Coactivator-Vitamin D Receptor Interactions Mediate Inhibition of the Atrial Natriuretic Peptide Promoter*

We have discovered a role for coactivators binding to the AF-2 surface of the vitamin D receptor (VDR) in its negative effects on gene transcription. We tested nine amino acid residues (Ser 235 , Ile 242 , Lys 246 , Asp 253 , Ile 260 , Leu 263 , Leu 417 , Leu 419 , and Glu 420 ) in human VDR which, based on homology to the human thyroid hormone receptor, would be predicted to lie in or near the coacti-vator-binding site. Mutation of six of these residues in VDR resulted in loss of both the activation (assessed with a transfected DR3 TK luciferase reporter) and inhibition (assessed with an hANPCAT reporter) functions of the receptor when tested in cultured neonatal rat atrial myocytes and HeLa cells. Collectively, these mutations also suppressed association of VDR with the coactivators GRIP1 and steroid receptor coactivator 1 in vitro but had little or no effect on ligand binding, heterodimerization with the retinoid X receptor, or association with a VDR-specific DNA recognition element. Co-transfection with GRIP1 or steroid receptor coactivator 1 amplified both the positive and negative re-sponses to wild type VDR but had little or no effect on the functionally impaired mutants described above. The interaction between VDR and GRIP1 proved to be heav-ily dependent upon the integrity of nuclear box III in the latter protein. Mutations in this region of GRIP1 impaired its ability to associate with VDR in vitro and to

In the myocardium hypertrophic growth is characterized by increases in the expression of certain genes, including those for atrial natriuretic peptide (ANP), 1 brain natriuretic peptide, and ␣-skeletal actin (1). We have shown previously that 1,25dihydroxyvitamin D 3 , the most active metabolite of the secosteroid hormone vitamin D 3 , acts through its receptor (VDR) to suppress these genes (2)(3)(4)(5). The vitamin D-dependent reduction in gene expression is accompanied by a decrease in cell size and protein synthesis. By inference, vitamin D or vitamin D analogues might have utility as therapeutic agents in the management of disorders associated with myocardial hypertrophy. In addition, a better understanding of this inhibitory effect could provide important insights into the basic pathophysiological mechanism(s) which are activated with hypertrophy.
The VDR is a member of a family of nuclear receptors each of which have three discrete structural domains. The amino-terminal domain is not conserved among the nuclear receptors and its function is poorly understood, although in certain instances it harbors an activation domain, termed AF-1 (6). The DNA-binding domain, which lies next to the amino-terminal domain, is highly conserved across different receptor species. The carboxyl-terminal domain is the ligand-binding domain (LBD). It also contains the hormone-dependent transcriptional activation function, termed AF-2. In genes which are stimulated by vitamin D 3 , for example, the osteocalcin gene, the VDR associates through its DNA-binding domain with specific DNA recognition elements in or close to the promoters of the target genes. The canonical sequence preferred by the VDR is composed of a direct repeat of the sequence AGGTCA spaced by three nucleotides (DR3) (7), which the VDR occupies as a heterodimeric complex with the retinoid X receptor (RXR). In genes which are repressed by vitamin D 3 the essential DNA elements are less well defined. There is evidence suggesting that VDR-dependent inhibition of the parathyroid hormone gene promoter is dependent upon two imperfect direct repeats separated by a three-nucleotide spacer (8); however, precise details of the mechanism(s) linking these repeats to the regulation of transcription are lacking.
Considerable attention has been devoted to elucidation of the mechanisms governing hormone-dependent transcriptional activation through the AF-2 of the nuclear receptor LBD. While the process remains incompletely understood, an evolving picture suggests that liganded receptors are linked to the transcriptional activation process by a growing number of accessory molecules, termed coactivators which, in many cases, physically contact the receptors and transmit the ligand-dependent signal to the core transcriptional machinery (9,10). GRIP 1 and SRC-1a comprise one class of coactivators. The recently described DRIPs (11), which appear to participate in both VDR and TR signal transduction, represent a second class of coactivators that may provide direct contacts with the RNA polymerase II complex (10). Recently, Feng et al. (12) used a scanning surface mutagenesis approach based on x-ray crystallographic data (13) to define the surface residues of the thyroid hormone receptor (TR) LBD that are required for association with coactivator proteins and for the hormone-dependent stimulation of transcription. They identified a collection of 6 amino acid residues on the surface which are essential for both coactivator binding and the activation function (12). Mutation of homolo-gous residues in the estrogen receptor (ER) effected similar reductions in coactivator binding and activation function (12). These residues form a small surface on each receptor, encircling a hydrophobic cleft thought to serve as the coactivator docking site. Subsequent studies of Darimont et al. (14) with TR␤1, Shiau et al. (15) with ER␣, and Nolte et al. (16) with the peroxisome proliferator-activated receptor ␥ confirmed this prediction with structural analyses demonstrating the alignment of specific regions of the coactivator, the so-called nuclear receptor or NR boxes (see below), within the hydrophobic cleft of these receptors.
