ASC-2 (activating signal co-integrator-2) is a transcriptional co-activator that mediates the transactivation of NRs (nuclear receptors) via direct interactions with these receptors. ASC-2 contains two separate NR-interaction domains harbouring a core signature motif, LXXLL (where X is any amino acid), named the NR box. Although the first NR box (NR box-1) of ASC-2 interacts with many different NRs, the second NR box (NR box-2) specifically interacts with only LXR (liver X receptor), whose transactivation in vivo requires heterodimerization with RXR (retinoid X receptor). Interestingly, RXR has been shown to enhance the LXR transactivation, even in the absence of LXR ligand via a unique mechanism of allosteric regulation. In the present study we demonstrate that LXR binding to an ASC-2 fragment containing NR box-2 (Co4aN) is enhanced by RXR and even further by liganded RXR. We also identified specific residues in Co4aN involved in its interaction with LXR that were also required for the ASC-2-mediated transactivation of LXR in mammalian cells. Using these mutants, we found that the Co4aN–LXR interaction surface is not altered by the presence of RXR and RXR ligand and that the Ser1490 residue is the critical determinant for the LXR-specific interaction of Co4aN. Notably the NR box-2, but not the NR box-1, is essential for ASC-2-mediated transactivation of LXR in vivo and for the interaction between LXR–RXR and ASC-2 in vitro. These results indicate that RXR does not interact directly with NR box-1 of ASC-2, but functions as an allosteric activator of LXR binding to NR box-2 of ASC-2.
The NR (nuclear receptor) superfamily of proteins regulates the transcription of target genes in a ligand-dependent manner by binding to specific DNA sequences, termed HREs (hormone response elements) [1–3]. NRs regulate critical biological events, including development, growth, differentiation and homoeostasis. In the absence of ligand, the apo forms of NRs repress transcription via direct interactions with transcriptional co-repressor proteins, such as N-CoR (NR co-repressor) and SMRT (silencing mediator of retinoid- and thyroid-responsive transcription), and subsequent recruitment of the histone deacetylase complex to target genes [4,5]. Upon binding to their cognate ligands, there is a conformational change within the LBD (ligand-binding domain) of the NR, which results in the release of the co-repressor complex and concomitant recruitment of various co-activator proteins [6,7].
The structure of the NR LBD consists of twelve highly conserved α-helices that are folded into a three-layered, antiparallel, α-helical sandwich, in which a central core of three helices is packed between two additional layers of helices to form the ligand-binding pocket . The ligand-dependent AF2 (activation function-2) of NRs requires helix 12 (the AF2 helix) in the C-terminus of the LBD, which adopts different conformations in the presence or absence of ligand. Genetic studies have identified transcriptional co-activators that lack specific DNA-binding activity, but are nonetheless essential for transcriptional activation. Subsequently, it has been shown that many of these transcriptional activators interact with the C-terminal ligand-dependent transactivation domains of NRs [8,9]. Co-activators such as SRC (steroid receptor co-activator) family proteins (p160 family), p300/CBP [CREB (cAMP-response-element-binding protein)-binding protein], p/CAF (p300/CBP-associated factor), TRAP (thyroid-hormone-receptor-associated protein)/DRIP (vitamin-D-receptor-interacting protein), and many others are thought to function by bridging transcription factors and the basal transcription apparatus, and/or by remodelling chromatin structures . The specific interaction of co-activator proteins with the LBDs of agonist-bound NRs is mediated by a core signature motif, the LXXLL motif, or NR box, within the IDs (NR-interaction domains) of co-activator proteins [6,7]. Based on structural analysis, the LXXLL motif adopts a two-turn helical conformation that allows it to interact with a hydrophobic groove on the surface of the agonist-bound NR LBD. NR box residues make direct contacts with helices 3, 4, 5 and 12 of the NR LBD, and a lysine residue in helix 3 and a glutamate residue in helix 12 form a charge clamp that interacts strongly with the LXXLL motif of the co-activator protein . The AF2 helix of apo- or antagonist-bound NRs is positioned outside of the LBD, which creates space for the binding of co-repressor motifs, which are longer than co-activator motifs. Thus the NR LBD acts as a signal-responsive regulatory module, adopting distinct conformations in the unbound, agonist- or antagonist-bound state [6,7].
LXRs (liver X receptors) belong to the NR superfamily and regulate gene expression in response to oxysterols. They play a critical role in cholesterol homoeostasis through the regulation of genes involved in cholesterol transport, catabolism and triacylglycerol synthesis . When bound to oxysterols and other synthetic agonists, LXRs activate transcription by recruiting co-activator proteins [11,12]. To date, two members of the LXR family have been identified, namely LXRα (NR1H3) and LXRβ (NR1H2). LXRα is mainly expressed in the liver, whereas LXRβ is ubiquitously expressed [13,14]. LXRs are ligand-dependent transcription factors that function as heterodimers with RXR (retinoid X receptor) . The LXR–RXR heterodimer binds to a sequence called the LXRE (LXR-responsive element) in the regulatory regions of target genes. LXRE is known to consist of direct repeats of a core motif, AGGTCA, separated by four nucleotides (termed DR4) [11,15]. LXR–RXR heterodimer bound to the DR4-response element can be transcriptionally activated by either LXR and RXR ligands, which are referred to as ‘permissive heterodimers’ . LXR can also be activated by heterodimerization alone, in the absence of ligand, via a mechanism termed ‘dimerization-induced activation’. This is a unique mechanism of NR activation that has not been described for NRs other than LXR . Dimerization-induced activation requires the AF2 domain of LXR, but not the RXR AF2 domain. It has been shown that heterodimerization is sufficient to recruit SRC-1 to the DNA-bound LXR–RXR heterodimers . Thus the interaction of RXR with LXR appears to induce a conformational change to the LXR-LBD, reminiscent of the structural changes induced by the binding of ligand, termed the phantom-ligand effect . In this model of LXR transactivation, which encompasses the permissive heterodimer model and the allosteric effects of RXR heterodimerization, the LXR–RXR heterodimer is activated by dimerization, and exhibits dual-ligand permissiveness and synergism [17–19]. However, the precise role of RXR in the interaction of the LXR–RXR heterodimer with co-activators remains poorly understood. In particular, it is unclear whether RXR is a direct binding partner of specific NR boxes of co-activators or functions as an allosteric activator of LXR by stimulating the interaction of LXR with co-activators.
