Heterodimeric DNA binding by the vitamin D receptor and retinoid X receptors is enhanced by 1,25-dihydroxyvitamin D3 and inhibited by 9-cis-retinoic acid. Evidence for allosteric receptor interactions.

Gel mobility shift analysis was utilized to investigate the molecular function of 1alpha,25-dihydroxyvitamin D3 (1,25-(OH)2D3) and 9-cis-retinoic acid (9-cis-RA) ligands in the binding of the vitamin D receptor (VDR) and retinoid X receptor (RXR) to mouse osteopontin and rat osteocalcin vitamin D-response elements (VDREs). At physiological ionic strength and reduced concentrations of expressed proteins, efficient binding to either VDRE occurs as a VDR. RXR heterodimer, not as a VDR homodimer. 1,25-(OH)2D3 dramatically enhances heterodimer-VDRE interaction, whereas somewhat higher concentrations of 9-cis-RA inhibit this association, perhaps related to the role of this retinoid in facilitating RXR homodimer formation. Interestingly, if VDR is occupied by 1,25-(OH)2D3 prior to complexing with RXR, the resulting heterodimer is relatively resistant to dissociation and diversion to other pathways by 9-cis-RA. Therefore, a proposed molecular action of 1,25-(OH)2D3 is to generate an allosteric switch in VDR to a form that not only binds to the VDRE with high affinity and specificity as a heterodimer with RXR, but also interacts with the RXR partner to conformationally restrict the action of its cognate ligand.

Gel mobility shift analysis was utilized to investigate the molecular function of 1␣,25-dihydroxyvitamin D 3 (1,25-(OH) 2 D 3 ) and 9-cis-retinoic acid (9-cis-RA) ligands in the binding of the vitamin D receptor (VDR) and retinoid X receptor (RXR) to mouse osteopontin and rat osteocalcin vitamin D-response elements (VDREs). At physiological ionic strength and reduced concentrations of expressed proteins, efficient binding to either VDRE occurs as a VDR⅐RXR heterodimer, not as a VDR homodimer. 1,25-(OH) 2 D 3 dramatically enhances heterodimer-VDRE interaction, whereas somewhat higher concentrations of 9-cis-RA inhibit this association, perhaps related to the role of this retinoid in facilitating RXR homodimer formation. Interestingly, if VDR is occupied by 1,25-(OH) 2 D 3 prior to complexing with RXR, the resulting heterodimer is relatively resistant to dissociation and diversion to other pathways by 9-cis-RA. Therefore, a proposed molecular action of 1,25-(OH) 2 D 3 is to generate an allosteric switch in VDR to a form that not only binds to the VDRE with high affinity and specificity as a heterodimer with RXR, but also interacts with the RXR partner to conformationally restrict the action of its cognate ligand.
The active metabolite of vitamin D, 1␣,25-dihydroxyvitamin D 3 (1,25-(OH) 2 D 3 ), 1 exerts its hormonal effects predominantly on intestine, bone, and kidney, where it plays a crucial role in calcium and phosphate homeostasis, bone mineralization, and the prevention of rickets (1). 1,25-(OH) 2 D 3 also has effects on many non-classical target tissues (2), such as those of the bone marrow/immune system, where it has been reported to be involved in a wide variety of differentiation processes (3,4). The cellular responses to 1,25-(OH) 2 D 3 are mediated by the vitamin D receptor (VDR) (1), which is a member of a superfamily of nuclear transcription factors that include the steroid, retinoid, and thyroid hormone receptors (5). Transcriptional control of specific 1,25-(OH) 2 D 3 -responsive genes is achieved by binding of the receptor, via its DNA binding domain, to a specific vitamin D-response element (VDRE) located in the 5Ј promoter region of the regulated gene. VDREs have been identified in positively controlled genes such as human and rat osteocalcin (6 -8), mouse osteopontin (9), avian integrin ␤ 3 (10), and rat vitamin D 24-hydroxylase (11,12), and comparison of these sequences indicates that they generally consist of a hexanucleotide direct repeat with a spacer of three nucleotides (DRϩ3).
In many studies with natural DRϩ3 VDREs, VDR has been reported to bind as a heterodimer with the retinoid X receptor (RXR) (reviewed in Ref. 13). Thyroid hormone receptor (TR) and retinoic acid receptor (RAR) also form RXR-containing heterodimers on positively controlled DRϩ4 and DRϩ5 response elements, respectively (14). Evidence has been provided, however, implying that VDR may also bind as a homodimer to certain DRϩ3 elements, primarily those containing the osteopontin-like half-site, AGTTCA (15,16).
