcloned into the pET15b vector with an N-terminal hexahistidine tag from mCAR cDNA kindly provided by Dr. Barry M. Forman. The human RXR␣ LBD (residues 225-462) was subcloned into the pACYC184 vector was a kind gift from Dr. Bruce Wisely (Glaxo Smith-Kline, Inc.). Residues 617-769 of SRC1 (accession no. Q15788) encompassing three nuclear receptor interacting motifs (RIDs) were subcloned into the pET-SUMO vector. CAR:RXR and SRC1 were separately isolated by affinity chromatography column using nickel-nitrilotriacetic acid resin (Qiagen Inc.) To prepare various CAR:RXR⅐SRC complexes, CAR:RXR and SRC(RID1-3) were mixed in a 1:2 molar ratio, after the addition of 2ϫ molar excess of ligands, and loaded onto an S200 Superdex 16/60 column for purification of the resulting complexes (data not shown). Fractions corresponding to the complexes were pooled, measured by Bradford assay, and concentrated for further analysis.
Small Angle X-ray Scattering-Measurements were recorded at the beamlines: SIBLYS at Laurence Berkeley National Laboratory and DND-CAT at Argonne National Laboratory. Scattering for CAR(tcp):RXR(9c) was collected at Argonne National Laboratory at three concentrations ranging from 1.2 to 8 mg/ml. Scattering for all other complexes was collected at SIBYLS with protein concentrations ranging from 0.8 to 5.4 mg/ml. Curves were normalized by concentration to eliminate interparticle effects at higher concentrations. No aggregation was evident within any sample. Data were analyzed using the ATSAS software package (26) and ScÅtter (Table 1). Threedimensional model building was performed using DAMMIN and MONSA (27) and visualized using Chimera. To generate envelopes, five DAMMIN runs per sample were used and averaged using DAMAVER. -Squared values from DAMMIN are in Table 1. Three separate MONSA runs were also performed on the complexes to ensure that the ab initio fitting converged to the relative orientations of the models generated by DAMMIN. Kratky plots were calculated for shape analysis (28,29). The theoretical values for spherical objects in the V c -based Kratky plot attain an ordinate maxima of 0.82, and for nonflexible scattering particles, q 2 ϫ V c ϭ ͌3, where q is the scattering vector (Å) and V c is defined by the correlation length of the scattering particle as the ratio of the zero angle scattering intensity, I(0), to its total scattered intensity (29). Excluded from these analyses is the CAR:RXR(9c)⅐SRC1 complex due to severe problems with protein aggregation.
Analytical Ultracentrifugation-Aliquots of CAR:RXR and SRC(RID1-3) were thawed and mixed in various ratios (1:3, 1:1, and 3:1), incubated briefly with ligands, and analyzed on a Beckman XL ultracentrifuge. Both sedimentation velocity and sedimentation equilibrium experiments were performed. Sedimentation equilibrium was performed at 8 -10°C and 3 rotor speeds, whereas sedimentation velocity was run at 20°C and 55,000 rpm. Data were analyzed using SEDFIT and SEDPHAT (30). We calculate the parameters, f/f 0 , frictional ratio; s w (20,w) , sedimentation coefficient under standard conditions; and r.m.s. deviations, which reports the quality fit of the solutions of the Lamm equations (30) to the data ( Fig. 2A and Table 2).
Reporter Gene Assays-These assays were performed as reported earlier (31). Briefly, CV-1 cells were maintained in DMEM/F-12 media containing 10% fetal bovine serum and 1000 units/ml of penicillin and 1 mg/ml of streptomycin. Immediately prior to the assay, the media was changed to DMEM/F-12 with 10% charcoal-dextran-treated FBS and no antibiotics. Effectene (Qiagen Inc.) was used to transfect cells with 50 ng/well of pCMX mCAR and 100 ng/well of pCMV-TK-luc containing three copies of the liver X receptor response element and 10 ng/well of pRL CMV expressing Renilla luciferase as an internal control. The cells were dispensed on 24-well plates and ligands were added 24 h post-transfection. The ligand concentrations used were TCPOBOP (1.0 M), 0.5 M 9-cis-retinoic acid. After 48 h, cells were lysed. Activity was determined using the dual luciferase assay kit (Promega Inc.) following the manufacturer's instructions. The reported results are the average from three separate experiments. Significance in differences in activity between liganded samples was measured by a two-tailed Welch test (32).