ASC-2, also referred to as NRC (NR co-regulator) , RAP250 (250-kDa receptor-associated protein) , TRBP (thyroid hormone receptor-binding protein)  and PRIP [PPAR (peroxisome-proliferator-activated receptor)-interacting protein] , is a transcriptional co-activator that was originally isolated as a gene AIB3, amplified in human cancer . Since then, ASC-2 has emerged as an important co-activator for NRs, and a number of other transcription factors, such as c-Fos, c-Jun, CREB, NF-κB (nuclear factor κB), ATF-2 (activating transcription factor-2), heat-shock factors, E2F-1, SRF (serum response factor) and the Rb (retinoblastoma gene product) . The interaction of ASC-2 with NRs is mediated by two IDs, each of which contains the canonical LXXLL motif required for the interaction. In the yeast two-hybrid system, the affinity of ASC-2 for ligand-bound NRs is similar to, or higher than, that of SRC-1 . The first NR box (NR box-1) of ASC-2 binds to a broad range of NRs, whereas the second NR box (NR box-2), which is located in the C-terminal region of ASC-2 (amino acids 1491–1495), appears to interact specifically with LXRα and LXRβ . TG (transgenic) mice expressing a DN (dominant-negative) form of ASC-2 encompassing NR box-1 (DN1, amino acids 849–929) inhibited the recruitment of endogenous ASC-2 to NRs. These TG mice exhibited a plethora of developmental phenotypes and other abnormalities . A DN form of ASC-2 encompassing NR box-2 (DN2, amino acids 1431–1511) exerted a potent DN effect on the transcriptional activity of LXR in transfection assays . The livers of DN2 TG mice displayed a phenotype of dysregulated lipid metabolism, highly homologous with that of LXR−/− mice. Thus ASC-2 appears to act as a physiological regulator of LXR and plays a pivotal role in liver lipid metabolism through the direct interaction of NR box-2 with LXRs .
In the present study we examined the role of RXR heterodimerization in the interaction of LXR with NR box-2 of ASC-2. We identified the key amino acids that contribute to the interaction, using the one- plus two-hybrid system, a recently developed selection method for specific missense mutations disrupting protein–protein interactions . From these mutational analyses, we have provided the detailed molecular basis for the interaction of the LXR–RXR heterodimer with the NR box-2 of ASC-2.
NR box-2 fragments of human ASC-2, Co4aN (amino acids 1429–1511) and Co4a (amino acids 1429–1730), were generated by PCR using the pcDNA3HA-ASC-2 plasmid as a template . The fragments were inserted into pRS424UB42-GBD [where GBD is Gal4-DBD (DNA-binding domain)] for screening of mutants and the yeast-interaction assay , and into pGEX4T-1 (Amersham Pharmacia Biotech) for the GST (glutathione transferase) pulldown assay, using appropriate restriction enzyme sites. The full-length ASC-2 mutants (LR1, LR4 and NR box-2 mutants) were created using QuikChange® (Stratagene) or Muta-Direct™ (iNtRON) site-directed mutagenesis and were cloned into the pcDNA3HA vector for mammalian expression and in vitro translation. Human LXRα (hLXRα) and hLXRβΔN (amino acids 59–446) fragments were obtained from pEG202-LXR constructs by restriction digestion and inserted into pRS325LexA  (for mutant screening and the yeast-interaction assay), pRS323U (for the yeast HRE assay), pGEX4T-1 (for the GST pulldown assay) and pcDNA3HA (for in vitro translation and transfection) using appropriate restriction sites. For the expression of RXR in yeast, a hRXRα gene fragment was obtained from YEp-hRXR (used for the yeast HRE system) and cloned into pRS426U (for the yeast-interaction assay). The pRS323U and pRS426U vectors were constructed by inserting the URA3 promoter region into pRS323 and pRS426 respectively, using appropriate restriction enzyme sites. The pSH18-34-ADE reporter plasmid was also constructed by substitution of an ADE2 gene fragment for the URA3 gene of pSH18-34 , using appropriate restriction enzyme sites. mRXRγ (mouse RXRγ) DNA was obtained by restriction enzyme digestion of pEG202-mRXRγ, and then inserted into pcDNA3HA for in vitro translation. To generate the ΔAF2 constructs, fragments of hLXRα and hRXRα that lacked the AF2 helix region (corresponding to the C-terminal 11 or 14 amino acids respectively) were generated by PCR, and cloned into pRS325LexA and pRS426U respectively. The hRARα (retinoic acid receptor) and hTRβLBD (where TR is thyroid hormone receptor) DNAs were subcloned into pcDNA3HA and pEG202 vectors for in vitro translation and the yeast two-hybrid assay respectively. All constructs were verified by DNA sequencing.
For one- plus two-hybrid screening of Co4aN mutants that were defective in LXR binding, yeast strain YOK400 (MATα, leu2, trp3, ura3, lexAop-LEU2, UASGAL-HIS3) was constructed by genetic manipulation of strain EGY48, as described previously . Yeast strain YLS500 (MATα, ade2, leu2, lys2, trp1, ura3, UASGAL-HIS3) was generated by genetic manipulation of strain YPH500 and used for the screening of mutant Co4aN constructs defective in LXR–RXR binding. Strains with proper integration of the UASGAL region upstream of the chromosomal HIS3 locus were selected based on their ability to grow on galactose medium lacking histidine, but not on glucose medium. Transformants that were highly resistant to 3AT (3-amino-1,2,4-triazol; a competitive inhibitor of His3p) were chosen as strains YOK400 (50 mM) or YLS500 (10 mM).
For the yeast LexA system, we used the yeast strains YPH500 (MATα, ade2, lys2, his3, trp1, ura3, leu2) and EGY48 (MATα, his3, trp3, ura3, lexAop-LEU2), which already harbour pSH18-34 (8xlexAop-LacZ reporter). These strains were then co-transformed with bait vectors expressing LexA-fusion proteins of the indicated NRs (pRS325LexA derivatives) and prey vectors expressing B42–GBD fusion proteins of Co4aN, using a standard lithium acetate method . For the yeast HRE system, yeast strain YPH500 harbouring the DR4-LacZ reporter gene was co-transformed with free forms of NR (pRS323U-hLXRα or YEp-hRXRα) and prey vectors expressing B42–Co4aN–GBD triple fusion. Plate and liquid assays for β-galactosidase activity for at least three transformants were performed as described previously . Experiments were repeated multiple times with similar results.