In addition to this issue of VDR homodimeric binding to VDREs, controversy exists as to the precise molecular role (if any) of 1,25-(OH) 2 D 3 in VDR⅐RXR heterodimer (or homodimer) formation and DNA binding. 1,25-(OH) 2 D 3 has been proposed to elicit high affinity association of the VDR with its heterodimeric partner (17,18), as well as induce a conformational change in VDR⅐RXR inferred from an increase in the mobility of the VDR⅐RXR⅐VDRE complex (8). Other studies indicated that the presence of both RXR and 1,25-(OH) 2 D 3 is required for optimal binding of VDR to a DRϩ3 from the osteocalcin VDRE (19) and to a 24-OHase VDRE (20). In contrast, it has been proposed that initial VDR⅐RXR DNA interaction occurs in a ligand-independent fashion (21). Still another variation was postulated by Freedman and co-workers (22), in which the role of 1,25-(OH) 2 D 3 was hypothesized to be the dissociation of the DNA-bound homodimer by decreasing the rate of conversion of DNA-bound monomer to homodimer, and ultimately promoting the formation of a DNA-bound VDR⅐RXR heterodimer.
There are also conflicting experimental data on the role played by 9-cis-retinoic acid (9-cis-RA), the cognate ligand for RXR, in the binding and transcriptional regulation of 1,25-(OH) 2 D 3 -responsive genes by the VDR⅐RXR heterodimer. 9-cis-RA has been observed to repress the 1,25-(OH) 2 D 3 -dependent accumulation of osteocalcin mRNA in osteoblast-like ROS 17/ 2.8 cells (19). Further analysis of VDRE binding in mobility shift assays (19) demonstrated that heterodimeric interactions between VDR and RXR were enhanced by 1,25-(OH) 2 D 3 , and were inhibited by high concentrations of 9-cis-RA, consistent with the observation that RXR-specific ligands can induce RXR homodimer formation in solution (23). In contrast, treatment of a variety of transfected cell types with both 1,25-(OH) 2 D 3 and 9-cis-RA was reported to enhance the expression of an osteopontin VDRE-linked reporter gene over levels observed with 1,25-(OH) 2 D 3 alone (24,25). An apparent synergistic co-stimulation by 1,25-(OH) 2 D 3 and 9-cis-RA in transfection systems employing AGGTCA synthetic DRϩ3 constructs has also been reported (26,27), and in at least one instance, 9-cis-RA was observed to have no effect on 1,25-(OH) 2 D 3 -stimulated transcription (28). Because of the wide range of cell types (insect to mammalian) utilized in these transfection studies plus the untold effects of variable receptor/coreceptor expression, as well as different levels of reporter plasmid, it is difficult to interpret the general significance of these findings. Therefore, to define at the biochemical level how 1,25-(OH) 2 D 3 , VDR, RXR and DNA interact ultimately to stimulate the transcription of hormone-responsive genes, the following three areas of conflicting experimental data were addressed: (i) the relative importance of VDR homodimer versus VDR⅐RXR heterodimer binding to DRϩ3 VDREs, including elements consisting of osteopontin half-sites; (ii) the effect of 1,25-(OH) 2 D 3 on VDR binding to RXR or VDR⅐RXR binding to DNA; and (iii) the role, if any, of 9-cis-RA on VDR⅐RXR DNA binding. To resolve these issues, we have utilized an electrophoretic mobility shift assay system that includes physiological salt and limited receptor concentrations to investigate the effects of cognate ligands on possible VDR homodimer and VDR⅐RXR heterodimer binding to their target response elements.
Reagents-Klenow enzyme was purchased from Promega Corp. (Madison, WI). Poly(dI-dC) was obtained from Boehringer Mannheim. Synthetic oligonucleotides corresponding to the respective hormone response elements were synthesized by Integrated DNA Technologies Inc. (Coralville, IA). [␣-32 P]dCTP (3000 Ci/mmol) was purchased from NEN Life Science Products.
Oligonucleotide Labeling-Double-stranded versions of each of the tested response elements (direct repeat hexanucleotides are underlined) with HindIII overhangs (lowercase) were created by annealing the following single-stranded synthetic oligonucleotides together with their complements (not shown). For the mouse osteopontin VDRE (9), the oligonucleotide was 5Ј-agctACAAGGTTCACGAGGTTCACGTTCT-3Ј. For the rat osteocalcin VDRE (7), the oligonucleotide was 5Ј-agctG-CACTGGGTGAATGAGGACATTACA-3Ј. For the rat cellular retinolbinding protein (CRBP) II RXRE (34), with a potential third half element indicated by a double underline, the oligonucleotide was 5Ј-agctGCTGTCACAGGTCACAGGTCACAGTTCA-3Ј.
The double-stranded oligomers were labeled with [␣-32 P]dCTP to a specific activity of Ն10 8 cpm/g as described by Nakajima et al. (35).
