Vitamin D-dependent Suppression of Human Atrial Natriuretic Peptide Gene Promoter Activity Requires Heterodimer Assembly*

Crystallographic structures of the ligand-binding domains for the retinoid X (RXR) and estrogen receptors have identified conserved surface residues that participate in dimer formation. Homologous regions have been identified in the human vitamin D receptor (hVDR). Mutating Lys-386 to Ala (K386A) in hVDR significantly reduced binding to glutathione S-transferase-RXRα in solution, whereas binding of an I384R/Q385R VDR mutant was almost undetectable. The K386A mutant formed heterodimers with RXRα on DR-3 (a direct repeat of AGGTCA spaced by three nucleotides), whereas the I384R/Q385R mutant completely eliminated heterodimer formation. Wild type hVDR effected a 3-fold induction of DR-3-dependent thymidine kinase-luciferase activity in cultured neonatal rat atrial myocytes, an effect that was increased to 8–9-fold by cotransfected hRXRα. Induction by K386A, in the presence or absence of RXRα, was only slightly lower than that seen with wild type VDR. On the other hand, I384R/Q385R alone displayed no stimulatory activity and less than 2-fold induction in the presence of hRXRα. Qualitatively similar findings were observed with the negative regulation of the human atrial natriuretic peptide gene promoter by these mutants. Collectively, these studies identify specific amino acids in hVDR that play a critical role in heterodimer formation and subsequent modulation of gene transcription.

The nuclear hormone receptors are a family of ligand-regulated transcription factors that associate with cognate recognition sequences in close proximity to target gene promoters and through an, as yet, incompletely understood process regulate their transcriptional activity (1,2). There are two major classes of nuclear hormone receptors. Class I receptors, which encompass the steroid hormone receptors (i.e. receptors for glucocorticoids, mineralocorticoids, progestins, androgens, and estrogens), typically bind as homodimers to palindromic sequences encoding the core recognition sequence. Class II receptors, which include the vitamin D receptor (VDR), 1 thyroid receptor (TR), and retinoic acid receptor (RAR), bind to direct repeat (DR) elements as heterodimeric complexes with unliganded retinoid X receptor (RXR) (see below). In contrast to class I receptors that invariably recognize a palindrome spaced by 3 base pairs, the class II receptors bind to DRs spaced by a variable length of nucleotides. This spacing contributes to specificity in the types of receptors that associate with a given recognition sequence (3).
VDR has been shown to interact with canonical recognition elements termed vitamin D response elements (VDREs) in a variety of target genes. In some cases these recognition elements function in a stimulatory mode (e.g. osteopontin (4), osteocalcin (5,6), calbindin (7), 24-hydroxylase (8), and ␤ 3 integrin (9)), whereas in others (e.g. parathyroid hormone (PTH) (10,11) and parathyroid hormone-related protein (PTHrP) (12,13)) it is clearly inhibitory. Although there is considerable sequence variation among the stimulatory VDREs, the general structure suggests conservation of two direct repeats of a consensus (A/G)G(G/T)TCA separated by a three-nucleotide spacer (DR-3). VDR typically associates with this element as a heterodimeric complex with RXR prior to effecting changes in transcriptional activity. VDR homodimers have been described (14,15), most notably with the receptors in an unliganded form (15); however, it is generally accepted that the ligand-dependent assembly of VDR-RXR heterodimeric complexes on the VDRE is the dominant pathway leading to vitamin D-dependent activation of gene expression. Less information is available regarding the inhibitory effects of liganded VDR on gene expression. In the case of PTH and PTHrP, the responsible element contains only one of the two tandem hexameric sites found in VDREs involved in positive gene regulation, and RXR does not appear to be involved in mediating the inhibitory effect (10 -13).
As already noted, RXR serves as a heterodimeric partner for a variety of different nuclear receptors (1,2). In addition, RXR can, in the presence of its cognate ligand 9-cis-retinoic acid (9-cis-RA), assemble as homodimers on a recognition sequence containing two DRs separated by a single nucleotide spacer (DR-1) (3). Thus, there are a number of pathways by which this receptor can regulate downstream transcriptional activity.