Mutagenic PCR and one- plus two-hybrid screening
Random mutagenesis of Co4aN and screening using the one- plus two-hybrid system were carried out as previously described . Briefly, mutagenic PCR was carried out using Taq polymerase and pRS424UB42-Co4aN-GBD as the template in the presence of 0.1 mM MnCl2. The cycling parameters were 30 cycles of 94 °C for 30 s, 55 °C for 30 s and 72 °C for 60 s. The two oligonucleotides used as universal primers were designed so that the resultant mutated fragments all contained approx. 100 bp of flanking sequence, identical with the linearized gap plasmid . To construct the library of randomly mutated Co4aN variants, we used a single-step method based on in vivo gap-repair . The products of the mutagenic PCR reaction (4 μg) were co-transformed with the gap plasmid (1 μg) into strain YOK400, which carried the reporter plasmid pSH18-34 and pRS325-LexA-hLXRβΔN, or strain YLS500, which carried the reporter plasmid pSH18-34-ADE and the indicated NR plasmids (pRS325-LexA-hLXRβΔN and pRS426U-hRXRα), for the screening of Co4aN mutants that were defective in LXRβ homodimer or LXRβ–RXR heterodimer binding respectively. His+ transformants of YOK400 or YLS500 were obtained after a 4-day incubation period at 30 °C on glucose medium lacking histidine, with or without 10 mM 3AT respectively, to select for missense mutants of Co4aN. Approx. 1000 transformants were selected and transferred on to a His− plate containing X-gal (5-bromo-4-chloro-indol-3-yl β-D-galactopyranoside), and colonies that showed a white or weak blue colour were designated as non-interactors. Prey vectors were rescued from the candidate mutants and individually tested for the integrity of the prey protein (one-hybrid test) and for the non-interactor phenotype (two-hybrid test), as previously described . Final candidate mutants were subjected to DNA sequencing to identify the mutation site(s).
In vitro GST pulldown assay
pGEX4T-1 derivatives expressing Co4a and hLXRα were introduced into Escherichia coli strain DH5α. Overexpression of GST alone or GST-fusion proteins was induced by incubation for 3 h in the presence of 0.1 mM IPTG (isopropyl β-D-thiogalactoside), and proteins were purified using glutathione–agarose beads (Promega) according to the manufacturer's instructions. Full-length ASC-2 derivatives or the indicated NRs (hLXRα, hLXRβΔN, mRXRγ, hRARα, hTRβLBD) were synthesized by in vitro translation using pcDNA3HA-based expression constructs, and the TNT transcription-coupled translation system (Promega). Radiolabelled proteins were added to equivalent amounts of GST or GST-fusion protein (2–4 μg) bound to glutathione–agarose beads pre-equilibrated with buffer A [150 mM Tris/HCl (pH 7.9), 5% glycerol, 1 mM EDTA, 1 mM dithiothreitol, 1× protease inhibitor (1 mM PMSF, 2 mM benzamidine, 5 μg/ml of leupeptin and 5 μg/ml pepstatin), 0.01% NP40 (Nonidet P40) and 150–200 mM KCl] in a final volume of 250 μl as described previously . The beads were washed three times in equilibration buffer (buffer A) and bound proteins were analysed by SDS/PAGE and autoradiography.
Cell culture and transient transfection assay
HeLa cells were grown in 24-well plates with Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum for 24 h and transiently transfected with 100 ng of LacZ expression vector pRSV-β-gal and 100 ng of p2xLXRE-Luc reporter plasmid, in combination with the indicated amounts of various mammalian expression plasmids using SuperFect® reagent (Qiagen). The total amounts of expression vectors were kept constant by adding appropriate amounts of pcDNA3. Luciferase assays and normalization of the results were performed as described previously . Similar results were obtained in more than two similar experiments.
Preparation of nuclear extracts, immunoprecipitation and Western blot analysis
All protein manipulations were carried out at 4 °C in the presence of 1× protease inhibitors. After transient transfection of HeLa cells with the indicated plasmids, nuclear extracts were prepared as previously described . Equal amounts of protein samples were pre-cleared by centrifugation (6000 g for 10 min at 4 °C), mixed with 30 μl of agarose beads (Sigma) conjugated to an anti-ASC-2 monoclonal antibody (A3C1), and incubated at 4 °C overnight with rotation. Beads were washed three times with IP-150 buffer [20 mM Hepes (pH 7.7), 150 mM NaCl, 2.5 mM MgCl2, 0.05% NP40, 10% glycerol, 1 mM dithiothreitol and 1× protease inhibitors]. Bound proteins were eluted with 0.1 M glycine/acetate (pH 3.0) and precipitated with 10% trichloroacetic acid. Protein pellets were resuspended in 2× SDS sample buffer, separated by SDS/PAGE (6% gel), and transferred on to Hybond-ECL (enhanced chemiluminescence) nitrocellulose membrane (Amersham Pharmacia Biotech). Membranes were probed with an anti-HA (haemagglutinin) monoclonal antibody (12CA5) and immunoreactive bands were developed with the Amersham ECL kit according to the manufacturer's instructions.
Allosteric activation of LXR binding to NR box-2 of ASC-2 by RXR heterodimerization
To investigate the role of RXR heterodimerization in LXR–ASC-2 binding, we examined the interaction of LXRα with ASC-2 Co4aN, (amino acids 1429–1511) (Figure 1A), which contains the NR box-2 of ASC-2, using the yeast two-hybrid assay, in the presence or absence of RXR and/or the cognate ligands for LXR and RXR. LexA–LXRα fusion protein was used as bait, and a B42–Co4aN–GBD triple-fusion protein was expressed as prey. The expression vector for B42–GBD was devised for the operation of the one- plus two-hybrid system, but also can be used for the conventional two-hybrid interaction assay using the lexA-LacZ reporter . ASC-2 Co4aN specifically interacted with LXRs, but not with other NRs, or with RXR, as reported previously (results not shown) . Notably, ASC-2 Co4aN had a similar affinity for LXR in yeast either in the presence or absence of its cognate ligand, 22R-HC [22(R)-hydroxycholesterol] (see below).
Allosteric activation of LXR–Co4aN binding by RXR heterodimerization
The LXR–RXR heterodimer has been shown to be transcriptionally activated by both LXR and RXR ligands, either alone or in combination, and also by heterodimerization itself. The latter mechanism of dimerization-induced activation is also called the ‘phantom-ligand effect’ . We used the yeast two-hybrid assay to examine whether the LXR–Co4aN interaction is stimulated by RXR and/or the cognate RXR ligand 9cRA (9-cis retinoic acid). The relatively weak interaction of LXRα and Co4aN was stimulated approx. 5-fold by LXR dimerization with RXR, and an additional 2-fold upon the addition of 9cRA (Figure 1B). These results suggested that the stimulatory effect of RXR on the interaction of LXR with the NR box-2 of ASC-2 is two-fold: a dimerization-induced, ligand-independent allosteric (phantom) effect, and a ligand-dependent allosteric effect. These modes of interaction of the LXR–RXR heterodimer with the NR box-2 of ASC-2 are reminiscent of the known modes of LXR transactivation by RXR and RXR ligands. Previous reports have suggested that the AF2 helix of LXR is essential for the transcriptional activity of the LXR–RXR heterodimer, whereas deletion of the AF2 helix of RXR has a minor effect. However, the AF2-deletion mutant of RXR (RXRΔAF2), which still bind 9cRA, does not elicit the stimulatory effect of 9cRA (the ligand-dependent allosteric effect of RXR) . Thus we next examined the effect of deleting the AF2 helix of LXR or RXR on the interaction of LXR–RXR with Co4aN. As shown in Figure 1(B), the deletion of RXR-AF2 did not affect the interaction of LXR with Co4aN in the absence of 9cRA (ligand-independent phantom effect), but abolished the 9cRA-stimulatory effect (ligand-dependent phantom effect). In contrast, the deletion of LXR-AF2 completely abolished the interaction of LXR with Co4aN. Overall, these results corroborate previous studies .