Ligand-dependent Gel Mobility Shift Assays-Previous VDR⅐RXR gel shift protocols were altered to reduce concentrations of receptors to approximate physiologic levels. Accordingly, hVDR (10 -100 ng ϭ 5-50 nM) was added to a mixture containing 10 mM Tris-HCl, pH 7.6, 150 mM KCl, 1.0 mM dithiothreitol, 15% glycerol, 1 mg/ml acetylated bovine serum albumin, and 50 g/ml poly(dI⅐dC), followed immediately by an amount of RXR␣ or RXR␤ equivalent to that of hVDR. Prior to the addition of receptor(s), 1,25-(OH) 2 D 3 was added to the reaction mixture in ethanol vehicle to give a final ligand concentration of 10 Ϫ10 to 10 Ϫ5 M. In experiments that involved the additional inclusion of 9-cis-RA (10 Ϫ10 to 10 Ϫ5 M), ethanol vehicle controls were included and the binding portion of the reaction (45 min at 22°C) was shielded from light. One l of radioactively labeled DNA probe (containing 0.5 ng DNA, or approximately 0.8 nM, and 50,000 -100,000 cpm) was added to give a final volume of 40 l, then incubated for an additional 30 min. Thereafter, the entire reaction was subjected to nondenaturing electrophoresis on a 4% polyacrylamide gel. The dried gels were used to expose Kodak X-Omat film at Ϫ70°C. The results of a number of assays were quantitated by densitometric analysis using an Apple OneScanner with Ofoto™ scanning software developed by Light Source, Inc.
A protocol identical to that described immediately above was utilized for the homodimeric association of RXR␤ with the CRBP II RXRE, except that 9-cis-RA was substituted for 1,25-(OH) 2 D 3 and no VDR was included. Finally, a separate set of experiments was carried out in which the order of addition of each of the receptors and their respective ligands to the reaction mixture was varied, with the total time of each reaction, including preincubation of receptor(s) with ligand, heterodimer formation, and DNA binding, not exceeding 110 min.
Preparation of Nuclear Extract from ROS 17/2.8 Cells-Cells were cultured in Dulbecco's modified Eagle's medium/Ham's F-12 medium (1:1) supplemented with 5% fetal bovine serum and 5% newborn calf serum. Log phase cultures were harvested by trypsinization and collected by centrifugation at room temperature for 10 min at 1000 ϫ g, followed by washing in five volumes of phosphate-buffered saline at 4°C. The nuclear extract was then prepared as described previously (36), with all steps performed at 4°C. Fig. 1A depicts a gel mobility shift experiment using the mouse osteopontin (MOP) VDRE and baculovirus-expressed VDR in the presence and absence of hRXR␣ or mRXR␤ (20 ng of each receptor). At limiting concentrations, neither the VDR alone (lanes 1 and 2) nor RXR alone (data not shown) (35) produce an observable DNA binding complex either in the absence or presence of 1,25-(OH) 2 D 3 . The only major DNA binding complex observed in this experiment is that produced by incubation of VDR plus either RXR␣ or RXR␤, and, by densitometric analysis, this complex is enhanced 23-fold in the presence of hormone (compare lane 3 to lane 4, and lane 7 to lane 8). The average degree of binding enhancement elicited by 1,25-(OH) 2 D 3 observed from more than 30 such experiments carried out at limiting receptor concentrations was 20-fold, although effects of ligand apparently as low as 2.5-fold have been recorded (see Fig. 4). The hormone effect is selective for 1,25-(OH) 2 D 3 , since the relatively inactive natural vitamin D metabolite 24(R)-25-(OH) 2 D 3 , its epimer, 24(S)-25-(OH) 2 D 3 , as well as the 25-(OH)D 3 , 1␣-(OH)D 3 , and vitamin D 3 hormone precursors, do not enhance DNA-binding complexes at 10 Ϫ8 M (data not shown). The hormone-stimulated, heterodimeric bands seen in lanes 4 and 8, are eliminated (lanes 5 and 9) by inclusion of the VDR-specific monoclonal antibody 9A7, which recognizes an epitope within the DNA binding domain. Moreover, the addition of an RXR␣-specific polyclonal antibody (lane 6) caused a majority of the complex to supershift, confirming the presence of RXR␣ in the complex. Interestingly, the RXR␤specific monoclonal antibody MOK 13.17 (lane 10) showed little interaction with the VDR⅐RXR␤ DNA binding complex. As discussed below, the MOK 13.17 antibody may not be able to access its epitope when RXR␤ is heterodimerized with VDR. display DNA complexes formed by VDR in combination with RXR␣ with or without hormone, whereas lanes 5 and 6 show the effects of a VDR-specific monoclonal antibody, 9A7 (0.6 g; added to the reaction after VDR and RXR are allowed to heterodimerize in the presence of 1,25-(OH) 2 D 3 ), and an RXR␣-specific polyclonal antibody (1.0 g; preincubated with RXR␣ alone), respectively. Lanes 7-10 repeat the conditions of lanes 3-6, but using mRXR␤ as the VDR partner and 1 g of an RXR␤-specific monoclonal antibody, MOK 13.17 (1 g; added to the reaction after VDR and RXR are allowed to heterodimerize in the presence of 1,25-(OH) 2 2-7). The gel mobility shift assay was carried out using 20 ng each of baculovirus expressed hVDR and hRXR␤, employing a salt concentration of 150 mM KCl as described under "Experimental Procedures." VDRE ( Fig. 2B) displays a slightly different pattern with an optimum at 10 Ϫ6 M hormone, followed by a more gradual decrease in binding as the ligand concentration is lowered. Den-sitometric scanning of these two gels, however, revealed that the ED 50 of 1,25-(OH) 2 D 3 for both VDREs was approximately 10 Ϫ9 M, representing a 13-fold increase in receptor heterodimer binding to the ROC VDRE and a 15-fold increase on the MOP element. Accordingly, the concentrations of 1,25-(OH) 2 D 3 chosen for most subsequent experiments were either 10 Ϫ9 or 10 Ϫ8 M.

