Retinoid X Receptor (RXR) Ligands Activate the Human 25-Hydroxyvitamin D3-24-hydroxylase Promoter via RXR Heterodimer Binding to Two Vitamin D-responsive Elements and Elicit Additive Effects with 1,25-Dihydroxyvitamin D3 *

We have previously shown that RNA levels of kidney 25-hydroxyvitamin D3-24-hydroxylase (24(OH)ase), a key metabolic enzyme for 1,25-dihydroxyvitamin D3 (1,25(OH)2D3), is up-regulated by retinoids in mice within hours. Deletion analysis of ∼5500 base pairs of the human 24(OH)ase promoter showed that the sequence between −316 and −142 contained the information necessary and sufficient for retinoid-induced activation of the promoter. This region contains two previously defined vitamin D-responsive elements (VDREs) at −294 to −274 and −174 to −151. Mutation of either VDRE diminished responsiveness of the −316 to −22 promoter sequence to retinoids or 1,25(OH)2D3, while mutation of both VDREs essentially abolished the activity of the ligands via the promoter. Heterologous promoter vectors driven by the VDREs were responsive to a retinoid X receptor (RXR)-selective ligand (LG100268), a retinoic acid receptor (RAR)-selective ligand (TTNPB), or 1,25(OH)2D3, while combinations of LG100268 with either TTNPB or 1,25(OH)2D3 resulted in additive increases in activity. Band shift analyses showed that vitamin D receptor, RAR, or RXR alone did not bind to the VDREs; however, the combination of either vitamin D receptor or RAR with RXR led to retardation of each of the labeled probes. Treatment of nontransfected CV-1 cells with retinoids or 1,25(OH)2D3resulted in induction of 24(OH)ase RNA, and ligand combinations led to increased RNA levels. These data imply that either or both of the heterodimer partners can be occupied with ligand to induce this enzyme, with dual receptor occupation leading to increased activation.

Both receptor subfamilies are thought to mediate the biological actions of retinoids in processes such as cellular growth and differentiation and development by altering the production of certain proteins in various cells at the level of gene transcription (1,6). The ligand-occupied receptors generally act by binding to retinoid receptor-responsive elements in the promoter regions of target genes. Ligand-bound RAR cooperates with RXR to form a heterodimer that is an efficient and high affinity binder of retinoic acid response elements, thereby activating transcription of certain tRA-responsive genes such as RAR␤ (7)(8)(9) and cellular retinoic acid binding protein II (10). RXR is postulated to function by at least two modes of action. RXR has been shown to act as a silent (nonliganded) partner with a number of other intracellular receptors, including RAR, vitamin D receptor (VDR), and thyroid hormone receptor, in response to their respective ligands (11)(12)(13)(14)(15)(16). RXR has also been shown to form homodimers in a complex with DNA upon binding to 9cRA (17). However, evidence for RXR homodimers as functional units in the transcription of biologically relevant target genes has yet to be demonstrated.
For example, while RXR binds to, and stimulates transcription from, an element within the cellular retinol binding protein II gene promoter in cotransactivation assays in mammalian cells (18) and in yeast cells (which do not contain RAR; Ref. 2), this gene has not been shown to be regulated by retinoids in nontransfected cells or in the animal. However, evidence is emerging that RXR may not function solely as a silent partner in hormone signaling pathways. Recently, 9cRA has been shown to activate RXR in heterodimeric interactions with two orphan receptors, LXR (19) and NGFI-B (20), to stimulate transcription from synthetic response elements in cotransfections assays. Additionally, RXR ligands have been shown to increase the effects of RAR ligands to induce certain RNA species (21). In these experiments, however, while an RXR ligand modulated the effects of an RAR ligand to activate certain genes, the RXR ligand alone did not induce RNA levels.
We show here that an RXR-selective ligand alone is able to activate transcription of the human 25-hydroxyvitamin D 3 -24hydroxylase (24(OH)ase) promoter by binding to the RXR partner of RXR⅐VDR or RXR⅐RAR heterodimers and acting through previously defined VDRE sequences (22,23) within the promoter. Either the RXR ligand alone or the VDR ligand alone leads to stimulation of the promoter, while the presence of both ligands leads to additive or more than additive induction of luciferase activity. These data indicate that ligand occupation of either or both heterodimeric receptor partners leads to a productive transcriptional event at this promoter, with maximal induction observed upon occupation of both receptors. Additionally, ligand-occupied RAR also activates this promoter through these sequences by binding with RXR, either with or with its ligand. Therefore, these two previously defined VDREs within the human 24(OH)ase promoter can also serve as retinoic acid response elements, since they are able to confer responsiveness of the promoter to retinoids.