Gal4-DBD Assays-The E456K mutation was made on pCMX-Gal4DBD-hRXR␣LBD (40 ng) (gift from Prof. Barry Forman). Transfection and assay were performed as described above with full-length CAR (50 ng), four copies of a Gal4 binding site (pUC8-MH100 ϫ 4-TK-Luc) (100 ng), and pCMV-Renilla luciferase (10 ng). As described above the reported results are an average of three separate experiments. Significance in differences in activity between liganded samples was measured by a one-tailed Welch test (32). Also, liganded values were found to differ significantly from unliganded (p Ͻ 0.05) in both CAR:RXR and CAR:RXR E456K .

CAR(tcp):RXR(9c)⅐SRC1 Assembles as a 1:2 Heterodimer⅐
Coactivator Complex-To develop an understanding of how the transcriptionally active CAR:RXR assembles with coactivators, we prepared CAR:RXR LBD and SRC1(RID1-3) in Escherichia coli and isolated multiple complexes of CAR:RXR⅐SRC1. Because CAR:RXR can also bind SRC1 in the absence of agonist, for the study here we isolated complexes CAR:RXR⅐SRC1, CAR(tcp):RXR⅐SRC1, and CAR(tcp):RXR(9c)⅐SRC1 representing the unliganded, singly liganded, and doubly liganded CAR⅐RXR complexes, respectively. SRC1(RID1-3) used in these studies comprises amino acids 617-769, and encompasses the three nuclear RIDs (33,34). Also, SRC1(RID1-3) has previously been shown to interact in a ligand-dependent manner with CAR:RXR (15,25). Using small angle x-ray scattering (SAXS), we have determined the global molecular assembly and structural properties of the CAR:RXR⅐SRC1, CAR(tcp): RXR⅐SRC1, and CAR(tcp):RXR(9c)⅐SRC1 complexes ( Fig. 2 and Table 1). For comparisons, we also measured scattering from CAR(tcp):RXR(9c) and SRC1 alone and the shapes of the scattering curves are distinct and typical of their molecular size and flexibility. Therefore, the scattering curve of the CAR:RXR heterodimer alone is characteristic of folded protein, whereas the scattering curve of SRC alone is representative of disordered proteins (Fig. 1B). We applied Kratky analyses (28) and shape comparisons to identify any noticeable difference in compactness between CAR(tcp):RXR⅐SRC1 and CAR(tcp):RXR(9c)⅐ SRC1. From these analyses we could conclude that the CAR(tcp):RXR(9c)⅐SRC1 complex is more elongated and shows greater flexibility than CAR(tcp):RXR⅐SRC1, which suggests the presence of a second SRC1 molecule within the CAR(tcp): RXR(9c)⅐SRC1 particle. We also compare the I(0)-scaled and V c -based Kratky scattering curves, which emphasize differences in size and geometry of the scattering particles (29). In the I(0)-normalized graphical plot, we compare the linearity and negative slope of the scaled intensity versus scattering angle for the complexes (Fig. 1B). The I(0)-normalized scattering plot of CAR(tcp):RXR(9c)⅐SRC1 complex is more linear with a sharper slope than the corresponding plots of either CAR(tcp):RXR⅐ SRC1 or CAR:RXR⅐SRC1 complexes. This clearly suggests that CAR(tcp):RXR(9c)⅐SRC1 is a relatively more extended molecule of higher molecular weight than the unliganded and singly liganded complexes. From the V c -Kratky plot using experimental SAXS data we are able to infer that CAR:RXR alone is mostly spherical with no apparent flexibility, whereas free SRC1 shows a hyperbolic plateau that is indicative of a highly flexible structure and with higher surface area to volume ratio (Fig. 1C). When comparing the heterodimer:coactivator complexes we note that there is a decrease in peak height between CAR(tcp): RXR⅐SRC1 and CAR(tcp):RXR(9c)⅐SRC1, which further indicates that CAR(tcp):RXR(9c)⅐SRC1 is a relatively extended particle (Fig. 1C). Overall, the observed scattering pattern is consistent with the presence of a second SRC molecule in the CAR(tcp):RXR(9c)⅐SRC complex. Furthermore, we note from a comparison of molecular envelopes of free and bound SRC1 that this molecule adopts a relatively more compact structure upon binding CAR:RXR, a feature that has been observed previously in the RAR:RXR⅐SRC1 complex (8).
