Vitamin D receptors repress basal transcription and exert dominant negative activity on triiodothyronine-mediated transcriptional activity.

We have examined vitamin D receptor (VDR), thyroid hormone receptor (TR), and retinoid X receptor beta (RXR beta) binding to vitamin D response elements (VDREs), two thyroid hormone response elements (TREs) (DR4 and F2), and a retinoic acid response element (DR5). VDR/RXR bound well to the VDREs and to DR4 and DR5 using the electrophoretic mobility shift assay. Surprisingly, VDR/RXR also bound well to F2, which contains half-sites arranged as an inverted palindrome. In co-transfection experiments using CV-1 cells, we observed that VDR repressed basal transcription in the absence of ligand on DR3 and osteopontin VDREs and F2, but had no effect on DR4 or DR5. VDR selectively mediated ligand-dependent transcription on only VDREs. VDR also exhibited dominant negative activity as it blocked triiodothyronine (T3)-mediated transcriptional activity on DR4 and F2. These results demonstrate that VDR/RXR heterodimers can bind promiscuously to a wide range of hormone response elements, including inverted palindromes. Moreover, they show that unliganded VDRs, similar to TRs and retinoic acid receptors, can repress basal transcription. Last, they also suggest a novel repressor function of VDR on T3-mediated transcription which may be significant in tissues where VDR and TR are co-expressed.

We have examined vitamin D receptor (VDR), thyroid hormone receptor (TR), and retinoid X receptor ␤ (RXR␤) binding to vitamin D response elements (VDREs), two thyroid hormone response elements (TREs) (DR4 and F2), and a retinoic acid response element (DR5). VDR/RXR bound well to the VDREs and to DR4 and DR5 using the electrophoretic mobility shift assay. Surprisingly, VDR/RXR also bound well to F2, which contains half-sites arranged as an inverted palindrome. In co-transfection experiments using CV-1 cells, we observed that VDR repressed basal transcription in the absence of ligand on DR3 and osteopontin VDREs and F2, but had no effect on DR4 or DR5. VDR selectively mediated ligand-dependent transcription on only VDREs. VDR also exhibited dominant negative activity as it blocked triiodothyronine (T 3 )-mediated transcriptional activity on DR4 and F2. These results demonstrate that VDR/RXR heterodimers can bind promiscuously to a wide range of hormone response elements, including inverted palindromes. Moreover, they show that unliganded VDRs, similar to TRs and retinoic acid receptors, can repress basal transcription. Last, they also suggest a novel repressor function of VDR on T 3mediated transcription which may be significant in tissues where VDR and TR are co-expressed.
Thyroid hormone receptors (TRs) 1 transactivate hormone response elements (HREs) that contain different half-site orientations: palindromes, inverted palindromes, and direct repeats (DRs) (2,3). Studies on the spacing of half-sites arranged as DRs have shown that TRs preferentially mediate ligand-dependent transactivation via a DR with a four-nucleotide gap (DR4), whereas vitamin D receptors (VDRs) and retinoic acid receptors (RARs) transactivate via DRs with three-and fivenucleotide gaps, respectively (DR3 and DR5) (2,4). The transcriptional specificity for these receptors is thought to depend upon optimal heterodimerization by these receptors with retinoid X receptors (RXRs) on these hormone response elements (2,4).
In general, nuclear hormone receptor binding to HREs correlates well with transcriptional activity, although there are examples of promiscuous receptor binding with variable effects on transcription. In particular, TRs can bind to direct repeats and inverted palindromes containing variable spacing (5)(6)(7)(8). TR/RXR heterodimers also can bind to the palindromic vitellogenin estrogen response element, but have weak transcriptional activity via this element (9,10). Additionally, TR and RAR can bind to the palindromic HRE, TREpal, and the rat growth hormone HRE, and both T 3 and retinoic acid can coregulate transcription via these elements (11)(12)(13)(14). It is likely that heterodimerization with RXR facilitates TR binding to these HREs. Although not studied extensively, it is possible that VDR/RXR heterodimerization also may promote similar promiscuous binding by VDR.
Recently, Schrader et al. (1) demonstrated that VDR and TR could form heterodimers on the mouse and rat calbindin HREs which are arranged as DR3 and DR4, respectively. Through cross-linking studies, they demonstrated that VDR/TR dimers have a 5Ј to 3Ј polarity on these elements similar to that described for TR/RXR and RAR/RXR heterodimers. In particular, on the rat calbindin HRE (DR3), VDR bound to the downstream half-site whereas in the mouse calbindin HRE (DR4), TR bound to the downsteam half-site. Moreover, both vitamin D and T 3 could co-regulate transcription via these elements. The demonstration of VDR:TR cross-talk is significant as it suggests that VDR may modulate T 3 -mediated transcription in target genes in cells that express both receptors such as bone, gut, and skin.
In order to understand further potential interactions between TR and VDR, we have studied the DNA binding and transcriptional activity of TR and VDR complexes on different HREs. We report that VDR can bind promiscuously to DR4 and DR5 HREs as well as the chicken lysozyme thyroid hormone response element (TRE), F2 (which contains two half-sites arranged as an inverted palindrome) (15). In co-transfection studies, VDR did not mediate ligand-dependent transcription via reporter plasmid containing these HREs although unliganded VDR could repress basal transcription via DR3, osteopontin VDRE, and F2. Interestingly, VDR also could block T 3 -mediated transcriptional activation on DR4-and F2-containing reporter plasmids. These findings suggest that, similar to TRs and RARs, VDRs can repress basal transcription. Furthermore, VDRs may have an important modulatory role in T 3 -mediated transcription.