Receptor and Reporter Vectors-Receptor expression vectors (pRShRXR␣, pRSmRXR␥, pRShRAR␥, and pRShVDR) were as described previously (2,27). 24(OH)ase promoter-driven reporter constructs were derived from an ϳ6-kb human genomic clone (23) that included sequence 3Ј to the start site of transcription. The 6-kb promoter fragment (ϳϪ5500 to ϩ455) was cloned into a promoterless luciferase expression vector, pLUCpl (2), at SalI and PstI sites upstream of the translation start site of the luciferase coding region of the plasmid. Deletion and mutant promoter constructs were generated from the 6-kb sequence with common 3Ј ends by digestion with NsiI at position Ϫ22 in the promoter sequence (ϳϪ5500 to Ϫ22)-LUC. (Ϫ1177 to Ϫ22)-LUC was generated by digesting (ϳϪ5500 to ϩ455)-LUC with Van91I (blunted) and NsiI, and the resultant fragment was subcloned into pLUCpl digested with XhoI (blunted) and PstI. (Ϫ316 to Ϫ22)-LUC, (Ϫ294 to Ϫ22)-LUC, (Ϫ261 to Ϫ22)-LUC, and (Ϫ143 to Ϫ22)-LUC were generated by PCR utilizing 5Ј primers, some with an overhanging SalI site immediately upstream of the 5Ј-most base of the deletion construct (others without an enzyme site) and a common 3Ј primer encompassing the NsiI site. PCR products were digested with SalI (or used bluntended) and NsiI and subcloned into pLUCpl digested with XhoI (blunted or not) at the 5Ј end and PstI at the 3Ј end. (Ϫ261⌬1)-LUC ( Ϫ168 CCC mutated to GTT) was created by site-specific oligonucleotide-directed mutagenesis of the (Ϫ261 to Ϫ22) PCR product in M13, digested with SalI and NsiI, and subcloned into pLUCpl cut with XhoI and PstI. (Ϫ316⌬1)-LUC ( Ϫ168 CCC mutated to GTT) was made by digestion of (Ϫ316 to Ϫ22)-LUC with HindIII/BstEII, digestion of (Ϫ261⌬1)-LUC with BstEII/BamHI, and ligation of the two fragments into pLUCpl digested with HindIII/BamHI. (Ϫ316⌬2)-LUC ( Ϫ289 CACC to AAAA) was created utilizing the MORPH site-specific plasmid DNA mutagenesis kit (5 Prime 3 3 Prime, Inc., Boulder, CO). (Ϫ316⌬1⌬2)-LUC ( Ϫ168 CCC to GTT and Ϫ289 CACC to AAAA) was generated by digestion of (Ϫ316⌬2)-LUC with HindIII/BstEII and (Ϫ261⌬1)-LUC with BstEII/ BamHI and combining the two fragments in pLUCpl digested with HindIII/BamHI. (Ϫ316⌬3)-LUC ( Ϫ308 GGGAGGCGCGTTCG mutated to AGAAGGCGCAAATT) was generated by using PCR; the 5Ј primer contained the individual base changes as well as an XhoI site immediately upstream of position Ϫ316, and the 3Ј primer contained the NsiI site at Ϫ22. The resultant PCR product was digested with XhoI and NsiI and subcloned into pLUCpl digested with XhoI/PstI. Oligonucleotides corresponding to the individual VDREs (Ϫ294 to Ϫ274 and Ϫ174 to Ϫ151) were synthesized with HindIII ends, annealed, and subcloned into the ⌬MTV-LUC vector. The identities of all constructs were confirmed by sequencing.