The regions of the hTR␤1, ER, and peroxisome proliferatoractivated receptor ␥ sequences which give rise to residues residing in the hydrophobic cleft are well conserved within the NR family (17). Thus, based on homology to the hTR␤1 primary amino acid sequence, as well as direct modeling of the VDR LBD (18), it is possible to identify the residues which are likely to form the equivalent surface in the hVDR. Targeted mutations of three residues within this AF-2 surface (two from the putative helix 12, Leu 417 and Glu 420 , and one from the putative helix 3, Lys 246 ) have been reported to reduce VDR-dependent activation of a conventional VDRE-driven reporter (11, 19 -23). Although these data suggest that the AF-2 of the VDR may be structurally similar to those for other nuclear receptors, limited information exists regarding the direct participation of these three residues or other residues positioned nearby in receptor-coactivator interactions (11,20,22). In addition, these questions have never been addressed in the context of the cardiac myocyte.
The structural basis of hormone-dependent inhibition by nuclear receptors is less well understood. In the case of hANP promoter repression by liganded VDR, preservation of the activation function of the receptor appears to be important (4), but the specific surface involved remains unexplored. As a first step in defining a mechanism for this repression, we wished to establish more definitively the AF-2 surface associated with activation and to investigate whether this surface plays any role in the repression function of the receptor. Therefore, we have introduced nine separate mutations into the background of the wild type VDR that are predicted to cluster on the VDR surface within or near to the putative coactivator-binding cleft. We have investigated the ability of each mutant to stimulate or inhibit promoter activity of co-transfected reporters. We have, in addition, examined the capacity of each of these mutant receptors to interact with known VDR coactivators (i.e. GRIP1 and SRC-1a) in vitro. Our findings demonstrate that the homologous cleft surface does play a critical role in the hormonedependent activation of target gene promoter activity. Surprisingly, the correlations between coactivator recruitment and the hormone-dependent inhibition of hANP promoter activity are the same as those for the hormone-dependent activation of the VDRE in every aspect tested, including the functional effects of each VDR mutation, the effects of coactivator overexpression, and the relative importance of the different NR boxes within the coactivator for conservation of functional activity. These results establish a critical role for coactivators and the VDR AF-2 surface in the inhibition of the ANP gene in cardiac cells. This finding is likely to be of more universal importance in defining the mechanism(s) underlying the repressive activities of other hormone-bound nuclear receptors.
GST Pull-down Assay-Wild type and mutant VDR proteins, GRIP1 and the GRIP1 NRBox mutant proteins were translated in vitro using [ 35 S]methionine and the TNT-Coupled Reticulocyte Lysate System (Promega, Madison, WI). GST-hRXR␣, GST-GRIP1, GST-SRC-1a, GST-TBP, GST-TFIIB, and GST-hVDR were prepared as described previously (28). The sonicated extracts were incubated with glutathione-Sepharose 4B beads for 2 h at 4°C. The concentrations of GST fusion protein bound to the glutathione-Sepharose beads were measured using the Coomassie protein reagent. The binding assay was performed using 15 g of GST protein and 3 l of 35 S-labeled soluble protein in 200 l of binding buffer (20 mM HEPES, pH 7.4, 80 mM KCl, 6 mM MgCl 2 , 10% glycerol, 1 mM dithiothreitol, 1 mM ATP, 0.2 mM phenylmethylsulfonyl fluoride, 2 g/ml aprotinin, 1 g/ml pepstatin, and 1 g/ml leupeptin) containing 2 g/ml bovine serum albumin in the presence or absence of 50 nM 1,25-dihydroxyvitamin D 3 where indicated. In selected experiments peptides encoding the nuclear receptor box II (KHKILHR-LLQDSS) or box III (ENALLRYLLDKDD) were included in the incubations. The samples were incubated for 2 h with mild agitation at 4°C. Beads were centrifuged and washed three times using 1 ml of ice-cold binding buffer. Samples bound to the beads were recovered by boiling 3 min in SDS sample buffer, size fractionated by 10% SDS-PAGE, and visualized by autoradiography. Signals were quantified by PhosphorImaging with ImageQuant (Molecular Dynamics).