To confirm the ligand-independent and -dependent allosteric effects of RXR on the LXR–Co4aN interaction, we performed an in vitro pulldown assay using a GST-fusion protein of the ASC-2 NR-box2 fragment and in vitro synthesized, radiolabelled LXR and/or RXR (Figure 1C). In this experiment, we used Co4a instead of Co4aN (Figure 1A) because Co4aN did not readily interact with LXR in vitro (results not shown). Consistent with these results, the level of interaction of Co4a with LXR or LXR–RXR was approx. 10-fold higher than that of Co4aN in the yeast two-hybrid assay (results not shown). These results suggest that the C-terminal region of Co4a probably has a stabilizing effect or essential roles in its interaction with LXR. As shown in Figure 1(C), Co4a weakly interacted with LXRα in vitro, and this interaction was stimulated by 22R-HC (Figure 1C), although the stimulatory effect of 22R-HC was hardly observed in yeast (Figure 1B). LXR–Co4a binding was stimulated by the addition of RXR, and further stimulated by 9cRA. These results demonstrate that the positive roles of RXR and liganded-RXR in the specific interaction between LXR and the NR box-2 of ASC-2 can be demonstrated in vitro as well.
Isolation of Co4aN mutants defective in the interaction with LXR homodimer
Since the NR box-2 of ASC-2 has been characterized as the specific LXR-interaction motif, we were interested in identifying residues in this region which play critical roles in the LXR interaction. To select Co4aN mutants that were specifically defective in LXR homodimer binding, we used a novel mutant screening system named the ‘one- plus two-hybrid system’. This system was recently developed as a method for the efficient selection of missense mutations that specifically disrupt protein–protein interactions . This system operates as a dual reporter system in the same cell; one is the modified one-hybrid system that positively selects missense mutants from a randomly generated mutant library, and the other is the two-hybrid reporter system which is used as a secondary screen for the identification of interaction-defective mutants among the isolated missense mutants.
LexA–LXRβ fusion protein was expressed in yeast as bait (pRS325LexA-hLXRβΔN), and the Co4aN triple-fusion protein was expressed as the prey protein (B42–Co4aN–GBD). We used PCR-mediated random mutagenesis and a gap-repair recombination method to generate a library of randomly mutated Co4aN fragments . Co4aN mutants were first generated via error-prone PCR in the presence of 0.1 mM MnCl2 and then co-transformed with linearized gap plasmid into strain YOK400, which harboured LexA–LXRβΔN, the episomal two-hybrid reporter plasmid (lexAop-LacZ), and the chromosomal one-hybrid reporter gene (UASGAL-HIS3). All of the transformants were grown in synthetic glucose medium containing 10 mM 3AT, but no histidine, to positively select intact prey fusion proteins. Among the surviving transformants, non-interacting mutants were selected by isolating white colonies on X-gal plates. We isolated white yeast colonies on X-gal plates from the 732 transformants that survived in the first positive screening for missense mutations. The plasmids encoding mutant prey fusion proteins were rescued and re-transformed into strain EGY48, which expressed LexA–LXRβΔN, and strain EGY-LG harbouring pLGSD5 reporter (UASGAL-LacZ), to confirm the results of the two-hybrid (non-interactors) and one-hybrid (integrity of the prey fusion protein) assays respectively. Candidates that exhibited white colour in the two-hybrid assay and blue colour upon re-transformation into strain EGY-LG were subjected to DNA sequencing. Most of the mutants contained a single-point missense mutation and no truncation mutants were isolated. Mutants that harboured double mutations were subjected to PCR-based site-directed mutagenesis to re-engineer each single-point mutation. As shown in Figure 2(A), most of the mutation sites were located within the LXXLL motif of Co4aN. We isolated two substitution mutants at position +3 (the first L of the LXXLL motif is ascribed as +1), which had not previously been implicated as a critical residue for NR interactions; Q+3L and Q+3R. Mutations were also found in the N-terminal flanking region of the LXXLL motif (P−3L and S−1P), but not in the C-terminal region. Although proline substitution mutants are generally regarded as uninformative because proline is known to disrupt helices, this result indicated that the N-terminal residues immediately flanking the LXXLL core motif of Co4aN are involved in its interaction with LXRβ. Using the yeast two-hybrid system, we confirmed that the Co4aN mutants were defective in LXR binding, independent of LXR subtype (Figure 2B). There were differences in the binding strengths of LXRα and LXRβ for Co4aN, which may be due to the presence or absence of the N-terminal region of the LXR protein, as LXRβ was expressed as an N-terminal deletion (amino acids 59–446). Similar results were observed in the in vitro binding assay (Figure 2C). We also tested the effect of these mutations on the interaction of Co4a with LXR, and observed a similar pattern of binding in yeast (results not shown).
Isolation of Co4aN mutants defective in LXRβ binding
Isolation of Co4aN mutants defective in binding to the LXR–RXR heterodimer
RXR heterodimerization stimulates the interaction of LXR with the NR box, possibly through conformational changes in the LXR LBD. We next examined whether RXR-mediated enhancement of LXR–Co4aN binding involved changes in the interaction specificity of Co4aN. The Co4aN–LXR or Co4aN–LXR–RXR interface should be similar, since RXR acts as an allosteric effector rather than a direct binding partner of Co4aN (Figure 1B). To test this hypothesis, we screened the Co4aN library for mutants that were defective in binding to the LXRβ–RXR heterodimer. We expressed LXRβΔN as a LexA-fusion protein along with native RXR, and randomly mutated Co4aN fragments were expressed as B42–GBD triple-fusion proteins. After the positive selection of missense mutants, a total of 1050 His+ transformants were subsequently screened on X-gal plates to identify loss-of-interaction mutants. From these experiments, we isolated a total of 12 Co4aN mutants defective in binding to the LXRβ–RXR heterodimer.