Cotransfection/Cotransactivation Assays-Receptor and reporter vectors were transfected along with carrier DNA and ␤-galactosidase internal control plasmid in COS-1 cells by the use of calcium phosphate and N,N-bis(2-hydroxyethyl)-2-aminoethane-sulfonic acid-buffered saline. Briefly, transfections were performed in triplicate in 12-well plates with a total of 20 g/ml DNA (0 -0.5 g of receptor expression vector, 5-10 g of reporter plasmid, 5 g of ␤-galactosidase plasmid, and pGEM carrier plasmid to a total of 20 g). Amounts of plasmids transfected in each experiment are listed in individual figure legends. Fifteen hours later, ligands were added, followed by an additional 30-h incubation. Cells were lysed, and measurements of luciferase and ␤-galactosidase activities were as described previously (4). Luciferase values were normalized with ␤-galactosidase values to control for variable transfection efficiencies. -Fold induction was calculated by dividing the maximal response by the response elicited with vehicle. Statistical analyses were performed using analysis of variance and Fisher's protected least squares determination.
Electrophoretic Mobility Shift Assays-Oligonucleotides (Ϫ294 to Ϫ274 with BamHI and HindIII sites at each end and Ϫ174 to Ϫ151 with HindIII ends) and a PCR fragment (Ϫ314 to Ϫ121) spanning the retinoid and 1,25(OH) 2 D 3 -responsive region of the human 24(OH)ase promoter were synthesized. Oligonucleotides containing a consensus DR3 sequence (GGGAGGTCATTTAGGTCAGGG) or a DR1 motif (AGGTCA-GAGGTCA) were synthesized with HindIII, or SalI overhanging ends, respectively. The oligonucleotides and PCR fragment were end-labeled with 32 P-labeled nucleotide triphosphates. Protein extracts were prepared from the yeast strain BJ2168 transformed without or with VDR, RXR␣, or RAR␣ expression vectors (2). In reactions including antibodies, anti-VDR 9A7␥, anti-RXR␣, anti-RAR␣, or anti-estrogen receptor antibodies (28,29) were preincubated with crude protein extracts (ϳ1 l) for 1 h on ice prior to the addition of DNA probe. Reactions performed with ligand present included 9cRA (1 M, 0.1% ethanol) or ethanol vehicle and RXR-containing extracts; incubation was at 4°C for 1 h prior to the DNA addition. Protein-DNA binding reactions were carried out in buffer containing 50 -100 mM KCl, 20 mM Hepes, pH 7.4, 20% glycerol, 12.5 mM MgCl 2 . Oligonucleotide probe (ϳ20,000 dpm) was added to protein mixtures along with 1 g of poly(dI-dC) competitor DNA/reaction and incubated on ice for 20 min, followed by 2 min at room temperature. Electrophoresis of protein-DNA complexes was performed on 6% acrylamide, 0.5 ϫ TBE gels (Novex, San Diego) in 0.5 ϫ TBE running buffer at 4°C (ϳ15 mA). Gels were dried and exposed to autoradiographic film for 1-2 h at Ϫ80°C.
CV-1 Cell Studies-CV-1 cells were maintained in Dulbecco's modified Eagle's medium supplemented with L-glutamine and 10% fetal bovine serum. Cells at ϳ40% confluency were incubated in media containing 10% charcoal-stripped fetal bovine serum for 48 h prior to treatment with ligands. Cells were ϳ80% confluent upon the addition of ligands in 0.1% ethanol vehicle. 1,25(OH) 2 D 3 , LG100268, TTNPB, 9cRA, or combinations thereof were added in fresh media containing charcoal-stripped fetal bovine serum and incubated with cells for 6 h prior to harvest. Concentrations of ligands are indicated in the figure legends. Total cellular RNA was extracted, and Northern analysis was performed as per standard methodology. Probes included a 900-base pair EcoRI/XbaI DNA fragment of the rat 24(OH)ase cDNA (30) and a 1.4-kb human glyceraldehyde-3-phosphate dehydrogenase fragment (Clonetech). Hybridization was performed at 65°C in Quik-Hyb solution (Amersham Corp.) for 2 h followed by washing (0.1 ϫ SSC, 0.1% SDS at 65°C for 1 h). Quantitation was by PhosphorImager analysis (Molecular Dynamics).