Electrophoretic Mobility Shift Assay-Assays were performed using 35 S-labeled proteins and nonradioactive DNA as described previously (5). Briefly, 2 l of 35 S-labeled VDR (either wild type or mutant) was incubated, in the presence or absence of 2 l of unlabeled hRXR␣, with 10 ng of double-stranded DR3 oligonucleotide and 50 nM 1,25-dihydroxyvitamin D 3 . Reactions were carried out in binding buffer (10 mM NaHPO 4 , pH 7.6, 0.25 mM EDTA, 0.5 mM MgCl 2 , 5% glycerol) for 20 min at 23°C. DNA-protein complexes were resolved by electrophoresis on FIG. 1. Alignments of conserved amino acid residues from the LBD of hVDR and hTR␤. Alignment taken from helices 3, 4, 5, and 12 of the TR LBD x-ray structure. Residues Val 284 , Lys 288 , Ile 302 , Leu 454 , and Glu 457 in hTR␤ have previously been shown to cluster on the surface of the hydrophobic cleft which binds to p160 coactivators (12,13). Positions of homologous residues mutated in VDR (I242R, K246E, I260R, L417R, and E420K) are indicated. Leu 263 in hVDR represents the homologue of Leu 305 in hTR␤, a residue which lies deep and in the center of hydrophobic cleft but was not formally tested in the study cited above. Residues Ser 235 , Asp 253 , and Leu 419 in VDR, by analogy to their TR␤ homologues, would be expected to lie on the receptor surface adjacent (Ser 235 ) or in close proximity (Asp 253 and Leu 419 ) to the coactivator-binding site. Single letter nonmenclature for the individual amino acids is used. 5% nondenaturing polyacrylamide gels run in TEA buffer (67 mM Tris, pH 7.5, 10 mM EDTA, 33 mM sodium acetate) at 240 V for 3 h at 4°C. The gel was washed (3 times) with 30% methanol and 10% glacial acetic acid, amplified for 30 min (Amplifier, Amersham Pharmacia Biotech), dried, and subjected to autoradiography overnight with an intensification screen. The results were also quantified by PhosphorImaging analysis as described above.
Cell Culture and Transfection-Atrial myocytes from 1-2-day-old neonatal rat hearts were separated by alternate cycles of trypsin digestion and mechanical disruption (31). Myocytes were transfected by electroporation (280 V and 250 microfarads), using plasmids indicated.

FIG. 3. Primer extension analysis demonstrates that liganded VDR directly suppresses the hANP promoter.
Atrial myocytes were transfected concomitantly with hANP CAT (40 g) and/or RSV-CAT (20 g) reporters, in the presence or absence of hVDR (10 g) and hRXR␣ (10 g) expression vectors. Twenty-four hours following transfection, cultures were changed to serum substitute medium, pretreated with cycloheximide (CHX;10 g/ml) or vehicle for 1 h, then treated with 1,25-dihydroxyvitamin D 3 (10 nM) for 6 h. Total RNA was then collected and primer extension was carried out as described under "Materials and Methods." Lanes at the right show RNA from cells transfected with RSV-CAT or hANP CAT alone. Representative experiment is presented in panel A. Pooled data from four independent experiments (mean Ϯ S.D.) are presented as a normalized hANP CAT/RSV CAT ratio in Panel B.

FIG. 2. Functional activity of individual hVDR mutants. Panel
A, HeLa cells were co-transfected with 1.5 g of wild type hVDR (or the mutant indicated), 2.5 g of DR3-LUC, in the presence or absence of 1.5 g of hRXR␣. RSV-␤-galactosidase (0.2 g) was included to control for transfection efficiency. After transfection, cells were treated with 10 nM 1,25-dihydroxyvitamin D 3 or vehicle for 48 h. Lysates were generated and luciferase and ␤-galactosidase activities were measured. Luciferase data are normalized for ␤-galactosidase expression. Pooled data from four to five experiments are shown. Panel B, atrial cells were transfected with 2.5 g of wild type VDR (or the mutant indicated), 5 g of DR3-luciferase, and 2.5 g of hRXR␣, where indicated. Transfections also contained 0.4 g of RSV-␤-galactosidase for normalization purposes. Twenty-four hours following transfection, cells were treated with 10 nM 1,25-dihydroxyvitamin D 3 for 48 h. Pooled data from four independent experiments are shown. Panel C, wild type VDR or one of its mutants (3 g) was co-transfected into atrial myocytes, in the presence or absence of RXR␣, with 12 g of Ϫ466 hANP CAT and 0.4 g of RSV-␤-galactosidase. Cells were cultured as described in panel B. CAT activity was normalized for ␤-galactosidase expression. Data shown were pooled from six independent experiments and represent mean Ϯ S.D.