To visualize the assembly of these complexes we generated molecular envelopes of CAR:RXR⅐SRC1, CAR(tcp):RXR⅐SRC1, and CAR(tcp):RXR(9c)⅐SRC1 (Fig. 2, B and C). The molecular shapes of CAR:RXR⅐SRC1 and CAR(tcp):RXR⅐SRC1 are of an elongated species and are nearly identical at SAXS resolution (Fig. 2B). The molecular shape of the CAR(tcp):RXR(9c)⅐SRC1 complex is also elongated and is ϳ1.4-fold larger than the CAR: RXR⅐SRC1 and CAR(tcp):RXR⅐SRC1 complexes (Fig. 2, C and  D). To depict the assembly of each complex, the molecular envelopes of the heterodimer and coactivator were superimposed upon the envelope of each CAR:RXR⅐SRC1 complex. Both manual and automated fitting (see "Experimental Procedures") suggest that both the CAR:RXR⅐SRC1 and CAR(tcp): RXR⅐SRC1 envelopes can encompass the CAR:RXR heterodimer bound to a single SRC1 molecule (Fig. 2B). On the other hand, the molecular shape of CAR(tcp):RXR(9c)⅐SRC1 readily corresponds to a single CAR:RXR heterodimer that is bound to two SRC1(RID1-3) molecules (Fig. 2, C and D). Thus, these SAXS analyses suggest molecular complexes with stoichiometries of 1:1 heterodimer:coactivator for the singly liganded CAR(tcp):RXR⅐SRC1 and 1:2 heterodimer:coactivator and CAR(tcp):RXR(9c)⅐SRC complexes.
Analytical Ultracentrifugation for Size Determination-To establish if the shapes of CAR:RXR⅐SRC1, CAR(tcp):RXR⅐ SRC1, and CAR(tcp):RXR(9c)⅐SRC1 correspond to their relative sizes, these complexes were independently analyzed through sedimentation velocity and equilibrium studies by analytical ultracentrifugation (AUC). The sedimentation velocity data consistently shows a species of higher sedimentation coefficient with CAR(tcp):RXR(9c)⅐SRC1 than CAR(tcp):RXR⅐SRC1 ( Fig. 3A and Table 2). From sedimentation equilibrium analyses we confirmed the molecular masses of these species to be 84.0 kDa (CAR:RXR⅐SRC1 and CAR(tcp):RXR⅐SRC1), which corresponds to the size of one CAR:RXR heterodimer (55 kDa) bound to a single SRC1 (30 kDa) and 113.0 kDa for the CAR(tcp):RXR(9c)⅐SRC1 complex, which corresponds to one CAR:RXR heterodimer bound to two SRC1 molecules (Fig. 3A). The equilibrium data therefore suggests a strong preference for 2:1 complex formation in the doubly liganded state. Taken together, the ultracentrifugation data confirm that the molecular envelopes determined by SAXS correspond to the molecular weights of these complexes as determined by AUC. Therefore, we hypothesize that transactivation by CAR(tcp): RXR(9c)⅐SRC1 relies on coactivator binding to both CAR and RXR.