RESULTS
Two regions of the human 24(OH)ase promoter are responsive to retinoids and 1,25(OH) 2 D 3 . The cloning of the rat (31, 32) and human (22, 23) 24(OH)ase promoters has been reported previously. Two VDREs have been defined within both the rat and the human promoters that confer responsiveness of the promoters to 1,25(OH) 2 D 3 (22,32). We have previously reported that retinoids induce kidney 24(OH)ase activity in mice (27), and cursory experiments indicated that the human 24(OH)ase promoter was stimulated by retinoids in CV-1 cell cotransactivation assays (27). The in vivo effects were observed with a synthetic RXR-selective ligand, LG100268 (24), a synthetic RAR-selective compound, TTNPB, and the endogenous retinoid ligands, 9cRA and tRA. To determine the mechanism of retinoid activation of the 24(OH)ase promoter, an in depth promoter study was undertaken. Various deletions of the previously cloned ϳ6-kb human 24(OH)ase promoter (23) were constructed into a promoterless luciferase reporter vector (2) and tested for the ability to respond to various ligands in kidney cell lines cotransfected with retinoid receptors and/or VDR. The promoter fragment containing ϳ5.5 kb upstream of the start site of transcription ((Ϫ5500 to Ϫ22)-LUC) yielded 3-fold induction of luciferase activity by the RXR-selective ligand, LG100268, in either of two kidney cell lines, CV-1 (data not shown) or COS-1 (Fig. 1A). COS-1 cells were used for the remainder of the study, since they gave identical results to CV-1 cells and were more stable in the transfection assays than CV-1 cells. The magnitude of the response (ϳ3-fold) to LG100268 with (Ϫ5500 to Ϫ22)-LUC was identical to that previously observed by us in the mouse (27) and was also identical to that observed with (Ϫ1177 to Ϫ22)-LUC and (Ϫ316 to Ϫ22)-LUC. (Ϫ261 to Ϫ22)-LUC yielded a less efficacious response (ϳ1.8-fold activation; p Ͻ 0.05) with LG100268, and (Ϫ143 to Ϫ22)-LUC and pLUCpl did not respond to LG100268 (Fig. 1A). Therefore, two regions within the promoter conferred responsiveness to the RXR-selective ligand LG100268, one between Ϫ316 and Ϫ261 and the other between Ϫ261 and Ϫ143 (see Fig. 2). Interestingly, these two regions are also responsive to 1,25(OH) 2 D 3 (Fig. 1B) and contain previously defined VDREs (see Fig. 2; Refs. 21 and 22). Sequence upstream of Ϫ1177 may contain an additional VDRE (Fig. 1B). The two LG100268-responsive regions also confer responsiveness of the promoter to TTNPB (Fig. 1C; p Ͻ 0.05) and to 9cRA treatment (Fig. 1D). COS-1 cells contain endogenous VDR, RAR, and RXR proteins (data not shown), and ligand-induced luciferase activity driven by the 24(OH)ase promoter sequences was also observed without transfected receptor expression vectors. The -fold induction varied from 20 to 80% of that in the presence of transfected receptors, depending on the ligand (Fig. 3). The endogenous retinoid receptor pan-agonist, 9cRA, elicited greater induction of luciferase activity through each of the promoter constructs both in the presence (ϳ6-fold; Fig. 1D) and in the absence (ϳ3-fold; Fig. 3) of transfected receptors, than either TTNPB or LG100268 alone, implying that occupancy of both RAR and RXR leads to greater activation of the promoter than either individual liganded retinoid receptor.