VDR Ligand Enhances Binding of RXR⅐VDR Heterodimers to the Mouse Osteopontin and Rat Osteocalcin VDREs-
Effect of 1,25-(OH) 2 D 3 to Reduce Dissociation of the VDR⅐RXR Heterodimer from a VDRE-Labeled MOP VDRE was incubated with VDR and RXR␤ in the absence and presence of 1,25-(OH) 2 D 3 , followed by inclusion of a 100-fold excess of competing unlabeled MOP VDRE at time zero. As shown in Fig. 3A, the presence of ligand not only greatly potentiates initial (time 0) DNA binding, as seen previously, but also results in an apparent slower rate of dissociation of the heterodimeric receptor complex from the labeled response element. Densitometric analysis of this experiment (Fig. 3B) indicates that the liganded VDR⅐RXR complex exhibits a halflife of Ն11 min (by extrapolation), whereas the unliganded VDR⅐RXR heterodimer dissociates with an estimated half-life of only 3.0 min.
Higher Receptor Levels Generate Apparently Artifactual VDR Homodimers and Attenuate the VDR Ligand Dependence of VDR⅐RXR Binding to the ROC and MOP VDREs-The experiment depicted in Fig. 4 was performed using both 20 ng and 100 ng of VDR, to determine if homodimeric binding, which was not observable using limiting receptor concentrations (Fig. 1A), could be elicited at receptor concentrations significantly in excess of physiologic levels. On the weaker ROC response element (Fig. 4, left panel), neither 20 ng (lanes 1 and 2) nor 100 ng (lanes 3 and 4) of VDR exhibited any observable monomeric or homodimeric DNA binding species, even in the presence of 1,25-(OH) 2 D 3 . We have previously observed VDR homodimer binding to the ROC VDRE at very high VDR amounts of 260 ng (35), but the stability of such VDRE complexes is quite minimal, with an apparent half-life of less than 0.5 min when assayed in a mobility shift dissociation assay similar to that shown in Fig. 3. 2 On the stronger MOP response element (Fig.  4, right panel), 100 ng of VDR produced a faint DNA binding complex, apparently corresponding to a homodimeric species (lanes 3 and 4). Moreover, by densitometric analysis, the puta- After complex formation, but before undergoing non-denaturing gel electrophoresis, a 100-fold excess (50 ng) of unlabeled competing MOP probe was added and individual reactions were allowed to proceed for the indicated time intervals, followed by electrophoresis on 4% polyacrylamide gels. B, densitometric scanning results of top bands in A are shown, expressed as percent of maximum VDR⅐RXR binding. EtOH, ethanol vehicle.
FIG. 4. Limiting concentrations of VDR bind to VDREs as a heterodimer rather than as a homodimer. Standard gel mobility shift assays were carried out on two different VDREs (0.5 ng) using the indicated amounts of receptors. Each receptor combination was tested in the presence (ϩ) or absence (Ϫ) of 1,25-(OH) 2 D 3 . Note that, as observed in Fig. 1B and 3A, and discussed under "Results," a second, faster migrating DNA complex appears with mRXR␤, which likely represents a species composed of VDR and truncated mRXR␤. ments, 20 ng of VDR in combination with 20 ng of RXR␤ produced DNA binding heterodimeric species on the ROC and MOP VDREs which were enhanced by ligand 10-fold and 2.5fold, respectively (Fig. 4, lane 6 versus lane 5 in each panel). At higher, nonphysiological amounts (100 ng of each receptor), however, the binding of this heterodimeric species was not significantly enhanced (1.3-fold for both ROC and MOP) by the presence of 1,25-(OH) 2 D 3 (Fig. 4, lane 8

versus lane 7 in each panel).