We have recently demonstrated that formation of VDR-RXR heterodimers is important for activation of a DR-3-dependent reporter in cultured neonatal rat atrial myocytes (16). However, whereas VDR-dependent inhibition of ANP promoter ac-tivity is amplified by cotransfection with RXR, the dependence of this inhibition on heterodimerization of these two receptors remains unclear. A VDR mutant (L262G), which demonstrates impaired heterodimerization with RXR (17), retains the ability to suppress human atrial natriuretic peptide (hANP) promoter activity (16) in transfected myocytes. Thus, the dependence of hANP promoter suppression on heterodimer formation remains open to question.
The crystallographic structures of RXR␣ (18) and the estrogen receptor (ER) (19) imply an important role for several surface residues within helix (H) 10 (for RXR) and 11 (for ER) in homodimer formation. This region resides within the 9th heptad repeat proposed for the TR (20) and has been shown to be highly conserved in other receptors of this class, suggesting conservation of the structural determinants that govern dimerization in this family of regulatory proteins. In fact, mutations have already been reported in this region of VDR (21), RXR (22), and TR␤ (23) that appear to interfere with dimer formation. However, based on the available structures (e.g. that for TR␣), a number of these mutations target residues placed internally in the receptor molecule (24) where they might easily effect disruption of receptor folding and structural integrity. We (25) have recently shown that mutation of selected surface residues in the 9th heptad of TR interferes with dimer formation yet preserves other receptor functions such as ligand binding, DNA binding, and coactivator interactions. Based on analogy to TR, we have placed homologous surface mutations in human (h) RXR and hVDR in positions predicted to interfere selectively with dimer interactions but not with binding to DNA, ligand, or the relevant coactivators. We have examined the effects of these mutations on dimer formation and functional activity in a transfected atrial myocyte model.
GST Pull-down Assay-Wild type or mutant pSG5 hVDR vectors and wild type or mutant pEThRXR␣ vectors were used to produce radiolabeled full-length receptors in vitro using the TNT-Coupled Reticulocyte Lysate System (Promega, Madison, WI) and [ 35 S]methionine. GST-hVDR, GST-hRXR␣, and GST-GRIP 1 fusion proteins were prepared using conventional protocols (29). Briefly, the plasmids were transformed into HB101, amplified in culture, pelleted, resuspended in buffer IPAB-80 (20 mM HEPES, 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) and sonicated (three times) for 10 s. The debris was pelleted; the supernatant was incubated for 2 h with 500 l of glutathione-Sepharose 4B beads and equilibrated with 5 volumes of IPAB-80. GST fusion protein beads were washed with 5 volumes of phosphate-buffered saline containing 0.05% Nonidet P-40 and resuspended in 0.5 ml of IPAB-80. All procedures above were carried out at 4°C. The concentrations of GST fusion proteins were measured using the Coomassie protein reagent.
For the binding assay, the glutathione bead suspension containing 10 g GST protein was incubated with 2 l of 35 S-labeled protein in 150 l of IPAB-80 buffer containing 2 g/ml bovine serum albumin, in the presence of 10 nM 1,25-dihydroxyvitamin D 3 or vehicle. After incubation for 2 h at 4°C, the beads were washed (three times) using 1 ml of IPAB-80 buffer. The beads were then heated to 100°C for 3 min; associated proteins were subjected to 10% SDS-PAGE and visualized by autoradiography. The results were analyzed using NIH image. Electrophoretic Mobility Shift Assay-Gel shift assays using 35 Slabeled proteins and nonradioactive DNA were performed as described previously (31). In brief, 3 l of 35 S-labeled proteins were incubated with 10 ng of oligonucleotide in DNA binding buffer (10 mM NaHPO 4 , pH 7.6; 0.25 mM EDTA; 0.5 mM MgCl 2 ; 5% glycerol) for 20 min at room temperature in the presence or absence of 100 nM of the appropriate ligand. The reaction mixtures were separated on 5% nondenaturing polyacrylamide gels in TEA buffer (67 mM Tris, pH 7.5; 10 mM EDTA; 33 mM sodium acetate). The gel was run at 240 V for 3 h at 4°C, washed extensively with 30% methanol and 10% glacial acetic acid, and amplified for 30 min (Amplifier; Amersham Pharmacia Biotech), dried, and exposed for autoradiography.