RXR Binds as a Heterodimer with VDR or RAR, but Not as a Homodimer, to Retinoid-responsive Sequences of the Human 24(OH)ase Promoter-Upon identification of the retinoid-responsive regions of the human 24(OH)ase promoter (see Fig. 2), the ability of those sequences to bind directly to RXR and RAR was tested. It was previously demonstrated that RXR⅐VDR heterodimers could bind to these regions (22,23). Oligonucleotides were synthesized spanning the regions from Ϫ294 to Ϫ274 and from Ϫ174 to Ϫ151. Wild type nontransformed yeast extracts did not display DNA binding activity via either of these sequences (Fig. 4, A-C). Extracts from yeast transformed with a human RXR␣ expression vector (2) were able to bind to each of these sequences in the presence of extracts from yeast expressing human VDR or human RAR␣ (Fig. 4, A-C) or human RAR␥ (data not shown). However, RXR did not bind alone to these sequences (Fig. 4, A-C) or to the entire responsive region between Ϫ314 and Ϫ121 (Fig. 4D) in the absence or presence of 9cRA ( Fig. 4D and data not shown). Conversely, RXR was able to bind to an oligonucleotide containing a consensus direct repeat separated by 1 base pair (DR1) (Fig. 4D), as previously shown (17,18). Therefore, we conclude that RXR is unable to bind as a homodimer to the retinoid-responsive elements of the 24(OH)ase promoter but that it does form heterodimers with either VDR or RAR. Also, VDR alone or RAR alone or the combination of the two receptors did not bind to any of the DNA probes (Fig. 4, A-C, and data not shown). Binding of RXR⅐RAR heterodimers to the human 24(OH)ase VDRE sequences was surprising in that the DR3 motifs that they contain are thought to bind preferentially to RXR⅐VDR heterodimers, while RXR⅐RAR heterodimers have been shown to prefer DR2 and DR5 type motifs (33). However, the 24(OH)ase promoter VDREs are not perfect consensus DR3 elements (see Fig. 2). Therefore, we tested the ability of RAR to heterodimerize with RXR on a consensus DR3 element and compared the binding with that via the 24(OH)ase VDREs. Fig.  4C shows that RXR-RAR interactions do occur on a consensus DR3-containing oligonucleotide probe (lane 4); however, this interaction is much weaker than the RXR-VDR interaction via the DR3 (lane 5). In contrast, the RXR-VDR and RXR-RAR interactions on the 24(OH)ase VDREs were both of similar high affinity (Fig. 4, A-C).
Mutations within the VDREs Abrogate Responsiveness of the Human 24(OH)ase Promoter to Retinoids-Two approaches were utilized to further delineate the cis-elements involved in the response of the promoter to retinoids: mutational analysis of the wild type promoter and the use of heterologous promoter constructs driven by the retinoid-responsive sequences. Mutations were made within the two previously defined VDREs and within DR-like elements between Ϫ316 and Ϫ291 in the context of the wild type human 24(OH)ase promoter-driven reporter vector. Point mutations within the 5Ј half-site of either the upstream VDRE (Ϫ316⌬2-LUC) or the downstream VDRE (Ϫ316⌬1-LUC) led to diminished -fold induction of luciferase activity by LG100268 or 1,25(OH) 2 D 3 (Fig. 5, A and B). Mutation of both VDREs (Ϫ316⌬1⌬2-LUC) essentially abolished the response to each of the retinoids as well as to 1,25(OH) 2 D 3 (Fig.  5, A-C). The mutations within the sequence between Ϫ316 and Ϫ294 (Ϫ316⌬3-LUC) had no effect on retinoid or 1,25(OH) 2 D 3 responsiveness (Fig. 5, A-C). Therefore, the two VDREs confer responsiveness of the promoter to retinoids as well as to 1,25(OH) 2 D 3 . The noninvolvement of the sequence between Ϫ316 and Ϫ294 was confirmed by a deletion construct containing promoter sequence from Ϫ294 to Ϫ22, which exhibited the same efficacy of response as the Ϫ316 to Ϫ22 construct to either retinoid or 1,25(OH) 2 D 3 (data not shown).
Either VDRE Confers Retinoid Responsiveness to a Heterologous Promoter: LG100268 and 1,25(OH) 2 D 3 Yield Additive Induction via Each VDRE-To confirm that the VDREs were able to function as cis-acting elements to confer responsiveness of the promoter to retinoids, oligonucleotides spanning the regions of activity were cloned into a ⌬MTV-LUC (MTV-LUC with the glucocorticoid response element deleted) reporter vector. Fig. 6 shows that either VDRE alone as a single copy ((Ϫ294 to Ϫ274)-LUC or (Ϫ174 to Ϫ151)-LUC) was able to drive increased luciferase activity in response to 1,25(OH) 2 D 3 (panels A and C) or LG100268 (panels A-C). 1,25(OH) 2 D 3 treatment resulted in 12-and 11.3-fold induction of luciferase activity from the 3Ј and 5Ј VDREs, respectively (Fig. 6A).