Effect of 9-cis-RA on RXR Binding to an RXRE as well as on RXR⅐VDR Association with a VDRE-9-cis-RA reportedly can augment the DRϩ1 binding efficiency of RXR␣ produced by in vitro transcription/translation, although no RXR ligand effect was observed with receptor produced in E. coli, Sf9, or COS-1 cells (37). Accordingly, we probed the possible influence of 9-cis-RA on the binding of RXR homodimers to a natural RXRE, a DRϩ1 element from the rat CRBP II gene (34). Using our conditions of reduced amounts of receptor and physiological salt concentrations, binding of BEVS mRXR␤ (40 ng) as a homodimer to the CRBP II RXRE was enhanced 16-fold by the presence of 10 Ϫ6 M 9-cis-RA (Fig. 5A, lane 2 versus lane 1).
Although it was noted in Fig. 1A that the MOK 13.17 RXR␤ antibody had relatively little effect on VDR⅐RXR␤ binding to a VDRE, an apparent stabilization as well as robust supershift of the DNA-bound, unliganded, and ligand-enhanced RXR homodimeric species by this antibody was observed (Fig. 5A,  lanes 3 and 4). Thus, RXR␤ appears to be more accessible to the MOK 13.17 antibody when configured in its homodimeric form than when it exists in a heterodimer with VDR. As expected, the ligand-dependent RXR homodimer complex with the RXRE was neither inhibited nor supershifted in the presence of a VDR antibody (Fig. 5A, lanes 5 and 6). Having demonstrated that, under our conditions, 9-cis-RA can facilitate RXR homodimerization, we next addressed the question of its influence on RXR⅐VDR heterodimers. As expected, 10 Ϫ9 M 1,25-(OH) 2 D 3 effectively promotes VDR⅐RXR binding to the MOP response element (Fig. 6A, lane 2 versus lane 1). Although 9-cis-RA alone does not have a significant effect on VDR⅐RXR binding to DNA (data not shown), this retinoid, at concentrations of 10 Ϫ5 and 10 Ϫ6 M, suppresses 1,25-(OH) 2 D 3 -stimulated DNA binding by the VDR⅐RXR heterodimer by 97% and 53%, respectively. Concentrations of 10 Ϫ7 and 10 Ϫ8 M each attenuate heterodimeric binding to the VDRE by 40%, whereas concentrations of 10 Ϫ9 M or below have no observable suppressive effect (Fig. 6A). In Fig. 6B, a ROS 17/2.8 nuclear extract (3.0 g of protein) was used to show that 9-cis-RA also diminishes the effect of 1,25-(OH) 2 D 3 on endogenous VDR⅐RXR complexing to a VDRE. Lanes 1 and 2 demonstrate the 2-3 fold enhancement of VDRE binding by the endogenous receptors elicited by the addition of hormone (10 Ϫ8 M), whereas lane 3 illustrates an 80% reduction in the 1,25-(OH) 2 D 3 -dependent enhancement by 10 Ϫ6 M 9-cis-RA. A similar reduction to that observed with hRXR␣ and endogenous rat RXR is seen when baculovirus-expressed mRXR␤ is used as the heterodimeric partner (data not shown). The inhibitory effect of 9-cis-RA on 1,25-(OH) 2 D 3 -induced binding can be overcome, however, by increasing the amount of RXR (Fig. 6C). The results in Fig. 6 (plus five other replicate experiments similar to that shown in Fig. 6A, along with the data in Fig. 5), are consistent with an ability of high concentrations of 9-cis-RA (10 Ϫ5 M, 10 Ϫ6 M and, to a lesser extent, 10 Ϫ7 M and 10 Ϫ8 M) to inhibit 1,25-(OH) 2 D 3 -dependent enhancement of VDRE binding by both overexpressed and endogenous VDR, most likely by occupying a limiting concentration of RXR and conformationally diverting it from heterodimerization with VDR toward homodimerization.