Cell Culture and Transfection-Atrial cells were obtained from 1-to 2-day-old neonatal rat hearts by alternate cycles of trypsin digestion and mechanical disruption as described previously (30). The cells were transfected by electroporation (280 V and 250 F) using the plasmids indicated. All transfections were normalized for equivalent DNA content with PUC18. After transfection, cells were resuspended in Dulbecco's modified Eagle's medium H21 containing 10% bovine calf serum (HyClone, Logan, UT) and cultured for 24 h. At that time medium was changed to Dulbecco's modified Eagle's medium/serum substitute (32), and the cultures were treated with 10 nM 1,25-dihydroxyvitamin D 3 alone or in combination with 10 nM 9-cis-RA for 48 h. Similar concentrations of ligand-free vehicle were used as controls. Cells were washed with phosphate-buffered saline and lysed with lysis buffer (250 mM Tris-HCl, pH 7.5; 0.1% Triton X-100). Soluble lysate protein concentration was determined using the Coomassie protein reagent (Pierce). Luciferase activity was measured on equal amounts of lysate protein using the Luciferase Assay System (Promega, Madison, WI). The CAT assay was performed as described previously (33).
Ligand Binding Assay-Wild type or mutant hVDR proteins, cloned in pSG5, were translated in vitro using the TNT-Coupled Reticulocyte Lysate System (Promega; Madison, WI) and cold methionine. Ten l of the translation products was incubated with increasing concentrations of 1,25-(OH) 2 -23,24[ 3 H]vitamin D 3 (98 Ci/mmol; Amersham Pharmacia Biotech) overnight at 4°C in the presence or absence of unlabeled 1,25-dihydroxyvitamin D 3 (100-fold molar excess). Bound and free ligand were separated with dextran-coated charcoal (Sigma) using the method of Dokoh et al. (34). Scatchard analysis was carried out using the Graphpad Prism program.

RESULTS
Crystallographic structures of RXR and ER have identified surface residues that participate in receptor dimerization. The majority of these residues lie in helix 10 of RXR (18) and helix 11 of ER (19). These, as well as homologous regions from the ligand-binding domains of RAR, TR, and VDR are aligned in Fig. 1 for comparison. We have made mutations in several surface residues in RXR␣ and homologous residues in VDR (identified by shaded boxes) to evaluate their roles in generating dimeric complexes in vitro and regulating transcription in vivo.
By using a GST pull-down assay, a method that identifies FIG. 1. Alignment of amino acid residues from the ligand binding domains of hRXR␣, hRAR␥, hTR␤, hER␣, and hVDR which, based on available structural information, are thought to participate in dimer formation. Shaded boxes identify conserved residues targeted for mutation in hRXR␣ and hVDR in the present study. Single letter nomenclature for the individual amino acids is used. Numbers at right identify the carboxyl-terminal residue in each sequence.
protein-protein interactions in solution, independent of the presence of a DNA recognition element, we examined the ability of wild type and mutant VDRs to form complexes with RXR. As shown in Fig. 2A, GST-VDR showed little propensity to self-associate with either of the VDR mutants. There was, however, a small amount of homodimeric complex formed with wild type VDR, and this association increased modestly with the addition of 1,25-dihydroxyvitamin D 3 . GST-RXR␣ strongly associated with wild type VDR, and again, this interaction was ligand-dependent. There was a reduced level of interaction with the K386A mutant of VDR (ligand-dependent) but virtually no interaction with the I384R/Q385R mutant. In each instance there was a modest increase in the presence of ligand. Of note, both the wild type and the mutant VDR proteins associated with GST-GRIP to an equivalent degree and in a ligand-dependent fashion, indicating that overall structure and function of these mutants were preserved. Thus, both VDR mutants appear to selectively impair heterodimerization with RXR in solution.
Similar findings were obtained when the RXR mutants were examined (Fig. 2B). Wild type RXR bound to GST-VDR in a 1,25-dihydroxyvitamin D 3 -dependent fashion. The R421A and L419R/L420R mutants each displayed impaired capacity to associate with VDR, and in both instances this limited interaction was ligand-dependent. The L419R/L420R RXR mutant also demonstrated poor heterodimerization with TR␤ and peroxisome proliferator-activated receptor ␥ (data not shown).