LG100268 alone yielded 4-and 3.5-fold induction from the 3Ј and 5Ј VDREs, respectively, while the combination of 1,25(OH) 2 D 3 and LG100268 resulted in 17.4-and 16-fold responses from the 3Ј and 5Ј VDREs, respectively, which represent additive increases in luciferase activity (Fig. 6A). TTNPB also acted through these elements on its own and additively increased the activation elicited by LG100268 alone (Fig. 6B). In these experiments (Fig. 6, A and B), receptor plasmids were used at 0.1 g/ml, and reporter constructs were at 5 g/ml, typical concentrations used in our cotransfection experiments. To ensure that the additive effects observed were not due to monomeric receptor activation of individual reporter templates instead of heterodimer action on common templates, the amount of reporter used was decreased to 0.1 g/ml. Fig. 6C shows that, using the VDRE1-⌬MTV-LUC reporter construct at this concentration, the overall luciferase values fall substantially, as expected (Fig. 6, compare panels A and B with panel  C). However, the additive effect of 1,25(OH) 2 D 3 with LG100268 or 9cRA is still observed, implying that the activity is on common templates and most likely through heterodimers. Therefore, these data show that each of the human 24(OH)ase VDREs is able to confer retinoid and 1,25(OH) 2 D 3 responsiveness to a heterologous promoter and that saturating concentrations of each ligand in combination elicit greater reporter activity than either compound alone, implying that both receptors can be occupied with ligand to yield greater activation of the promoter.  2. Human 24(OH)ase promoter sequence. The human 24(OH)ase promoter has been previously cloned (22,23). A portion of the promoter sequence (Ϫ320 to ϩ10) that includes the regions that are responsive to retinoids and 1,25(OH) 2 D 3 is illustrated. VDRE sequences are boxed; dots over bases in the VDREs denote residues that were changed in mutation constructs (see "Experimental Procedures"). The region between Ϫ316 and Ϫ291 contains sequences that are homologous with various direct repeat motifs. Potential response elements are underlined and overlined; dots over bases denote residues that were changed in mutation constructs (see "Experimental Procedures"). The TATA box is underlined.  3, 5, and 7; lanes  3, 6, and 8; lanes 4, 5, and 9; and lanes 4, 6, and 10). Similar effects were also observed with the combination of TTNPB and LG100268 versus either ligand alone (compare lanes 5, 11, and  13; lanes 6, 11, and 14; lanes 5, 12, and 15; and lanes 6, 12, and  16). The naturally occurring bifunctional retinoid, 9cRA, was also an efficacious inducer (9 -10-fold) of 24(OH)ase RNA ( lanes  17 and 18), an effect that was greater than that observed with either the RXR-selective (lanes 5 and 6) or the RAR-selective (lanes 11 and 12) ligand alone. 9cRA at 10 nM was as efficacious as 10 nM 1,25(OH) 2 D 3 in induction of 24(OH)ase RNA levels in nontransfected CV-1 cells (compare lanes 3 and 17). Additionally, 9cRA led to increased levels of 24(OH)ase RNA in combination with 1,25(OH) 2 D 3 (lanes 19 and 20). These data corroborate the information elucidated from the cotransfection/ cotransactivation assays, since they indicate that endogenous 24(OH)ase production in kidney cells can be effected either by a retinoid alone or by 1,25(OH) 2 D 3 alone at nanomolar concentrations. Furthermore, the data also show that combination treatment with retinoids and 1,25(OH) 2 D 3 or two selective retinoids leads to increased levels of 24(OH)ase RNA, as was also demonstrated in the cotransfection assays. This information taken together with the DNA-binding data lead to the conclusion that ligand occupation of either or both receptor partners results in activation of 24(OH)ase, with increased levels of stimulation observed upon liganding of both receptors.