Transactivation by CAR:RXR E456K Is Distorted from the Native CAR⅐RXR Complex-There are multiple factors within the cell that function to regulate transactivation by NRs (35,36). Of these factors, the SRC family of coactivator proteins are recruited specifically to the agonist-bound conformation of the NR LBD (37). In this agonist-bound conformation the AF2 domain is realigned along the receptor surface, and in doing so creates a new interface that can bind SRC proteins (38). The amino acid Glu 456 is within the AF2 domain of RXR and interacts with SRC1 (16). To better understand the role of the RXR coactivator binding site within CAR:RXR, we compare the transactivation of CAR:RXR E456K with the native protein complex in a cell-based reporter system. This Glu 3 Lys amino acid substitution has previously been shown to disrupt SRC1 recruitment by RXR (39 -42). Therefore, we predicted that the native and CAR:RXR E456K would have distinct transcriptional responses to RXR-specific agonists. When tested in CV1 cells, we observe that the transcriptional response of CAR:RXR and CAR:RXR E456K to tcp (relative to no exogenous ligand) is similar. However, transactivational levels in response to exogenously applied 9c alone (Fig. 3B) and to the combination of 9cϩtcp by the native CAR:RXR and mutant CAR:RXR E456K receptor complexes are markedly different. Specifically, 1) transactivation of CAR(tcp):RXR E456K is greater than CAR: RXR E456K , which is analogous to the trend observed with the native protein where the increased activity is potentiated by tcp (Figs. 1A and 3B); 2) transactivation of CAR:RXR(9c) E456K is at the same level as the unliganded CAR:RXR E456K , which suggests that in the mutant complex 9c is unable to increase transactivation above the basal level of unliganded CAR:RXR E456K , unlike within the native protein; and 3) transactivation by CAR(tcp):RXR(9c) E456K is at the same level as CAR(tcp): RXR E456K , but lower than by the native CAR(tcp):RXR(9c) (Figs. 1A and 3B). The measurements of SAXS and AUC above propose a distinct role for RXR within the CAR:RXR complex. Together, these results suggest that in CAR:RXR the SRC-binding site within RXR is essential for the heterodimer to achieve maximum transcriptional activity. Thus, relative to other NR heterodimers such as RAR:RXR and VDR:RXR (8,43), RXR can undertake a more significant role in transactivation by CAR:RXR.

DISCUSSION
Nuclear receptors are a superfamily of structurally and functionally conserved proteins that have evolved to regulate transcription in response to small molecule ligands through multiprotein assemblies. Hormonal agonist molecules evoke the The following abbreviation are used: I(0), interpolated intensity at 0 angle; R g , radius of gyration; V c , volume of correlation; M p , molecular weight of protein; V, volume; D max , maximum pairwise distance; P(r), pair-distance distribution function; R c , cross-sectional radius of gyration; P x , Porod exponent. A noticeably larger R g , D max , and a translated P(r) peak in CAR(tcp):RXR(9c)⅐SRC suggest a significantly larger species than CAR(tcp):RXR⅐SRC. Close agreement of real and Guinier-derived values confirm that the data is internally consistent. We use the convention of 4 ϫ sin/. 2 (Fig. 3C). Both CAR and RXR in the CAR:RXR heterodimer can independently bind their respective agonists (24). Also, crystallographic studies have shown that CAR(agonist):RXR(agonist) can bind two 13-mer LXXLL coactivator-derived peptides through binding sites on both CAR and RXR (16,17). This structural assembly has also been observed in other permissive NR heterodimers such as LXR:RXR (44) and PPAR␥:RXR (45). It is now clear from our data that these LXXLL motifs that are bound to permissive NR heterodimers are derived from two independent SRC1 molecules, although the intact SRC1 molecule has three distinct LXXLL-containing RIDs (46). This agonist-mediated heterodimer:coactivator stoichiometry has important mechanistic implications for transactivation and in pharmacology. First, among the