All transfections were normalized for equivalent DNA content with pUC18. Transfected cells were cultured in Dulbecco's modified Eagle's medium-H21 (DME-H21) containing 10% bovine calf serum (Hyclone, Logan, UT) for 24 h before switching to serum-substitute medium (32). Cultures were then treated with 10 nM 1,25-dihydroxyvitamin D 3 or vehicle for 48 h. HeLa cells were maintained in DME-H21 medium containing 10% fetal bovine serum. Cells were transfected by electroporation (250 V, 960 microfarads) and immediately exposed to 1,25dihydroxyvitamin D 3 or vehicle for 48 h. CAT assay was carried out on cellular lysates as described previously (33). Luciferase activity was measured using the Luciferase Assay System (Promega). Transfection efficiency was normalized for ␤-galactosidase activity in the individual cultures. ␤-Galactosidase activity was measured using the the Galactolight Plus chemiluminescence assay (Tropix, Bedford, MA). To assess efficacy of cycloheximide treatment, myocytes were cultured in 24-well dishes at a density of 2 ϫ 10 5 cells/well in DME-H21 containing 10% bovine calf serum for 24 h. Medium was changed to DME-H21/serum substitute at that point and cultures were continued for an additional 40 h. Cells were treated with different concentrations of cycloheximide for 1-2 h, then pulsed with 2 Ci/ml [ 3 H]leucine in leucine-free medium containing the same concentration of cycloheximide for 4 h. Culture media was removed, cells were washed three times with phosphatebuffered saline and extracted with 10% trichloroacetic acid for 30 min at 4°C. Cellular residues were rinsed with 95% ethanol, solubilized in 0.25 N NaOH at 4°C for 2 h, and neutralized with 2.5 M HCl, 1 M Tris-HCl (pH 7.5). Radioactivity was then measured in a liquid scintillation counter.
Primer Extension Analysis-For primer extension, Ϫ466 hANPCAT and RSV-CAT were co-transfected, in the presence or absence of VDR or RXR expression vectors into neonatal atrial myocytes. Cells were cultured in DME-H21, 10% bovine calf serum for 24 h before switching to DME-H21, 10% serum substitute for 40 h. At that point cells were pretreated with cycloheximide, at a concentration (10 g/ml) which provided Ͼ95% reduction in [ 3 H]leucine incorporation into protein in the experiments described above (see "Cell Culture and Transfection"), or vehicle alone for 1 h, then treated with 1,25-dihydroxyvitamin D 3 or vehicle for 6 h. Total RNA was extracted using QIAshredder and the RNeasy Mini-Kit (Qiagen; Valencia, CA). A cDNA oligonucleotide primer encoding sequence from the 5Ј end of the CAT coding sequence (5Ј-TATCAACGGTGGTATATCC-3Ј) was end-labeled with T4 polynucleotide kinase and [␥-32 P]ATP and separated from free nucleotide using the QIAquick Nucleotide Removal Kit (Qiagen). Primer extension was carried out using the Primer Extension-AMV Reverse Transcriptase System according to the instructions provided by the manufacturer (Promega). DNA products were heated for 10 min at 90°C and run at 250 volts in 0.5 ϫ TBE on 8% acrylamide gels containing 7 M urea. Gels were dried and exposed to x-ray film with intensification for 2-3 days. Radiographic signals were also quantified by PhosphorImaging with ImageQuant.

RESULTS
We carried out a sequence alignment (hTR␤1 versus hVDR) to identify probable homologues of those residues recently identified in hTR␤1 (12) as being critically involved in the activation function of that receptor (Fig. 1). We identified Ile 242 , Lys 246 , Ile 260 , Leu 417 , and Glu 420 in hVDR which, based on homology to Val 284 , Lys 288 , Ile 302 , Leu 454 , and Glu 457 in hTR␤1, would be predicted to cluster on the surface of hVDR in the region believed to participate in coactivator interactions. Residue Leu 263 was also identified as a likely participant in forming the hVDR coactivator-binding surface because the equivalent residue in the hTR␤ (L305), although it was not tested directly in the study of Feng et al. (12), is positioned within the center of the coactivator-binding hydrophobic cleft. Mutations were created in each of these sites to examine their effects on the functional properties of the resultant receptors. Residues Ser 235 , Asp 253 , and Leu 419 in hVDR (homologues of Thr 277 , Glu 295 , and Leu 456 in hTR␤) are predicted to lie on the surface adjacent to, but outside, the coactivator-binding site (based on the hTR␤1 analysis). Mutations in these sites were created as comparative controls.