Figure 3(A) shows the mutation sites and the corresponding amino acid changes for the LXRβ–RXR-binding mutants. The mutational profile of these mutants was similar to that of LXR homodimer binding mutants. Substitution of T−2 (T−2P) was isolated only in the screen for LXRβ–RXR-binding mutants. We examined the ability of these mutants to bind to LXRα and LXRβ, either alone or in combination with RXR, using the yeast two-hybrid assay. Although we identified several mutants (P−3Q, T−2P, S−1L and L+1V) only in the screen for LXRβ–RXR-binding mutants, they were also defective in binding to LXR homodimers, with the exception of S−1L, which retained the ability to bind to LXRα (approx. 20% of the level of wild-type) (Figure 3B). All of the mutants, with the exception of S−1L and Q+3R, were defective in binding to LXR–RXR heterodimers (Figure 3C). S−1L and Q+3R exhibited a relatively stronger level of interaction with LXRα–RXR compared with LXRβ–RXR, which corresponded to the isoform of LXR used in the original screen. P−3L, which was only identified in the screen for LXR homodimer-binding mutants, was also defective in LXR–RXR binding (Figure 3C). Overall, these results strongly suggest that the interaction of LXR with NR box-2 of Co4aN is mediated by specific amino acids within or closely flanking NR box-2 of ASC-2, and the specificity of these determinants is not altered upon allosteric activation by RXR, and are independent of LXR subtype.
Co4aN mutants defective in binding to LXR–RXR heterodimers
In vitro and HRE assays of the interaction of Co4aN mutants with LXR–RXR heterodimer
To investigate the effect of RXR ligand on the interaction of Co4aN mutants with the LXR–RXR heterodimer, we performed in vitro binding assays using GST-fused Co4a variants and in vitro synthesized LXRα–RXR or LXRβ–RXR, in the presence or absence of RXR ligand (Figure 4A). We selected four Co4aN mutations (P−3L, S−1P, L+1F and Q+3L) and introduced each mutation into Co4a. Wild-type Co4a alone interacted weakly with LXR–RXR heterodimers. Binding was stimulated by 22R-HC, and further stimulated by 9cRA (Figure 4A, see lanes 1, 2 and 7), indicating that there is a synergistic effect of LXR and RXR ligands on the Co4a-interaction with LXR–RXR. In the absence of 9cRA, binding of all Co4a mutants to LXR–RXR was very low to undetectable (Figure 4A, lanes 3–6). In the presence of 9cRA, all four mutants exhibited significant defects in LXR–RXR binding (Figure 4A, lanes 8–11). The residual interaction between Co4a mutants and LXR–RXR appeared to be non-specific, as RXR alone was able to bind to Co4a in a 9cRA-independent manner (Figure 4A, lanes 12 and 13). These results suggest that specific residues of Co4aN are involved in its interaction with LXR, independent of LXR subtype (LXRα compared with LXRβ), the dimeric state of LXR (homodimer compared with heterodimer), or the presence or absence of RXR ligand. Thus the mechanism underlying the ligand-independent or -dependent RXR phantom effect does not appear to involve a change in the specificity of the interaction of LXRs with NR box-2 of ASC-2.
In vitro and yeast HRE assays for the interaction of LXR–RXR with Co4aN mutants
Next, we examined the effect of Co4aN mutations on LXR–RXR-binding in the presence or absence of RXR ligand in the yeast HRE-interaction assay. The yeast HRE system is more physiologically relevant than the yeast two-hybrid system, because it utilizes the LacZ reporter gene under the control of a natural response element recognized by native forms of NRs. We examined the interaction of DR4-bound LXRα–RXRα heterodimers with various Co4aN mutants, which were expressed as B42–GBD triple-fusion proteins. As shown in Figure 4(B), all of the Co4aN mutants, with the exception of S−1L, exhibited significant defects in LXR–RXR binding, similar to the results of the yeast LexA two-hybrid system, and this pattern was not changed in the presence of 9cRA (compare Figures 3C and 4B). The interaction of wild-type Co4aN with native heterodimers was 5-fold lower than that with LexA–LXR–RXR in the absence of 9cRA. In contrast, the 9cRA responsiveness of native heterodimers was much stronger than that of LexA–LXR–RXR (2-fold compared with 5.2-fold; compare Figures 1B and 4B). It is possible that the fusion of LexA to LXR may lead to a partial activation of the transcriptional activity of LXR by allosterically altering the conformation of LXR to a ligand-bound form.
Identification of the Ser1490 residue (S−1) of NR box-2 as a key determinant for the LXR-specificity of Co4aN
The S−1L mutant of Co4aN consistently exhibited a relatively strong interaction with LXR–RXR as compared with other NR box-2 mutants (Figures 3C and 4B). To confirm the mutational effect of S−1L in vitro, we performed GST pulldown assays. Consistent with observations in yeast, the S−1L mutant, but not the S−1P mutant, showed significant binding activity to LXR–RXR in vitro, regardless of the LXR subtype or the presence of cognate ligands (Figure 5A).
Identification of the Ser1490 residue (S−1) as a key determinant for LXR-specificity of Co4aN
Based on sequence analysis of the flanking regions of the core LXXLL motifs, Chang et al.  and Bramlett and Burris  have grouped the NR boxes of various co-activators into four classes. For example, NR box-1 of ASC-2 is classified as a typical class II motif. In this regard, the NR box-2 of ASC-2 is unique in its primary sequence which does not resemble any of the known NR box motifs present in other NR co-activators or discovered by affinity selection using combinatorial phage peptide libraries. Since threonine and serine residues at −1 (Ser1490) and −2 positions (Thr1489) respectively, are distinguishing features of the ASC-2 NR box-2, substitution of Ser1490 to a leucine residue (S−1L mutant) creates a typical class III NR box sequence [(S/T)ΨLXXLL; Ψ, hydrophobic residue]. Class III NR boxes are found in various co-activator proteins [e.g. NR box-4 of SRC-1, NR box-3 of SRC-2 and SRC-3, and PGC-1 (PPARγ co-activator 1)] and interact with a broad range of NRs. This observation explains why the S−1L mutant interacts relatively strongly with LXR or LXR–RXR. In fact, it was reported that LXR can interact with co-activator fragments containing a class II or class III NR box .
From these results we hypothesized that the S−1L mutant of Co4aN would interact with many NRs, including LXR, if the Ser1490 residue of ASC-2 contributes to the LXR-specific interaction of Co4aN. To test this possibility, the Co4aN fragment harbouring the S−1L mutation was tested for its ability to bind to TR and RAR in a yeast two-hybrid assay. As shown in Figure 5(B), wild-type Co4aN failed to interact, or only barely interacted, with these NRs in yeast, as compared with the Co2cN fragment containing the NR box-1 of ASC-2 (a positive control). In particular, the S−1L mutant showed significant binding activity to RAR and TR (30–40% the level of Co2cN), whereas another mutant with a substitution of Ser1490 to a proline residue (S−1P) completely abolished its interaction with these NRs. We also observed that other Co4aN mutants could not interact with RAR or TR (results not shown). To confirm the effects of the S−1L mutation in vitro, in vitro synthesized RAR and TR were tested for their binding to GST-fused Co4a fragments containing the S−1L or S−1P mutation in the presence or absence of each ligand (Figure 5C). Consistent with the yeast data, the S−1L mutant acquired significant binding activity for RAR and TR only in the presence of ligands as compared with Co2cN binding to these NRs. As predicted, neither the wild-type Co4a nor the S−1P mutant could bind to RAR or TR in vitro (Figure 5C). Collectively, these results strongly suggest that the Ser1490 residue located at −1 position of LXXLL motif plays an essential role in establishing the LXR-specificity of NR box-2 of ASC-2.