DISCUSSION
The experiments described here indicate that RXR-selective ligands are able to activate the human 24(OH)ase promoter in cotransactivation assays in COS-1 cells through binding to RXR, which has the ability to form heterodimers with either VDR or RAR. These heterodimers form on previously defined VDREs (22,23), which as demonstrated here, also act as retinoic acid response elements. RXR is not observed to homodimerize on the retinoid-responsive sequences of the promoter as determined by electrophoretic mobility shift assays. Activation via the VDREs is achieved by specific ligands for either receptor, and the presence of both ligands leads to increased stimulation. Pan-agonists such as 9cRA lead to greater activation than either retinoid receptor-specific ligand alone. While we describe ligand-bound RXR interacting with VDR or RAR to activate this promoter, we cannot rule out the possibility of the occurrence of another partner for RXR in vivo, such as an orphan receptor (19,20). However, the experiments in nontransfected CV-1 cells (Fig. 7) showing that a combination of FIG. 4. RXR forms heterodimers with VDR or RAR, but not homodimers, on ligand-responsive human 24(OH)ase promoter sequences. Oligonucleotides corresponding to the 3Ј VDRE (VDRE1; Ϫ174 to Ϫ151; A and C), 5Ј VDRE (VDRE2; Ϫ294 to Ϫ274; B), or sequence from Ϫ314 to Ϫ121 (D) of the human 24(OH)ase promoter were radiolabeled and incubated with protein extracts (ϳ1 l) from nontransformed yeast (WT) or from yeast transformed with VDR, RXR␣, RAR␣, or combinations thereof as denoted in the figure. Receptor amounts in each extract aliquot added to the binding reaction were approximately equal as determined by ligand binding assay. Additional oligonucleotide probes included DNA sequences containing a DR3 (C) or a DR1 motif (D) (see "Experimental Procedures"). ␣VDR, ␣RXR, ␣RAR, and ␣ER denote antibodies against VDR, human RXR␣, human RAR␣, and human estrogen receptor, respectively. 9cRA (1 M) or vehicle was preincubated with RXR␣ in some reactions (D). saturating amounts of 1,25(OH) 2 D 3 and an RXR ligand yields increased levels of 24(OH)ase RNA is difficult to reconcile with the involvement of another partner.
The formation of RXR⅐VDR and RXR⅐RAR heterodimers occurs with approximately equal affinity on each of the VDREs within the human 24(OH)ase promoter as determined by electrophoretic mobility shift assays. This was somewhat surprising, since the VDREs contain DR3 motifs that have been shown to be preferential binders of RXR⅐VDR heterodimers rather than other receptor combinations (33). However, Umesono et al. (33) used consensus DR sequences for their experiments. Upon comparison of consensus DR3-containing oligonucleotides and the nonconsensus 24(OH)ase VDREs, it was apparent that while the perfect DR3 sequence did have a higher affinity for RXR⅐VDR heterodimers than for RXR⅐RAR heterodimers, the human 24(OH)ase VDREs bound both heterodimer pairs with approximately equal affinity (Fig. 4). These VDRE sequences also confer responsiveness of the promoter to retinoids and 1,25(OH) 2 D 3 in cotransactivation assays. Therefore, the VDREs within the human 24(OH)ase promoter also function as retinoic acid response elements.
The stimulation of the human 24(OH)ase promoter in COS-1 cell cotransfection/cotransactivation assays extends our previous work, which showed that retinoids, including an RXRspecific ligand, induced kidney 24(OH)ase RNA in mice within hours (27). This effect was observed in normally fed or vitamin D-deficient mice, implying that 1,25(OH) 2 D 3 was not required for the activation by retinoids. We performed cursory promoter experiments in that report, which indicated that the activation by retinoids was not dependent on the 3Ј VDRE. Our present data show that while stimulation of the human 24(OH)ase (Ϫ316 to Ϫ22) promoter sequence by retinoids is retained with mutations in the 3Ј VDRE, it is diminished, and mutations in both VDREs essentially destroy the ability of retinoids to initiate transcription from the promoter. This work on the dissection of the human 24(OH)ase promoter in cotransfection/cotransactivation assays corroborates the effects previously observed in vitamin D-deficient and normally fed mice (27), i.e. an RXR-selective ligand is able to activate 24(OH)ase in the absence or presence of 1,25(OH) 2 D 3 .