We began by examining the ability of each of the mutants described to increase transcriptional activity from a conventional DR3-luciferase reporter in cultured HeLa cells and atrial myocytes. Liganded wild type hVDR effected a 2-3-fold ( Fig.  2A) or 3-4-fold (Fig. 2B) increase in reporter activity in HeLa cells or atrial myocytes, respectively. This was increased modestly by co-transfection with unliganded hRXR␣ in HeLa cells, while co-transfection with hRXR␣ had a more dramatic effect in amplifying wild type VDR activity in atrial myocytes, implying lower levels of endogenous RXR in the myocytes versus FIG. 4. Electrophoretic mobility shift assay of wild type and mutant VDR. 35 S-Radiolabeled wild type or mutant VDR was incubated with unlabeled RXR␣ and the DR3 oligonucleotide in the presence of 50 nM 1,25-dihydroxyvitamin D 3 . Reaction mixtures were then separated on non-denaturing polyacrylamide (5%) gels. Gels were dried, amplified, and subjected to autoradiography. The experiments were repeated four times and a representative autoradiograph is shown. HeLa cells (Fig. 2, A and B). Each of the candidate mutations with predicted impairment of the receptor-coactivator interaction (i.e. I242R, K246E, I260R, L263R, L417R, and E420K) demonstrated reduced stimulation of reporter activity relative to that seen with the wild type hVDR in HeLa cells or atrial myocytes (Fig. 2, A and B). In fact, in no instance (even in the presence of co-transfected hRXR␣) did the response appear to exceed that seen in cells lacking exogenous hVDR (endogenous VDR effected an ϳ2-fold increment in reporter activity in the presence of transfected hRXR␣). The negative controls D253R and L419R behaved like wild type VDR suggesting that these are truly "neutral" mutations. S235R, however, displayed a phenotype intermediate between the impaired mutants and wild type receptor, implying that it may play a limited role in coactivator binding and subsequent transcriptional activation. We were particularly interested in examining the effects of these mutations on the transcriptional inhibitory activity of the liganded VDR. As noted previously (2)(3)(4)(5), the VDR-dependent inhibition of hANP promoter activity is ligand-and RXR-dependent (Fig. 2C). The neutral mutations, D253R and L419R, demonstrated inhibitory activity which was close to that seen with the wild type receptor. Candidate functional mutations, on the other hand, were almost devoid of inhibitory activity when transfected alone. Some (e.g. K246E and I260R) displayed some inhibitory activity when co-transfected together with RXR␣ while others (e.g. I254R, L263R, L417R, and E420K) did not. Even in those instances where RXR-dependent activity was seen, it did not exceed that attributable to the presence of endogenous VDR. Once more, the S235R mutant displayed an intermediate phenotype and, again, from a quantitative standpoint, this more closely approximated the response seen with the candidate mutations than the wild type receptor.
We were surprised that the same residues found to be essential for the activation function of the VDR were also required for transcriptional suppression. This apparent discrepancy might be explained if the suppression of hANP gene transcription resulted from a VDR-dependent increase in production of a suppressor protein which, secondarily, reduced ANP promoter activity. To address this issue, we examined the ability of liganded VDR to suppress the hANP promoter in the absence of protein synthesis. Primer extension analysis presented in Fig.   FIG. 5. Analysis of wild type and mutant VDR binding to nuclear receptor coactivators. Panel A, wild type and mutant VDR were synthesized in vitro and incubated with Sepharose beadimmobilized GST-RXR␣, GST-GRIP1, GST-SRC-1a, GST-TBP, GST-TFIIB, or GST alone in the presence of 50 nM 1,25dihydroxyvitamin D 3 . After extensive washing, bound radiolabel was removed from the beads by boiling and subjected to 10% SDS-PAGE. Gels were then dried and subjected to autoradiography. A representative autoradiograph is shown.
Panel B, bar graph shows results from four to six experiments, quantitated by PhosphorImager analysis. Data represent mean Ϯ S.D.
3 demonstrates that the vitamin D-dependent inhibition of hANP promoter activity persists in the presence of cycloheximide. Thus, the inhibition does not require new protein synthesis, arguing in favor of a primary rather than a secondary effect.
A trivial explanation of our data would apply if any of these mutations, all of which lie in the LBD of hVDR, perturbed binding of vitamin D to the receptor. To explore this issue, we synthesized wild type and mutant receptors in vitro and examined their ability to bind [ 3 H]1,25-dihydroxyvitamin D 3 using a conventional dextran-coated charcoal assay (34). As shown in Table I, wild type receptor associated with the radioligand with a K D of 470 pM and a total binding capacity of 950 pM. Each of the mutants displayed a binding affinity and binding capacity similar to that shown for the wild type receptor. The most variant mutant, L263R, demonstrated a modest (Ͻ2-fold) increase in binding affinity but ϳ50% of the binding capacity seen with the other receptors. Thus, with the possible exception of L263R, impairment of ligand binding does not seem to account for the loss of functional activity seen with any of the candidate mutants.