Since the mutants seemed to disrupt dimerization differentially and because DNA may provide support for heterodimer interactions (35,36), we decided to test the ability of these mutations to disrupt homo-and heterodimerization on DNA using conventional electrophoretic gel mobility shift assays (EMSA). As shown in Fig. 3A, wild type RXR␣ effectively formed ligand-dependent homodimers and hTR␤1-dependent heterodimers on a conventional DR-4 element. Selective muta-tion at position 421 (R421A) in RXR␣ resulted in a loss of homodimeric complexes while, if anything, it increased heterodimer formation. Mutation at positions 419 and 420 (L419R/ L420R), which are exposed on the surface of RXR␣ at or near the dimeric interface (18), completely eliminated homodimer assembly and severely reduced the formation of heterodimers. Of note, these latter reductions were accompanied by the appearance of RXR monomers on the DR-4 template. Next, we studied RXR␣ dimer assembly on a DR-1 template, a template that favors formation of homodimeric complexes of liganded RXR (37). On the DR-1 template wild type RXR␣ assembled as a homodimer in the presence or absence of ligand (Fig. 3B) Reaction mixtures were subjected to 5% nondenaturing PAGE. The gel was amplified, dried, and exposed for autoradiography. B, gel mobility shift assay was performed in the same fashion as in A except that DR-1 was used instead of DR-4. Similar results were obtained from two additional experiments. homodimers on DR-1, as well as the weak TR␤1-RXR␣ heterodimers noted above. Again, as in Fig. 2A, most of the L419R/ L420R mutant assembled as monomers on the DR-1 template.
As expected, wild type VDR formed a stable complex on a DR-3 template with wild type RXR␣ (Fig. 4). The R421A mutant of RXR␣ also formed stable heterodimeric complexes with VDR on this template. On the other hand, the L419R/L420R RXR␣ mutant formed only very low levels of heterodimeric complex with wild type VDR. Mutation of VDR at position 386 (K386A), analogous to the R421A mutation in RXR␣, still bound to wild type RXR␣ and the R421A mutant of that protein (albeit in a somewhat reduced fashion relative to that seen with wild type VDR), and like wild type VDR, it failed to complex with the L419R/L420R mutant of RXR␣. No homodimeric complexes were seen with either wild type VDR or its K386A mutant. Mutation of VDR at positions 384 and 385 (I384R/ Q385R), analogous to the L419R/L420R mutation of RXR␣, displayed modest heterodimer formation only with wild type RXR␣. Neither of the RXR mutants showed appreciable interaction with I384R/Q385R. These complexes were replaced by monomers of the VDR mutant on the DR-3 template.
Each of these mutants was next tested for the capacity to activate a reporter plasmid linking the DR-3 sequence upstream from a thymidine kinase promoter-driven luciferase reporter (DR-3-TKLuc). As shown in Fig. 5, liganded VDR alone effected an ϳ3-fold increment in luciferase activity compared with the untreated control. K386A was somewhat less effective, and I384R/Q385R was devoid of activity. Transfection with wild type RXR␣ alone led to DR-3-TKLuc activities that were not different from control (ϩVD 3 ). Activity declined still further with the R421A and L419R/L420R mutants. Wild type RXR␣ significantly amplified the VDR effect, as did RXR R421A, whereas RXR L419R/L420R effected only a 2-fold increment in reporter activity over that seen with VDR alone. A similar activity profile was seen with the K386A mutant when it was substituted for wild type VDR. The I384R/Q385R VDR mutant, on the other hand, proved incapable of interacting functionally with either the wild type RXR␣ or the RXR␣ mutants.