Confirmation of mutational effects of the NR box-2 mutations within the context of full-length ASC-2
To confirm the effects of NR box-2 mutations in a physiologically relevant system, we introduced the NR box-2 mutations into full-length ASC-2 and tested the abilities of these ASC-2 mutants to serve as co-activators for LXR–RXR transcriptional activity in mammalian cells. A reporter gene under the control of 2× LXRE was transfected into HeLa cells together with an expression vector for LXR. Cells were then incubated in the presence or absence of ligand, and reporter gene activity was measured. Ectopic expression of LXR in HeLa cells resulted in a 2-fold activation of the LXRE-Luc reporter gene in the absence of ligand. This activation was further stimulated by treatment with ligand, resulting in dual-ligand activation (22R-HC or 9cRA) and synergism (both ligand) as reported previously (Figure 6A). The transactivation probably involved LXR–RXR heterodimer formation with an endogenous RXR, since it was reported that LXRs can bind to the LXRE as heterodimers with RXR. In the presence of both ligands, co-transfection of wild-type ASC-2 with LXR enhanced reporter gene activity approx. 3-fold, which corresponds to the well-known stimulatory effect of the NR co-activator in transfection assays. Interestingly, all of NR box-2 mutants, except for S−1L, exhibited significant defects in the stimulation of LXR–RXR transactivation (Figure 6B), which is in good agreement with their mutational effects in yeast using the small fragment (Co4aN). As predicted, the S−1L mutant had the ability to enhance LXR–RXR transactivation to 70–80% of the wild-type level. These results indicate that the residues that were identified as being required for the LXR–Co4aN interaction in yeast are also essential for ASC-2-mediated stimulation of LXR–RXR transcriptional activity in mammalian cells. We tried to check the expressions of ASC-2 constructs from crude extracts by Western blotting, but failed owing to the low expression levels of HA–ASC-2 proteins under these experimental conditions (results not shown). Therefore we prepared the nuclear extracts from the transfected cells and immunoprecipitated ASC-2 proteins with anti-ASC-2 antibody followed by Western blot analysis with an anti-HA antibody (Figure 6B, lower panel). From this result, we could confirm that the mutational effects of ASC-2 mutants were not due to the differences in their expression levels.
Functional defects of NR box-2 mutants of ASC-2 correlate with their reduced binding for LXR–RXR heterodimer
To further delineate the functional defects of NR box-2 mutants, we examined the physical interaction between these ASC-2 mutants and the LXR–RXR heterodimer by co-immunoprecipitation of the transiently expressed proteins. However, unfortunately, we could not observe their physical interaction via co-immunoprecipitation, despite the several attempts using various combinations of antibodies or tags. We presumed that this might be due to the low level of ASC-2 expression (as mentioned above) and/or that the ASC-2–LXR interaction in solution is too weak to be detected by co-immunoprecipitation. For that reason, we employed the GST pulldown analysis instead of a co-immunoprecipitation assay. Full-length ASC-2 derivatives and RXR were synthesized in vitro and tested for their binding to GST-fused LXR in the presence of both ligands (22R-HC and 9cRA). As shown in Figure 6(C), radiolabelled RXR strongly bound to GST–LXR in all cases, indicating efficient heterodimerization of RXR with LXR (Figure 6C, +RXR). Consistent with the transfection results in HeLa cells, NR box-2 mutants exhibited a significant but incomplete loss of LXR–RXR binding, except for the S−1L mutant (see the quantified data in Figure 6D). From these results, we concluded that P−3, T−2 and Q+3 residues, in addition to the LXXLL motif, are required for the co-activator function of ASC-2 in the potentiation of LXR transactivation because they are important determinants of the LXR–ASC-2 interaction.
NR box-2, but not NR box-1, is required for ASC-2-mediated transactivation of LXR and for the LXR–ASC-2 interaction in vitro
The results of the above study consistently indicated that NR box-2 mutants of ASC-2 were defective in both binding to LXR and mediating LXR transactivation, even though these mutants still contained a functional NR box-1 that exhibited strong affinity and preference for RXR . This suggests a model for the interaction of ASC-2 with LXR–RXR heterodimers: LXR exclusively interacts with NR box-2 of ASC-2, whereas RXR functions as a phantom-ligand effector, and does not directly bind to NR box-1. To directly test this hypothesis, we carried out transient transfection assays using two ASC-2 mutants, LR1 and LR4, in which the core sequence motif of NR box-1 and -2 was mutated to LXXAA respectively (Figure 7A). The level of reporter gene activity was enhanced by more than 3-fold following overexpression of wild-type ASC-2. This was absolutely dependent on NR box-2, but not on NR box-1, as judged by the impaired co-activator function of LR4 but not of LR1 (Figure 7B). This result confirms the role of RXR as an allosteric activator for the LXR–NR box-2 interaction, rather than as a direct binding partner of NR box-1.
NR box-2, but not NR box-1, is required for ASC-2-mediated transactivation of LXR and for the LXR–ASC-2 interaction in vitro
To further establish the role of the NR box-2 in the LXR-co-activator function of ASC-2, we performed a GST pulldown assay to examine the interaction between GST–LXR and in vitro synthesized RXR and ASC-2 LR1 and LR4 (Figure 7C). Radiolabelled wild-type ASC-2 and LR1, but not LR4, bound to GST–LXR in a 22R-HC-dependent manner (Figure 7C, lanes 1 and 2). We also observed that radiolabelled RXR efficiently bound to GST–LXR, indicating that LXR–RXR heterodimers were successfully formed, in the presence or absence of ligands or ASC-2 (Figure 7C, lanes 3–6, RXR band). The RXR-mediated enhancement of LXR binding to either wild-type ASC-2 or LR1 was observed in the presence or absence of LXR ligand (Figure 7C, compare lanes 1 and 3 with lanes 2 and 4). In addition, the interaction of wild-type ASC-2 and LR1 with LXR–RXR was stimulated by the addition of each ligand, and was synergistically enhanced in the presence of both ligands (compare lanes 3–6). These results are correlated with the ligand-dependent and -independent allosteric effects of RXR on LXR–ASC-2 binding in vitro. In contrast, LR4 binding to LXR–RXR was inefficient, and was not stimulated by 22R-HC, but was slightly enhanced by liganded RXR (compare lanes 3–6; also see the quantified results in Figure 7D). This result was somewhat inconsistent with the results of the mammalian system, and may be due to a low level of binding of the relatively higher amount of RXR (present as a LXR–RXR heterodimer) to the NR box-1 of ASC-2.