Next, we examined the ability of each of the mutants to bind to a cognate VDR recognition element (DR-3) in the electrophoretic mobility shift assay. As expected, VDR alone was incapable of binding to DNA but bound very avidly in the presence of RXR (Fig. 4). Each of the mutants, with the exception of L263R, bound to DNA as well as the wild type VDR when RXR␣ was included in the reaction mixture. The L263R mutant showed a partial (ϳ50%) reduction in DNA binding. Thus, again with a single exception, the candidate mutations behave like the wild type receptor making it unlikely that differences in DNA binding can account for the loss of functional activity.
The interaction of the hVDR mutants with several different proteins was explored further using a GST pull-down assay (Fig. 5A). As shown in the top panel, modest impairment (55%) of RXR interaction was observed only with L263R. When the same wild type versus mutant receptors were examined for their ability to associate with the coactivators GRIP1 or SRC-1a, much more pronounced effects were seen (Fig. 5, A and B). The candidate mutants, but not the neutral mutants, uniformly demonstrated significant reductions in coactivator binding. In concert with the functional data presented above, S235R displayed an impaired, but not absent, ability to interact with either GRIP1 or SRC-1a. Low level associations with the core transcription factors TBP and TFIIB were identified with the wild type receptor, the latter confirming the observations of Blanco et al. (35) and MacDonald et al. (36). However, none of the mutant VDRs, with the possible exception of K246E which consistently demonstrated a modest increase in TFIIB binding, displayed binding activity which was substantially different from that seen with the wild type receptor (Fig. 5B). Collectively, these data offer a mechanistic explanation for the observed reduction in functional activity described above (i.e. impaired coactivator binding leads to a reduction in functional activity). It also accounts for the unexpected reduction in S235R activity identified in Fig. 2 since this mutation clearly results in unanticipated interference with the coactivator interaction. Inferentially, the data also argue that it is coactivator binding and not binding to the core transcription factors (i.e. TFIIB and TBP) at this receptor interface which ultimately governs the functional response to the receptor. A similar conclusion has been reached by others (19,20).
If the functional impairment is a reflection of reduced coactivator binding, one might predict that increasing cellular levels of the coactivator could, at least partially, offset the functional defect (12). GRIP1 effected a dose-dependent increase in liganded VDR-dependent DR-3 luciferase activity in HeLa cells when co-transfected with either the wild type receptor or neu-tral mutations, or when expressed solely with endogenous receptor (control group) (Fig. 6A). The candidate mutations increased activity modestly in the presence of increasing concentrations of GRIP1 but, in no case, did the increase approach wild type levels. In most instances this increment did not exceed that seen in the absence of transfected receptor, implying that the GRIP1 effect results from amplification of endogenous receptor activity rather than partial recovery of activity in the mutant receptors. Again, S235R displayed an intermediate phenotype which approached complete normalization in the presence of the highest concentration of GRIP1 (8 g). The GRIP1-dependent increment in promoter activity was not cell type specific. Virtually identical findings were obtained using cultured atrial myocytes, although, in most cases, the   FIG. 7. SRC-1a amplifies wild type and mutant VDR activity. Panel A, increasing concentrations of SRC-1a were transfected into HeLa cells together with the DR3-luciferase reporter (2.5 g), RSV-␤-galactosidase (0.2 g), and either wild type or mutant VDR (1.5 g). Luciferase measurements were normalized for ␤-galactosidase activity and expressed as fold activation over controls. Panel B, atrial myocytes were transfected as in A except that Ϫ466 hANP CAT (12 g) was substituted for DR3 luciferase. Data (mean Ϯ S.D.) presented in A and B were each pooled from three independent experiments. magnitude of the induction was less than that observed in HeLa cells (Fig. 6B).
To our surprise, the GRIP1 dependent activity also extended to VDRs suppression of hANP promoter activity. As shown in Fig. 6C, the addition of increasing concentrations of GRIP1 expression vector led to amplification of the wild type VDR-dependent reduction in hANP promoter activity. In the case of most of the mutants, hANP promoter activity fell appreciably but did not approach that seen with the wild type receptor. Again, S235R displayed a response intermediate between the candidate mutations and the wild type receptor.
A second coactivator, the steroid receptor coactivator 1 (SRC-1a), demonstrated similar effects on VDR-dependent promoter activity in HeLa cells. As shown in Fig. 7A, SRC-1a amplified the response to endogenous as well as transfected VDR. The neutral mutation D253R responded to SRC-1a co-transfection in a manner equivalent to that seen with wild type VDR while the candidate mutation I242R displayed a truncated response similar to that seen in the control group where only endogenous receptor is present. S235R responded in a fashion intermediate between the wild type VDR and I242R.