several functions ascribed to SRCs are the recruitment of the cellular transcriptional machinery to activated NRs (21,22,47) and as histone acetyltransferases (23). The polarity of heterodimers such as CAR:RXR on the direct repeat response element places CAR toward the 3Ј end of the promoter (48 -50). Therefore, the most likely function of the coactivator molecule bound directly to CAR is to assemble the transcriptional machinery through interactions with p300/cAMP-response element-binding protein (22). The addition of 9c to permissive NR heterodimers allows for the recruitment of a second SRC coactivator molecule directly to RXR on the 5Ј end of the promoter. From this location, the second SRC can function as and recruit other histone acetyltransferases thereby enhancing transactivational levels (Fig. 3C), also observed with the progesterone and gluco-  Table 2). Fitting of CAR(tcp):RXR(9c)⅐SRC to heterodimer:coactivator stoichiometry of 1:1 (r.m.s. deviations ϭ 0.010, top) and 1:2 models (r.m.s. deviations ϭ 0.0068, bottom). The better fit has a lower overall r.m.s. deviation, and this discrepancy is consistent across three data sets. B, transactivation by CAR:Gal4 DBD-RXR LBD and CAR:Gal4 DBD-RXR E456K LBD measured in CV-1 cells on four copies of a Gal4 binding site (mh100 ϫ 4-tk-luc) response element. Data are normalized to no-ligand activity ϭ 1.0 and absolute luciferase activity is in parentheses above each column. Statistical significance was measured by a one-tailed Welch test and differences are listed as highly significant (***, p Ͻ 0.05), moderately significant (**, p ϭ 0.198), and not significant (*, p ϭ 0.5). C, models of the NR transactivation complexes. Top, CAR: RXR⅐SRC1 and/or CAR(tcp):RXR⅐SRC1. Bottom, CAR(tcp):RXR(9c)⅐SRC1.

TABLE 2 AUC velocity data
The abbreviations use are: f/f 0 , frictional ratio; s w (20,w) , sedimentation coefficient under standard conditions; root meant square deviations reports the quality fit to the data. A higher sedimentation coefficient for 9c and tcpϩ9c complexes indicates a larger species, whereas a slightly higher frictional coefficient indicates a more extended structure.  (51). Second, this mechanism of transactivation is a distinct alternative to that proposed for the RAR: RXR heterodimer (8,43,52). In RAR:RXR, a single coactivator molecule is recruited directly to RAR upon activation by agonist in a conformation similar to CAR(tcp):RXR⅐SRC. However, the recruitment of a second coactivator to RAR(agonist): RXR(agonist) is restricted through long-range, agonist-induced conformational changes that disrupt the RXR coactivatorbinding site (53). Third, targeting coactivators for therapy is of growing interest (54,55), thus requiring a detailed knowledge of such binding events. These coactivators display different binding specificities for the receptors both independently (56) and within the heterodimer (15), thus, it is likely that the specific interactions between each SRC1 molecule and the two binding sites on CAR:RXR are distinct. The activation of CAR:RXR is not always beneficial as hepatic metabolism can convert certain therapeutic drugs to potent toxins. For instance, CAR:RXRmediated metabolism of acetaminophen results in a reactive quinone metabolite (N-acetyl-p-benzoquinone imine). This metabolic by-product promotes acute liver failure by binding to cellular macromolecules and generating reactive oxygen species (10,57). The hepatotoxic effects of cocaine are also mediated via a CAR:RXR-dependent pathway (9). As a result, the activity of CAR:RXR can have either protective or deleterious consequences to the organism depending on the particular chemical challenges faced. Also, the discoveries of endogenous RXR ligands such as polyunsaturated fatty acids (58 -60) increase the likelihood of two agonists binding the CAR:RXR heterodimer at once. Consequently, this novel assembly has important implications for the design of small molecules directed at regulating transactivation by modulating the formation and composition of the NR-coactivator assembly.