Effect of Different Orders of Receptor-Ligand Assembly-In a typical experiment depicted in Fig. 7, we employed baculovirus expressed mRXR␤ as the heterodimeric partner for VDR. In lanes 1-3, the individual receptors and the indicated ligands were added simultaneously to the reaction mixture. Lanes 4 -6 contain reactions in which VDR and RXR were first allowed to associate for 30 min and were then added to a reaction mixture containing either 1,25-(OH) 2 D 3 (lane 5) or a combination of 1,25-(OH) 2 D 3 and 9-cis-RA (lane 6). Finally, each receptor was individually preincubated with the indicated ligands and, after a 30-min period, the respective receptor partner was added to the reaction (lanes 7-10). Densitometric analysis of four such independently performed experiments indicated 9-cis-RA inhibits 1,25-(OH) 2 D 3 -mediated VDRE binding by an average of 78% when each component is added simultaneously to the reaction (Fig. 7, lane 3 versus lane 2). Similar repression of heterodimeric binding (85%) is observed when the preincubated VDR⅐RXR mixture is combined with ligands (Fig. 7, lane  6 versus lane 5), and a significant inhibitory effect (60%) also is observed when RXR is incubated initially with 9-cis-RA to apparently form homodimers (Fig. 7, lane 10 versus lane 8). However, the 9-cis-RA inhibitory effect is nearly prevented (an average of only 14% reduction) whenever VDR is allowed to preincubate in a mixture containing 1,25-(OH) 2 D 3 and 9-cis-RA, with RXR being added to the system after 30 min (Fig. 7,  lane 9 versus lane 8). Further experiments (data not shown) indicated that if a preformed heterodimer is incubated with either 1,25-(OH) 2 D 3 alone for 30 min or with 1,25-(OH) 2 D 3 followed by the addition of 9-cis-RA, or even first with 9-cis-RA alone for 30 min followed by exposure to 1,25-(OH) 2 D 3 , this heterodimer is still strikingly susceptible to inhibition by 9-cis-RA (approximately 88% to 95% reduction). Thus, the situation in which VDR is allowed to bind 1,25-(OH) 2 D 3 before heterodimer formation is unique in that the resulting heterodimer complex is largely resistant to dissociation by 9-cis-RA. DISCUSSION An experimental system has been devised to investigate the interaction of VDR, RXR, 1,25-(OH) 2 D 3 , and 9-cis-RA with DNA, and to assess the contribution of each component to overall complex formation. Previous studies investigating the role of ligand in the interaction of VDR with DNA have produced somewhat conflicting results and conclusions (see Introduction). Indeed, even in our hands, VDR complex formation on a VDRE can become ligand-and RXR-independent either when salt concentrations are non-stringent (Ͻ100 mM KCl; data not shown) or when receptor levels are increased to Ն50 nM (Fig. 4). The parameters used in the present gel mobility shift assay system were carefully chosen to approach those found within the cell, including 150 mM KCl and receptor concentrations in the range of 5-10 nM. Under these conditions, association between VDR and the MOP or ROC VDREs occurs only in the presence of RXR, and is greatly enhanced by 1,25-(OH) 2 D 3 (Fig.  1, A and B). Further, the fact that both types of VDRE display an average of Ͼ10-fold ligand-dependent enhancement of receptor-DNA binding under near-physiologic conditions suggests that this effect is not gene-or element-specific. Endogenous VDR and RXR from a ROS 17/2.8 cell nuclear extract (Figs. 1C and 6B) also exhibit a ligand-dependent enhancement of binding to the VDRE, although the magnitude of the effect is lower (2-3-fold). This lower degree of ligand stimulation possibly reflects the presence of proteins that nonspecifically bind to 1,25-(OH) 2 D 3 or interfering DNA-binding proteins in the cruder nuclear preparations. It is also conceivable that ligand  7. Preincubation of VDR with its ligand blunts inhibition by 9-cis-RA. Ligand-dependent gel mobility shift assays were carried out using different orders of receptor and ligand addition to the incubation mixture. VDR and RXR␤ (20 ng each) were used in reactions together with 0.5 ng of labeled MOP under standard binding conditions (see "Experimental Procedures"). Lanes 1-3 display reactions in which the receptors and the indicated ligands are added simultaneously to the incubation mixture. In lanes 4 -6, a preincubated VDR⅐RXR mixture was added to a reaction that contained the indicated ligands. In lanes 7 and 8, VDR was allowed to preincubate for 30 min in the absence (lane 7) or presence of 10 Ϫ9 M 1,25-(OH) 2 D 3 (lane 8) and then RXR was added to the reaction mixture. In lane 9, a indicates that VDR was preincubated for 30 min in a reaction mixture containing both 1,25-(OH) 2 D 3 and 9-cis-RA and then RXR was added to the reaction. In lane 10, b indicates that RXR was allowed to incubate in a mixture containing 1,25-(OH) 2 D 3 and 9-cis-RA and, after 30 min, VDR was added to the reaction. acts on overexpressed receptors to facilitate their folding into a native conformation, thereby enhancing hormone-dependent DNA binding. Nevertheless, the fact that we observe a consistent effect of ligand using endogenous VDR in ROS 17/2.8 cell nuclear extracts argues that the enhancement by ligand of receptor binding to VDREs is biologically relevant. This conclusion is independently supported by in vivo footprinting experiments (38), which reveal that 1,25-(OH) 2 D 3 is required for VDR occupation of the VDRE in the native rat osteocalcin promoter.