As noted above with GRIP1, a parallel, albeit inverted, response was seen when SRC-1a was co-transfected with the hANP CAT reporter into cultured atrial myocytes. As shown in Fig. 7B, SRC-1a effected a dose-dependent reduction in reporter activity in the presence of either endogenous or transfected wild type VDR. I242R, the functionally impaired mutant, displayed only a modest response to co-transfected SRC-1a while S235R, again, showed the intermediate phenotype (between I242R and wild type VDR).
We next focused on identification of the VDR-binding sites on the coactivator molecule. Previous studies have reported that receptor interactions involve contacts with the NR boxes presumed to lie on the coactivator surface (26,37,38). These NR boxes harbor a canonical sequence (LXXLL, where L is leucine and X is any amino acid) flanked by amino acids which appear to impart specificity to the protein-NR box interaction. Three NR boxes have been identified in the GRIP1 molecule. Mutation of either NR Box II or III leads to a moderate-to-severe decrease in the functional interaction between the GRIP1 coactivator and the ER, TR, or glucocorticoid (GR) receptor in transfected HeLa cells (26). The ER and TR are preferentially affected by mutations in NR Box II while the GR is more severely impacted by mutations in NR Box III. We explored the role of NR Box II and NR Box III in promoting the VDR interaction using GRIP1 molecules harboring mutations at one or both sites. As shown in Fig. 8A, GST-VDR successfully pulled down wild type GRIP1 and, to a lesser extent, a GRIP1 mutant harboring a mutation in NR Box II. Mutation of NR Box III, on the other hand, almost completely abrogated the GST-VDR/GRIP1 interaction. Mutation of both NR boxes effected the highest level of inhibition (Fig. 8B). This was confirmed using synthetic competitor peptides encoding either the NR box II or NR box III core (Fig. 8, C and D). In this case, the interaction of GST-GRIP with wild type VDR was almost completely blocked by inclusion of NR Box III peptide in the incubation. NR Box II was much less potent in blocking this interaction.
At a functional level, mutation of NR Box III in the GRIP1 molecule significantly reduced amplification of VDR-stimulated, DR3-dependent promoter activity in HeLa cells (Fig. 9A). Mutation of NR Box II had minimal effect and mutation of Box II and III together effected a reduction which was no greater than that seen with the NR Box III mutation alone. Similar findings were obtained using the hANP promoter in cultured atrial myocytes (Fig. 9B). Mutation of NR Box III virtually completely eliminated GRIPs ability to amplify VDR-dependent inhibition of the ANP promoter. Mutation of NR Box II had little effect on GRIPs ability to inhibit the ANP promoter. DISCUSSION In the present study we have explored the molecular mechanism(s) underlying both the activating and inhibitory functions of the liganded VDR. Based on analogy to the published structures of TR alone (13) and ER (15) and peroxisome proliferator-activated receptor ␥ (16) bound to coactivator fragments, the residues selected for mutation are predicted to lie on a surface surrounding a hydrophobic cleft thought to function as a coactivator docking site. Our findings indicate that these mutations (putatively residing within helices 3, 5, and 12 in the VDR) impair both the activation and inhibitory functions of the liganded VDR. This loss of activity correlates with a decrease in coactivator (GRIP1 or SRC-1a) binding but not with changes in receptor binding to ligand or its cognate recognition site on DNA, nor with changes in receptor-dependent interactions with RXR␣, TBP, or TFIIB. Collectively, these data support the hypothesis that it is coactivator binding that mediates both of FIG. 8. Role of NR boxes in binding of GRIP1 to VDR. Panel A, radiolabeled GRIP1 or one of its NR-box mutants was incubated with bead-immobilized GST-VDR in the presence of 50 nM 1,25-dihydroxyvitamin D 3 . Beads were washed, boiled, and size-fractionated on denaturing polyacrylamide gels. Dried gels were then subjected to autoradiography. A representative autoradiograph is shown. Panel B, results of three experiments were quantitated by PhosphorImaging and pooled for analysis. Panel C, radiolabeled VDR was synthesized in vitro and incubated with GST-GRIP1 and 50 nM 1,25-dihydroxyvitamin D 3 in the presence or absence of synthetic peptide encoding NRBox II or NRBox III. Bound VDR was detected as in Panel A. Panel D, results from three experiments were pooled for analysis. All data represent mean Ϯ S.D. the hormone-dependent functions, positive and negative, of the VDR.