The ANP gene promoter has been shown previously to be a target for the liganded VDR (16,27,38,39). 1,25-Dihydroxyvitamin D 3 , as well as a number of non-hypercalcemic analogues of vitamin D, effects a VDR-dependent reduction in hANP promoter activity. With this in mind, we examined the ability of the various VDR and RXR␣ mutants to impact on this inhibitory activity. As shown in Fig. 6, liganded VDR effected ϳ50% inhibition in hANP-CAT reporter activity, whereas liganded RXR␣ produced only a 20% reduction, levels that are in agreement with those previously reported (27). The VDR K386A mutant was slightly less effective than wild type VDR in promoting the inhibition, whereas the I384R/Q385R mutant was virtually devoid of activity. Neither of the RXR␣ mutants proved capable of inhibiting hANP promoter activity.
When used in combination, wild type VDR and RXR␣ effected a Ͼ90% inhibition of Ϫ1150 hANP CAT activity (Fig. 7). RXR␣ R421A was less active than wild type in amplifying VDR activity, whereas inhibition in the presence of I384R/Q385R was reduced to ϳ50%, the level seen with wild type VDR alone (see above). In agreement with the observations made with DR-3 TKCAT (see above), the combination of VDR and the RXR mutants resulted in a stepwise loss of ANP promoter inhibition that varied as a function of the "severity" of the mutation. The most significant loss of inhibitory activity was seen with the combination of RXR␣ L419R/L420R and VDR I384R/Q385R. This combination had virtually no effect on the Ϫ1150 hANP CAT reporter. These findings support the hypothesis that residues critical for heterodimerization in these two nuclear receptors are also critical for maintenance of transcriptional regulatory activity (in this case, either stimulatory or inhibitory in nature).
A trivial explanation of these findings arises from the possibility that the mutations, which we assume are selectively targeted to the dimer interface, actually lead to global changes in receptor structure. Since these mutations are positioned in the ligand-binding domain of the receptor, alterations in ligand binding could account for both the loss of heterodimerization and the impairment in functional activity. To address this question, we examined the ligand binding properties of both the wild type and mutant VDRs in a cell-free system. As shown in Fig. 8, affinity of the receptors for [ 3 H]dihydroxyvitamin D 3 was almost identical for each of the three receptors while, if anything, total binding capacity was modestly increased with the mutants. This, together with the observation that each of the VDRs bound equivalently to GRIP-1 ( Fig. 2A), argues against major structural changes as accounting for the loss of functional activity in the mutants and implies that the latter results from selective impairment in the ability of these mutants to form heterodimers.
We employed a double mutation to probe the VDR heterodimerization function to maximize the probability of interfering with the dimer interface. One of the amino acids mutated here (Gln-385) has previously been shown to reduce VDR interaction with an auxiliary factor (presumably RXR) present in COS-7 cells (21). To address the selective role of Ile-384 in the dimerization process, we introduced a site-directed mutation at this position, and we examined the effects of this perturbation on the RXR binding and functional properties of VDR. I384R, like the double mutant (I384R/Q385R), was ineffective in activating DR-3-TK-Luc (Fig. 9A) or inhibiting Ϫ1150 hANP CAT (Fig. 9B) in atrial myocytes. In addition, this mutant displayed a markedly reduced affinity for RXR␣ in the GST pull-down assay (Fig. 9C). Placed in the context of the earlier results of Nakajima et al. (21), it would appear that both Ile-384 and Gln-385 play equivalently important roles in heterodimer assembly. DISCUSSION Recent crystallographic analyses of individual nuclear receptors suggest conservation of structural features involved in dimer assembly. Specifically amino acid residues in helix 10 of hRXR␣ (18) (these residues are located in helix 11 in hTR and hER) and helix 11 of hTR␤ (24) and hER (19) appear to play a FIG. 6. Effect of VDR, RXR␣, or their mutants on ؊1150 hANP-CAT activity. Five g VDR, RXR␣, or the relevant mutants was cotransfected with 20 g of hANP-CAT into neonatal rat atrial myocytes. After 24 h, cells were treated with vehicle, 10 nM 1,25-dihydroxyvitamin D 3 , or 10 nM 9-cis-RA for 48 h. Cells were harvested, and CAT activity was measured. Pooled data, derived from four independent experiments, are presented as means Ϯ S.D.