The present results indicate that VDR homodimerization is not a physiologically relevant event on either the ROC or MOP response elements. Previous studies utilizing an artificial osteopontin-like VDRE (15,22), and the actual mouse osteopontin VDRE (16), suggested that these elements represented a target for VDR homodimeric binding, although the salt concentrations used during gel mobility shift assays were not specified in every case, and very high (micromolar) concentrations of VDR were utilized in gel filtration experiments (22). We have used both the natural MOP and ROC VDREs (Fig. 4), which typify the two main subtypes of VDREs reported to date. ROC is an imperfect direct repeat with a G present as the third base in each half element, whereas MOP is a perfect direct repeat with a T present in this position. It is this latter feature of the MOP element that was proposed to distinguish it for VDR homodimeric binding (15). The current results (Fig. 4) employing limited amounts of receptor (20 ng or 10 nM) fail to show binding of any homodimeric form of VDR to either VDRE, arguing against any distinction between these positive VDREs in their binding of VDR homo-versus heterodimers. Only at supraphysiological levels of receptor (approximately 50 nM) can a faint homodimeric species be observed to form on the MOP VDRE; further, this weak binding is rela-tively unaffected by the presence of 1,25-(OH) 2 D 3 (see Fig. 4, lanes 3 and 4 in the right panel). Finally, MacDonald et al. (39) have demonstrated that VDR does not interact with itself in the yeast two-hybrid system, arguing against homodimerization of VDR in solution as well.
Although the heterocomplex of liganded VDR with RXR appears to constitute the physiologically significant DNA binding species, readily detectable VDRE association does occur in the absence of 1,25-(OH) 2 D 3 when both receptors are present in excess, in vitro (Fig. 4, lanes 7 and 8 in each panel). Loss of the 1,25-(OH) 2 D 3 requirement at higher receptor levels probably explains previous reports that VDR⅐RXR, and perhaps other heterodimeric DNA complexes such as those with TR⅐RXR and RAR⅐RXR, do not require ligand for specific DNA binding. We propose a model (see Fig. 8) in which VDR⅐RXR heterodimers exist in equilibrium between two distinct states, with the addition of 1,25-(OH) 2 D 3 shifting the equilibrium of the heterodimer to a form that binds to the VDRE with not only specificity, but also greater affinity (Fig. 3). A supraphysiologic concentration of receptor would, by mass action, artifactually increase the amount of receptor species already in the allosteric state that binds DNA, thus explaining the lack of hormone effect under conditions of vast VDR and RXR excesses.
The ligand-dependent gel mobility assay system has also been utilized to examine the ligand responsiveness of RXR, the heterodimeric partner of VDR, in binding as a homodimer to its cognate response element. Employing an RXRE of the DRϩ1 type (34), we find that RXR␣ and RXR␤ 3 homodimers form a It is proposed that both VDR and RXR, upon binding their respective cognate ligands, undergo distinct conformational changes, creating two alternative allosteric pathways for VDR⅐RXR association and response to ligand. A, an unliganded VDR binds its RXR partner to generate an apoheterodimer, which, upon subsequent binding of 1,25-(OH) 2 D 3 , undergoes a conformational change, presumably in its ligand binding domain, that results in enhanced association of the VDR⅐RXR-1,25-(OH) 2 D 3 complex with its specific DNA target sequence. In this form of the heterodimer, RXR is proposed to exist in a conformational state in which it can be readily dissociated from the VDR⅐RXR heterodimer by the addition of 9-cis-RA, leading to the formation of RXR homodimers that mediate distinct retinoid-responsive pathways. This 9-cis-RA-receptive conformation is proposed to exist also in monomeric RXR or in the apoRXR⅐VDR heterocomplex. B, when VDR binds 1,25-(OH) 2 D 3 before heterodimer formation, it is postulated to acquire a conformation distinct from that in pathway A. This 1,25-(OH) 2 D 3 -generated conformational change in monomeric VDR is then proposed to influence the state of the RXR ligand binding domain after heterodimerization. Essentially, occupied VDR is postulated to conformationally switch the RXR partner, rendering it hyposensitive to the effects of its own ligand and thus making this form of the heterodimer resistant to 9-cis-RA-elicited dissociation.
DNA-bound complex preferentially in the presence of 9-cis-RA (Fig. 5, A and B). In contrast to 1,25-(OH) 2 D 3 , the concentrations of 9-cis-RA (10 Ϫ5 to 10 Ϫ6 M) required to produce a striking ligand-dependent effect are higher than what are normally considered physiological concentrations (low nanomolar) of this retinoid. However, it should be noted that the K d (1.2 ϫ 10 Ϫ8 M) for 9-cis-RA binding to RXR (40) is 2 orders of magnitude higher than that for 1,25-(OH) 2 D 3 binding to VDR. It is also possible that the extreme photolability of 9-cis-RA may lower the effective concentration present in these in vitro reactions, and in vivo, target tissues may require local concentrations that are much higher than circulating levels for a hormonal response. Nevertheless, our results agree qualitatively with a previous report that 9-cis-RA enhances the binding of RXR␣ homodimers to a DRϩ1 element (41).