The residues which we selected for mutation were largely based on the structural (13,39) and subsequent functional (12) analyses of TR and ER. Thus, it appears that the predicted structural homology subserves a conserved functional role. A possible exception is residue Ser 235 in VDR, whose homologue in TR␤ (Thr 277 ) appears not to be involved in the TR-coactivator interaction (13). Noteworthy, the structural data (13) demonstrate that Thr 277 in TR␤ contacts a second residue (Leu 454 ) which is located on the coactivator-binding surface of TR. We would speculate that similar contacts between homologous residues in VDR might play a more important role in supporting the receptor-coactivator interaction (i.e. that Ser 235 contacts a second residue on the coactivator-binding surface of VDR).
S235R displayed an intermediate phenotype between wild type receptor and the more severely affected mutants when assessed for either stimulatory or inhibitory activity. The functional discrepancy (i.e. VDR versus TR) likely reflects subtle differences in the surface structures of the two nuclear receptors (i.e. mutations that disrupt contacts between Ser 235 and residues on the coactivator-binding surface in VDR perturb receptor structure to a greater extent than do homologous mutations in Thr 277 of the TR␤). By way of contrast, VDR residues Asp 253 and Leu 417 , where mutations have no effect on any VDR function, are predicted to lie beyond contact with the coactivatorbinding surface.
As predicted, L263R displays a "null" phenotype with regard to DR3-luciferase activation or hANP-CAT inhibition (see Fig.  2). However, it also displays a modest reduction in ligand FIG. 9. Mutation of NR Box III eliminates GRIP-1 amplification of VDR activity. Panel A, HeLa cells were cotransfected with DR3 luciferase (2.5 g), RSV-␤-galactosidase (0.2 g), and increasing concentrations of wild type GRIP-1 or GRIP-1 mutated at NR Box II, NR Box III, or both. Luciferase activity is normalized for ␤-galactosidase expression. Pooled data (mean Ϯ S.D.) from three independent experiments are shown. Panel B, atrial myocytes were cotransfected with Ϫ466 hANP CAT (12 g) and the GRIP-1 expression vectors described above. Normalized data (mean Ϯ S.D.) from three independent experiments are shown.
binding activity (ϳ50% of wild type), diminished heterodimer assembly with RXR␣, almost complete loss of interaction with the coactivator GRIP1, and partial reduction in DNA binding, presumably a reflection of the impairment in heterodimer formation. Both the reduction in ligand binding and the inhibition of heterodimeric assembly with RXR␣ are moderate in magnitude (ϳ50% reduction) and the functional impact of the mutation was not overcome by inclusion of supraphysiological concentrations of 1,25-dihydroxyvitamin D 3 or higher levels of co-transfected VDR or RXR␣ in these cultures (data not shown). The coactivator interaction, on the other hand, was inhibited almost completely by the mutation, a finding that parallels more closely the reduction in functional activity.
Thus, it appears probable that the defect in coactivator binding is of dominant importance here. The homologue of Leu 263 in the TR (Leu 305 ) is at the center and deep within the hydrophobic, coactivator-binding cleft (12,13). It is possible that Leu 263 , in addition to supporting the VDR-coactivator interaction, also helps to maintain the overall structural conformation of the LBD. Thus, mutation of this residue to Arg could perturb the LBD structure sufficiently to explain the effects on ligand binding and heterodimerization.
To our surprise we found that the vitamin D 3 -dependent repression of the ANP gene promoter requires the same coactivator binding surface of the VDR essential for activation at the VDRE. Furthermore, the activation and inhibition functions of the receptor were similarly impacted by overexpression of GRIP1 or by mutation of NR Box III in the GRIP molecule. Collectively, these findings support a role for coactivators in vitamin D-mediated inhibition of gene transcription, suggesting that coactivators participate in both the inhibitory and activation functions of nuclear receptors. This seemingly paradoxical finding implies that the functional outcome of the nuclear receptor-coactivator interaction must be determined, at least in part, by other factors. A similar situation involving an activation function of the unliganded TR following corepressor association has been reported (40). Thus, it appears likely that the activation properties of nuclear receptors rely on more than a single mechanism. The same conclusion may be drawn from studies of NR-dependent suppressor activity. T 3 -dependent repression of AP-1 elements by TR␣ (41) and TR␤ (12), for example, does not appear to require an intact coactivator binding surface.
In summary, we have established that residues residing in putative helices 3, 5, and 12 of hVDR collectively constitute a coactivator-binding surface. Several of these residues have not previously been linked to the activation function of VDR. Intriguingly, these same residues appear to be involved in mediating VDRs inhibitory transcriptional properties. This latter finding implies the existence of an undiscovered pathway linking nuclear receptor recruitment of coactivators to repressive gene regulation.