FIG. 7. VDR-dependent suppression of ؊1150 hANP-CAT activity requires an intact heterodimerization function. Five g of VDR or the relevant mutant was cotransfected with 20 g of hANP-CAT, with or without wild type or mutant RXR␣ (5 g), into neonatal rat atrial myocytes. After 24 h of culture, cells were exposed to vehicle, 10 nM 1,25-dihydroxyvitamin D 3 , or vitamin D plus 10 nM 9-cis-RA for 48 h. Cells were lysed, and CAT activity was measured. Pooled data from 4 to 7 experiments are presented as means Ϯ S.D.

FIG. 8. Ligand binding assays of wild type and mutant VDRs.
Ten l of programmed reticulocyte lysate for VDR or a VDR mutant were incubated with increasing concentrations of 1,25-(OH) 2 -23,24[ 3 H]vitamin D 3 at 4°C for 14 -16 h, as described under "Materials and Methods." Bound and free ligand were separated and quantified. Scatchard analyses of the data are presented in A-C. The same experiment was repeated twice with similar results. key role in establishing contacts required for dimer formation. The present study demonstrates that mutation of two residues (Leu-419 and Leu-420), which structural studies place on the surface of RXR␣ at or near the dimer interface (18), participate in functional heterodimer assembly in the atrial myocyte, as do two homologous residues in VDR (Ile-384 and Gln-385). Contiguously positioned residues in RXR␣ (Arg-421) or VDR (Lys-386) do not inhibit heterodimer formation on DR-3 and consequently have very little impact on receptor-dependent transcriptional regulation.
A number of mutations in helices 10 and 11 have been shown to alter receptor dimerization for different nuclear receptors (21)(22)(23). However, interpretation of these findings is complicated by the fact that, at least in some instances, they have targeted nonsurface residues of the receptor, a manipulation that is likely to disrupt protein folding and/or alter ligand binding. For example, the leucines that border the heptad repeats of TR␣ were previously thought to be involved in dimerization (20), but structural studies have shown that these leucines are engaged in intramolecular interactions and do not participate directly in dimer formation (24). For this reason we confined our mutations to those residues that are known or, based on sequence homology, would be predicted to lie on the surface of RXR␣ and VDR, respectively.
On the DR-4 template, RXR␣ R421A demonstrated impaired formation of RXR␣ homodimers, whereas heterodimer formation with TR␤1 was relatively normal. If anything, the latter was modestly increased, perhaps reflecting diversion of the mutant RXR␣, incapable of assembling with itself, into complexes with a heterodimeric partner. The double mutation (L419R/L420R) further upstream completely eliminated both homo-and heterodimer assembly with RXR␣. Noteworthy, the double mutation still permitted monomer binding to the DR-4 template, implying that the DNA binding function, per se, is not perturbed in this mutant.
We noted no VDR homodimer formation on the DR-3 template in the EMSA and only minimal interaction in the GST pull-down assay. VDR homodimers have been identified in in vitro binding studies by others (14,15), and in the unliganded form they may function as suppressors of gene transcription (15). However, most studies suggest that the functionally relevant complex in transducing the positive vitamin D signal in the target cell is the liganded VDR-RXR heterodimer (39). VDR I384R/Q385R completely disrupted heterodimeric pairings with either wild type RXR␣ or the RXR mutants. Mutations in this region of the hVDR molecule have been reported previously to interfere with heterodimer formation (21). Specifically, VDR mutations K382E, M383G, Q385K, and L390G reduced assembly of a VDR/accessory factor complex on VDRE. Studies with the Q385K mutant (21) support our findings with I384R/ Q385R. Additional studies focusing specifically on Ile-384 (mutation I384R) indicate that this residue, as well, plays an important role in heterodimer formation. Thus, it would appear that both Ile-384 and Gln-385 participate directly in dimer assembly. Lys-386, on the other hand, despite its contiguous location on the receptor surface, does not appear to play a critical role in this process.