Collectively, these results provide evidence that the specific role of RXR in the interaction between ASC-2 and LXR–RXR is the enhancement of LXR binding to the NR box-2 of ASC-2 via allosteric activation (ligand-independent and -dependent phantom effects) of LXR, rather than a direct interaction with the NR box-1 of ASC-2.
In vivo, LXRs function as obligate heterodimers with RXR and regulate target genes by binding to their DR4 elements. Interestingly, LXR–RXR heterodimers have been shown to be activated by heterodimerization itself, and then further activated by their cognate ligands . ASC-2 has been shown to specifically interact with LXRs via its NR box-2 . A fragment encompassing ASC-2 NR box-2 (DN2) actively represses LXR transactivation in co-transfection assays . Moreover, DN2-TG mice exhibit phenotypes that are highly similar to those previously observed for LXRα−/− mice . Although these results suggest that ASC-2 has a pivotal role as a physiological co-activator protein for LXR transactivation, the underlying molecular mechanism of the interaction of LXR with ASC-2 is largely unknown. In the present study we examined the role of RXR heterodimerization in LXR binding to the NR box-2 of ASC-2 by identifying the molecular determinants of NR box-2 required for its interaction with LXR in the presence or absence of RXR. We demonstrated that the AF2 of LXR is essential for the interaction of the heterodimer with Co4aN. The RXR AF2 was required for the 9cRA-dependent, but not the 9cRA-independent allosteric effect of RXR on Co4aN binding to LXR. We also demonstrated that the specific residues of Co4aN involved in its interaction with LXR are also required for its interaction with the LXR–RXR heterodimer, regardless of the presence of 9cRA. Finally, we found a pivotal role of the NR box-2 of ASC-2 in transactivation of LXR–RXR in mammalian cells, and in the binding of full-length ASC-2 to the LXR–RXR heterodimer in vitro. These results suggest that RXR does not directly bind to the NR box-1 of ASC-2 but rather enhances the ability of LXR to bind to the NR box-2 of ASC-2 via allosteric activation of LXR.
Using yeast systems for the study of the LXR–co-activator interaction
To understand the molecular mechanism of LXR binding to the NR box-2 of ASC-2, we used a yeast-interaction system and attempted to identify the molecular determinants for the interaction. Yeasts lack endogenous NRs and NR-specific co-activators (e.g. SRC family proteins, ASC-2), but contain conserved general co-activator proteins that mediate the effects of NR-specific co-regulators [36–38]. Thus the advantage of studying NR–co-regulator interactions in yeast is that the system is free from any complication from NR co-regulators present in mammalian systems. In the present study, although the data from the yeast-interaction assays were generally consistent with the in vitro results, there were some discrepancies. LXR and Co4aN could interact in yeast even in the absence of exogenous LXR ligand, and there was no stimulatory effect of 22R-HC on the interaction. Previously, screening of a natural product library of microbial extracts identified steroidal and triterpenoidal fungal metabolites as LXR ligands . Thus although we did not examine other LXR ligands, it is possible that this lack of effect of 22R-HC is due to the presence of endogenous ligands in yeast. Another possibility is that the fusion of LXR to LexA may result in improper folding or alteration of the DNA binding orientation of LXR, and the conversion of the LXR LBD into a quasi-ligand-bound (activated) conformation. To avoid these potential artifacts, we carried out our analyses using the yeast HRE assay, using native forms of NRs and a DR4-LacZ reporter, and obtained similar results to those with the LexA reporter system.
Similar binding specificities of the two LXR isoforms for interaction with NR box-2 of ASC-2
LXRα and LXRβ have distinct expression patterns: LXRβ is ubiquitously expressed, whereas LXRα expression is relatively restricted to tissues involved in lipid metabolism [13,14]. Thus we considered the possibility that the two LXR isoforms have distinct binding preferences for co-activator proteins or that there are differences in the molecular determinants required for the specific interaction of each isoform with a given co-activator. However, the present study revealed no significant differences in the interaction specificities of ASC-2 toward both LXR isoforms. Interestingly, in contrast with other NR isoforms, sequence comparison of LXRα and LXRβ revealed that the highest level of sequence variability between the two isoforms is restricted primarily to the DBD region , which suggests an interesting possibility that any functional difference between LXRα and LXRβ may depend on isoform-specific LXREs. Sequence conservation is relatively high in the LBDs of the LXR isoforms, with 77% sequence identity in this region between the two human LXR isoforms. This conservation appears to define the ligand-binding characteristics of the LXR isoforms, as the two isoforms show similar binding affinities for a wide range of agonists . According to the crystal structures of LXR–RXR LBD heterodimers complexed with LXR synthetic ligand T0901317, common residues of the LBDs of different isoforms of LXR are involved in the interaction with T0901317 . In this respect, it is reasonable to suggest that similar agonist-bound conformations of the LBDs of LXRα and LXRβ determine the specificity of the Co4aN interactions with both LXR isoforms. As mentioned above, LXRs function as heterodimers with RXR and can be activated by either the ligand for LXR or the ligand for RXR (permissive heterodimer). In the crystal structure of LXRα–RXRβ LBDs, the binding interface between LXR and RXR involves the residues of helix 7 of RXRβ, but not those of helix 7 of LXRα . Conversely, the loop connecting H8 and H9 of LXR, but not the same loop in RXR, contributes to binding. This asymmetric interaction is a common feature of permissive heterodimers, such as LXR–RXR and PPAR–RXR . The asymmetry of the LXRα–RXRβ heterodimer is more prominent than that observed for other RXR heterodimers, as inferred from the dimerization interface of the LXRα–RXRβ which is approx. 20% smaller than those of other heterodimers . This asymmetric interaction may stabilize the AF2 helix of the interacting partner in a position that permits co-activator interactions even in the absence of the ligand of the partner, which might explain the molecular basis of ligand permissiveness and RXR-mediated allosteric activation [19,42]. In the case of LXR binding to NR box-2 of ASC-2, heterodimerization with RXR or liganded RXR may indirectly cause further structural change in LXR LBD, resulting in a more favourable conformation of LXR for NR box-2 binding. However, our results consistently indicated that the specific residues of Co4aN that are involved in its interaction with LXR are not changed by the presence of RXR (dimeric status) or liganded-RXR (ligand condition). This suggests that the mechanism of allosteric activation via the ligand-independent or -dependent RXR phantom effect does not appear to involve a significant change in the conformation of LXR-LBD, which could be the cause of the apparent altered specificity of the interaction between LXRs and NR box-2 of ASC-2. Detailed structural studies on the unliganded-LXR–RXR heterodimer will be helpful to prove our suggested mechanism.