The doses of 1,25(OH) 2 D 3 that were administered to the mice maximized the induction of 24(OH)ase RNA, and the addition of LG100268 with 1,25(OH) 2 D 3 had no effect on RNA levels at 32 h postdose (27). To test the effects of combinations of retinoids and 1,25(OH) 2 D 3 at shorter times post dosing, we used CV-1 kidney cells as a model. CV-1 cells were found to produce 24(OH)ase RNA, and the levels of the RNA were modulated by 1,25(OH) 2 D 3 and retinoids. 1,25(OH) 2 D 3 induced 24(OH)ase RNA in CV-1 cells in a dose-dependent manner ( Fig. 7 and data not shown). Each of the retinoids (LG100268, TTNPB, and 9cRA) tested also induced 24(OH)ase RNA in CV-1 cells. Combinations of 1,25(OH) 2 D 3 with either LG100268 or 9cRA at saturating doses gave additive or superadditive increases in the amount of 24(OH)ase RNA that was produced. Additionally, 9cRA yielded a greater induction of RNA than either receptor-selective retinoid alone at the same concentrations. The combination of the two selective retinoids (LG100268 and TTNPB) also elicited an additive effect. Therefore, from these experiments it was demonstrated that while either a retinoid or 1,25(OH) 2 D 3 alone induced 24(OH)ase RNA, the combination of both ligands gave an increased effect. Additionally, 1,25(OH) 2 D 3 and 9cRA were approximately equipotent activators of 24(OH)ase in nontransfected CV-1 cells, since each ligand at 10 Ϫ8 M gave a similar induction of RNA levels (9.1and 9.4-fold, respectively).
We have concluded from the data described herein that Ϫ308 GGGAGGCGCGTTCG mutated to AGAAGGCGCAAATT) in the context of the Ϫ316 to Ϫ22 sequence; Ϫ261 to Ϫ22 wild type sequence in pLUCpl (Ϫ261 to Ϫ22)-LUC or a mutant of VDRE1 (Ϫ261⌬1-LUC; Ϫ168 CCC mutated to GTT) (see "Experimental Procedures" and Fig. 3 for details).
1,25(OH) 2 D 3 and retinoids exert additive effects through RXR heterodimers at the VDREs within the 24(OH)ase promoter. To rule out the possibility that retinoids induce VDR or that vita-min D up-regulates retinoid receptors in these cells, CV-1 cells were treated for 6 h with 10 and 100 nM 9cRA or 1,25(OH) 2 D 3 and 100 nM and 1 M LG100268 (concentrations that gave effects in both analyses), and receptor levels were quantitated by ligand binding assays. Neither retinoid increased VDR levels, as assayed by specific binding of CV-1 cell extracts to tritiated 1,25(OH) 2 D 3 (data not shown). Additionally, 1,25(OH) 2 D 3 treatment of the cells did not increase RARs or RXRs as assayed by specific binding of the extracts to tritiated 9cRA (data not shown). Therefore, the additive effects of retinoids and 1,25(OH) 2 D 3 via this promoter are not due to ligandinduced up-regulation of the receptor proteins.
Two other vitamin D target genes have also been shown to be regulated by retinoids: osteopontin (34,35) and osteocalcin (36 -38). Osteopontin RNA was induced in rats after a 4-h treatment with tRA regardless of vitamin A or D status and cooperated with 1,25(OH) 2 D 3 to induce increased levels of osteopontin (34). Others have used a heterologous promoter containing two copies of the osteopontin VDRE to show differential effects of retinoids in cotransfection assays (35). Osteocalcin production has been shown to be stimulated by retinoids in primary human osteoblasts, and synergistic induction was observed with tRA and 1,25(OH) 2 D 3 (37), although others have reported down-regulation of osteocalcin by 9cRA in cultured ROS17/2.8 osteosarcoma cells (36). Additionally, the human osteocalcin promoter has been shown to be stimulated by retinoids in cotransactivation assays in osteosarcoma cells through a sequence containing the VDRE (38). RXR ligands may have the ability to regulate a number of vitamin D (and thyroid hormone) target genes through perturbation of the structure of the heterodimer, which may lead to activation or repression. The potential for dual hormone regulation may depend on a number of factors including hormonal status of the organism, cellular receptor complement, promoter context, and the presence of specific receptor-interacting cofactors.
Interestingly, it has been shown that combinations of retinoids and vitamin D compounds have additive or synergistic effects in promoting apoptosis or growth inhibition in breast cancer cells (39), prostate cancer cells (40), and leukemia cells (41) and in growth inhibition and differentiation of leukemia cells (42)(43)(44). Therefore, lower concentrations of two ligands together may achieve efficacies that would require increased amounts of either compound alone. Clinically, combination therapy of retinoid and vitamin D analogues may potentially provide a drug treatment regimen that would exhibit a greater therapeutic index than either agent could achieve alone in diseases such as cancer and leukemia.