A persistent area of controversy has been the effect of 9-cis-RA on VDR⅐RXR heterodimer binding to DNA and, hence, regulation of gene transcription. Previous studies from a number of laboratories have produced conflicting results ranging from evidence that 9-cis-RA can destabilize VDR⅐RXR heterodimers and presumably induce RXR homodimer complex formation (19), to a conclusion that this ligand interacts synergistically with 1,25-(OH) 2 D 3 to transcriptionally activate the VDR⅐RXR heterodimer (24,25). Our data from the gel mobility shift assay system tend to confirm that the effect of 9-cis-RA is to dissociate the VDR⅐RXR heterodimer and divert RXR to form homodimers. In combination with 10 Ϫ9 M 1,25-(OH) 2 D 3 , 9-cis-RA at concentrations of 10 Ϫ5 M to 10 Ϫ8 M, equal to or slightly lower than those that elicit homodimerization of RXR (Fig. 5), cause significant dissociation of the VDR⅐RXR␣ heterodimer (Fig. 6A). The inhibitory effect of 9-cis-RA is seen even in the absence of an RXRE (data not shown), consistent with previous reports that RXR homodimerization takes place in solution (23). Importantly, using endogenous receptors present in a ROS 17/2.8 cell nuclear extract, 10 Ϫ6 M 9-cis-RA inhibits 1,25-(OH) 2 D 3 stimulated binding to the VDRE by 80% (Fig. 6B). The fact that this inhibitory effect of 9-cis-RA can be overcome by the subsequent addition of excess RXR (Fig. 6C) argues against the possibility that 9-cis-RA may be altering the affinity of the heterodimer for the VDRE and suggests that this inhibition is due to ligand-induced diversion of limiting RXR from the VDR⅐RXR complex toward homodimers. These findings can be generalized to propose the existence of hormonal crosstalk between metabolites of vitamins A and D in controlling VDRmediated transcriptional regulation.
The present results further suggest that, in addition to the availability of target sequences, receptors, and hormonal ligands, another parameter regulating the transcriptional response to hormone may be the sequential order in which these components assemble. As can be seen from Fig. 7, the ability of 9-cis-RA to shift the equilibrium from DNA-bound VDR⅐RXR heterodimer to RXR homodimer formation depends either on 9-cis-RA first occupying RXR monomers to facilitate their homodimerization, or on VDR and RXR forming an apoheterodimer. Evidence that VDR and RXR can form apoheterodimers in solution is supported by yeast two-hybrid interaction studies (39), and protein-protein association assays using either VDR or RXR expressed as glutathione S-transferase fusion proteins (data not shown). Taken together, these findings suggest the existence of at least two allosteric pathways for VDR⅐RXR association and response to ligand, as presented in Fig. 8. This model predicts two allosterically distinct forms of the 1,25-(OH) 2 D 3 -occupied heterodimer: (i) a preformed VDR⅐RXR heterodimer that subsequently binds 1,25-(OH) 2 D 3 and hence is induced to bind its specific DNA target sequence, but can be dissociated by the addition of 9-cis-RA ( Fig. 8, pathway A) and (ii) a pre-liganded form of VDR that creates a heterodimer, which is largely resistant to 9-cis-RAinduced dissociation (Fig. 8, pathway B). This model may explain why 9-cis-RA is observed only to partially suppress 1,25-(OH) 2 D 3 induced transcription in transfection studies (19), since the intact target cell likely contains a mixed population of the proposed two allosteric forms of the heterodimer.
Recently, a type of converse regulation has been reported with RXR homo-and heterodimers (42). It was found that unliganded RAR and TR suppressed RXR-mediated transcriptional response to an RXR specific-ligand, indicating that RAR and TR function as attenuating partners of RXR. Further, TR, when bound to its cognate ligand, completely suppresses RXR ligand binding, similar to what is observed with pre-liganded VDR (Fig. 8, pathway B). However, in contrast to both VDR⅐RXR and TR⅐RXR heterodimers, 9-cis-RA binding to RXR is restored when its partner RAR binds its cognate ligand, resulting in a synergistic response to both ligands. It was therefore proposed (42) that the addition of the RAR ligand transmits a positive allosteric signal to the RXR heterodimeric partner, whereas similar addition of ligand to TR prior to heterodimerization transmits a negative allosteric signal to RXR, which prevents it from binding 9-cis-RA. As summarized in Fig. 8, the current experimental evidence for VDR suggests that only the pre-liganded form of VDR can effectively mediate a negative allosteric change in RXR, such that it can neither bind 9-cis-RA nor readily dissociate to form RXR homodimers.
In conclusion, we propose that a role of 1,25-(OH) 2 D 3 is to drive VDR⅐RXR heterodimers to an allosteric form which displays high affinity and specificity for its target response element. We also have presented evidence that the sequential order of assembly of ligand, receptor, and DNA apparently can result in distinct allosteric forms of the heterodimer which respond differently to the presence of 9-cis-RA, and perhaps reflect allosteric changes in VDR that are conveyed to its heterodimeric partner. This proposed conformational switch may be fundamental to transcriptional regulation by the VDR subfamily of nuclear receptors.