The EMSA analyses indicate that both VDR K386A and RXR R421A retain the capacity to interact with heterodimeric partners (RXR␣ in the case of VDR and TR␤1 in the case of RXR␣) at near wild type levels. The conclusion, at least as it applies to the RXR mutant, R421A, stands in contrast to those of Lee et al. (22) who identified this residue as critical for heterodimerization. This difference remains unexplained since both EMSA and the functional analyses indicated that this particular mutant displays close to wild type activity in our system. It should be noted, however, that the GST pull-down assays showed impaired heterodimeric interactions of these two mutants (i.e. VDR K386A and RXR R421A) (see Fig. 3). This discrepancy (GST pull-down versus EMSA) likely reflects differences in the end points being addressed in these two assays. The GST pull-down assay assesses the ability of proteins to associate in solution. Such associations, by definition, have to be of sufficient affinity to preclude disruption during the washing procedure used to reduce "nonspecific" protein-protein interactions. The EMSA is carried out in the presence of DNA template. Positioning of nuclear receptors next to each other on DNA may promote dimer contacts between the DNA binding domains of the subunits (35,36), and receptor-DNA contacts may further stabilize the dimer complex. Thus, the GST pull-down assay is probably a more sensitive method to detect subtle impairment of protein-protein interactions that might otherwise be obscured when the same proteins are bound to DNA. With reference to the current study, although VDR K386A and RXR␣ R421A displayed obvious impairment in their capacity to establish protein-protein interactions in solution, the impairment was not seen when they were permitted to assemble on DNA, and the latter, rather than the former, is probably most reflec- FIG. 9. Impact of Ile to Arg mutation at position 384 (VDR I384R: VDRm3) on the functional and RXR binding properties of VDR. Five g of wild type VDR or VDRm3, alone or together with RXR␣ (or one of the RXR␣ mutants), was cotransfected with DR-3-TKluciferase (10 g) or hANP-CAT (20 g) into neonatal rat atrial myocytes. Extracts were analyzed for luciferase or CAT activity 48 h later. Pooled data from 3 to 4 independent experiments are presented in A and B. C, 35 S-labeled VDR or VDRm3 was incubated with GST-RXR␣ in the presence or absence of 10 nM 1,25-dihydroxyvitamin D 3 for 2 h at 4°C. Bound protein was then analyzed by SDS-PAGE, as described in Fig. 2. The experiment was repeated twice, with comparable results. tive of their functional activity in the intact cell (see Figs. 4 and  7).
Collectively, our data suggest that the surfaces involved in homo-versus heterodimerization overlap (i.e. impairment of both homo-and heterodimerization is seen with the double mutants). They also reveal a critical role for selected residues in dimer assembly. Disruption of the hydrophobic residues in both VDR (Ile-384 and Gln-385) and RXR (Leu-419 and Leu-420) abolishes heterodimerization on DNA, whereas disruption of the charged residue flanking this hydrophobic patch (Lys-386 or Arg-421, respectively) has no effect on this process. Homodimerization appears to be equally affected by any of these mutations suggesting that homodimer assembly is less stable than that of heterodimers, a finding that is in agreement with the fact that heterodimers form preferentially (over homodimers) in solution or on DNA (see Figs. 2-4) (2).
We have shown previously that the liganded VDR exerts anti-hypertrophic activity and suppresses ANP gene transcription in cultured neonatal rat atrial (16,38,39) and ventricular (27) myocytes. This effect was clearly amplified by cotransfection with RXR␣ (16,27); however, a VDR mutant (L262G) with impaired capacity for heterodimer formation (17) was found to retain the ability to suppress the hANP gene promoter (16). This places into question the inferred requirement for VDR heterodimerization in generating the inhibitory effect. Our studies with the I384R/Q385R mutant clearly demonstrate that capacity for heterodimerization closely parallels the ability of the receptor to suppress hANP gene promoter activity. By inference, this would suggest that the L262G mutant described above retains the capacity to interact with a heterodimeric partner, albeit not RXR␣ (17), as a prelude to initiating its biological activity.
In summary, VDR and RXR mutations which, based on RXR and ER structural studies, would be predicted to disrupt protein-protein interactions involved in receptor dimerization do, in fact, demonstrate impairment in dimerization in two independent in vitro assays. This is accompanied by a commensurate reduction in functional activity, assessed through activation of a DR-3-dependent promoter or suppression of an hANPdependent reporter, in transiently transfected rat atrial myocytes. The studies provide support for the suggested conservation of structure-function relationships across different members of the nuclear receptor family and highlight the role of heterodimer formation in vivo as a prerequisite for functional activity of VDR in activating or repressing target gene expression.