Molecular determinants of the interaction between LXR and NR box-2 of ASC-2
The sequence of NR box is not merely a receptor-co-activator docking site, but also contains information that governs the specificity of the interaction of co-activators and NRs. The residues flanking the NR box sequence appear to have a great impact on the NR selectivity and binding affinity . We demonstrated that, in addition to LXXLL motif, the residues P−3, T−2 and Q+3 of NR box-2 are the critical determinants for LXR binding. The crystal structure of the LXRα–RXRβ heterodimer indicates that the liganded-LXRα LBD adopts the typical conformation of the transcriptionally activated state of other NRs bound to cognate ligands . In this structure, the GRIP-1 NR box peptide binds through its helical structure to the predominantly hydrophobic groove formed by H3, H4, H5 and H12 (AF2 helix) of the LXR LBD, which is reminiscent of the PPARγ–rosiglitazone complex bound by the SRC-1 NR box peptide. Considering this common feature of the liganded-LXR, the L+1, L+4 and L+5 residues of ASC-2 NR box-2 must participate in direct binding of ASC-2 to the hydrophobic pocket of the LXRα LBD, which would explain why mutations in these residues result in complete loss of interactions with LXRs. Previous structural and biochemical studies revealed that the preferred interaction of the NR box-2 of GRIP1/SRC-2 with TRα can be ascribed to favourable interactions between the basic residues positioned upstream of its core LXXLL motif and acidic residues in the AF2 helix of the TR . Conversely, mutational studies on another member of the SRC family (SRC-1) demonstrated that specific amino acids in the C-terminal region of the LXXLL motif are involved in differential interactions of this motif with target NRs . We could isolate Co4aN mutations located in the N-terminus, but not in the C-terminal, flanking region of the core LXXLL motif, which is reminiscent of the preferential interaction of the TR with the NR box-2 of GRIP1. The −3 position of the ASC-2 NR box-2 contains a proline residue that acts as a helix-breaker and it is located far from the core motif. For these reasons, we suppose that the P−3 residue may affect the orientation of NR box-2 binding to LXR rather than directly making contact with the LXR. If this hypothesis were correct, the P−3Q mutant would not bind to LXR, since it would have lost its ability to adopt the proper orientation upon docking to the LXR-LBD. As to the mutational effect of T−2P, due to the P−3 residue, another proline residue at the −2 position may render the NR box-2 motif too rigid to accommodate an induced fit for binding to the LBD pocket. To more precisely define the role of the P−2 residue, we generated a T−2A mutant in the context of Co4aN and ASC-2. Since alanine has a small side chain, the mutational effect of an alanine substitution would not be expected to change the global structure of NR box-2. We examined the T−2A effect on LXR binding in the yeast-two hybrid assay and its co-activator function in mammalian cells. The T−2A mutant displayed reduced interaction with the LXR and resulted in LXR transactivation that was less than 50% of the wild-type level, raising the possibility that the threonine residue at the −2 position may directly participate in the LXR interaction (results not shown). Interestingly, our data consistently showed that the Q+3 residue, but not S+2, plays a critical role in NR box-2 binding to LXRs. Structural studies indicate that the side chains of the residues at the +2 and +3 positions are solvent-exposed  and that these residues are not directly involved in the NR box–NR interaction since this interaction is unaffected by the mutations in these residues . Currently, it is unclear whether the Q+3 residue plays a role in direct binding to the LXR or indirectly contributes to the LXR-specificity of NR box-2. A more detailed structural analysis will be needed to answer this question.
Identification of the Ser1490 (S−1) residue as a key determinant for the LXR-specificity of NR box-2
Although the upstream sequence of the core NR box-2 is well conserved across mammalian ASC-2 homologues (dog, rat and mouse), these residues are not conserved in other NR box motifs and are not important residues for the classification of NR boxes. From these facts, we proposed that the unique sequence of the ASC-2 NR box-2 appears to provide ASC-2 the molecular specificity for LXR-binding. We suggested that the Ser1490 residue of ASC-2 (S−1 residue of NR box-2 motif) may significantly contribute to LXR-specific binding, based on observations that the S−1L mutant had substantial binding activity for the TR, RAR, as well as for the LXR. Interestingly, most of NR box motifs found in co-activator molecules contain a hydrophobic or a positively charged amino acid at the −1 position, which is thought to enhance NR box motif binding to various NRs . This may provide a molecular basis for why the S−1L mutant of the NR box-2 lost its LXR-specificity but could still interact with the LXR (in other words, it has the general NR binding activity of class III NR boxes). Although the S−1L mutant gained the ability to interact with the TR and RAR, it still could not bind to PPAR and VDR (results not shown). Although class III NR boxes cannot interact with VDR , this result raises the intriguing possibility that residue(s) other than S−1 in the ASC-2 NR box-2 might prevent it from binding to PPAR and VDR. A more detailed analysis will be needed to fully define the molecular determinants involved in the LXR-specific binding of NR box-2.
In addition to ASC-2, SRC co-activators have also been shown to have a role in transcriptional regulation by LXR–RXR heterodimers. In vitro, DR4-bound LXR–RXR heterodimers exhibited a three-fold activation pattern by RXR and their ligands, which correlated with the recruitment of SRC-1 . SRC-1, either alone or in combination with p300, can potentiate LXR-mediated transcription of the gene encoding ABCA1 . In the yeast-interaction assay, we found that LXR interacts with SRC-1 or TRAP220 fragment containing NR boxes, although it is not clear whether TRAP220 acts as a co-activator of LXR (Y. L. Son, M. J. Park and Y. C. Lee, unpublished work). Interestingly, all of these interactions were stimulated by the addition of RXR and 9cRA, as was the case for ASC-2. Although the NR boxes of ASC-2 are located 600 amino acids apart, the spacing of the NR boxes of SRC-1 and TRAP220 is approx. 50 and 40 amino acids respectively. The NR boxes of SRC-1 and TRAP220 are believed to interact co-operatively with both subunits of the NR homodimer or heterodimer. Actually, in the crystal structure of LXRα–RXRβ LBD heterodimer complexed to two GRIP-1 NR box peptides, each LBD of the NRs has a typical agonist conformation, in which the co-activator-binding sites of both NRs are bound by GRIP-1 peptides . This model indicates that RXR may actively participate in binding to these co-activators via a direct interaction with NR box sequences, rather than functioning as an allosteric activator. Currently, we are carrying out a direct comparative analysis of the contributions of LXR and RXR in binding to the NR boxes of SRC-1 and TRAP220 in order to define the molecular determinants of the three-fold mechanism of activation of LXR–RXR heterodimers by various co-activator proteins.
This work was supported by grants from Korea Research Foundation (2002-070-C00068 and 2006-005-J03003) to Y. C. L. and by National Institutes of Health Grant DK064678 to J. W. L. Y. L. S. and G. S. K. are supported in part by the Second Stage BK21 Programme.
activation function 2
activating signal co-integrator-2
hormone response element
liver X receptor
retinoic acid receptor
retinoid X receptor
steroid receptor co-activator
thyroid hormone receptor