Preparation of Vectors and in Vitro
Mutagenesis of VDR-pSG expression vectors encoding chick TR␣ (kind gift of Dr. D. Barettino, Heidelberg University, Germany) and human VDR (kind gift of Dr. W. Hunziker, Hoffman-La Roche, Basel, Switzerland) and human RXR␤ in pCDNA were used in these studies (1,16). An expression vector for GAL4-mouse RXR␤ which encodes the GAL4 DNA-binding domain and the RXR␤ ligand-binding domain from amino acids 209 -448 in pAB as well as GAL4 in pAB also were used in some experiments (17) (kind gift of Dr. Ming-Jer Tsai, Baylor University). Oligonucleotides containing DR4 and F2 (AGCTACTTATTGAGGTCACATGAGGTCAAGTTACG, AGCTTATTGACCCCAGCTGAGGTCAAGTTACG) have been described previously (18,19). DR3 and DR5 (AGCTTACTTATTGAGGT-CACTGAGGTCAAGTTACG, AGCTACTTATTGAGGTCACACTGAG-GTCAAGTTACG) contain nucleotide gaps of 3 and 5 nucleotides, respectively, but are otherwise identical with DR4 (contain the consensus half-site sequence AGGTCA arranged as a direct repeat, flanked upstream and downstream by F2 sequences). An oligonucleotide containing the GAL4 binding site: upstream activating sequence, a previously described 17-base pair sequence (20), also was used. Each of these oligonucleotides as well as the mouse osteopontin HRE (imperfect DR3 with half-site sequence AGTTCA) (21) contained BamHI and EcoRI restriction sites on either end and were subcloned in the reporter vector, PT109 (22) which contains a viral thymidine kinase promoter and the firefly luciferase cDNA, as described previously. Clones were isolated, sequenced, and maxiprepped by affinity chromatography (Qiagen) before used in transfections.
The method for preparation of the VDR DNA-binding mutant, DBD mut , is similar to the method that we have used previously (23,24). Briefly, mutagenesis of human VDR in pSG was performed using 30base pair mutagenic oligonucleotide and restriction site oligonucleotide according to the manufacturer's instructions (Clontech). Final constructs were sequenced to verify mutations. DBD mut contains a single nucleotide change in the codon for the fourth coordinating cysteine (resulting in Cys to Ser change) of the first zinc finger of the DNAbinding domain which has been shown previously to be important for TR binding to TREs (24,25).
Preparation of in Vitro Translated Receptors-cDNA clones of TR, VDR, RXR␤, and DBD mut described above were used in these assays. Unlabeled and [ 35 S]methionine-labeled receptors were produced from rabbit reticulocyte lysates using expression vectors encoding these cDNAs according to the manufacturer's instructions (Promega). Unprogrammed reticulocyte lysate also was incubated under the same conditions. [ 35 S]Methionine-labeled receptor protein was quantitated by SDS-polyacrylamide gel electrophoresis analysis which showed labeled proteins of expected molecular weights.
In vitro translated receptor and 10,000 cpm of oligonucleotide probe were mixed and incubated together before being subjected to electrophoresis and autoradiography as described previously (26).
Co-transfection Studies-cDNA clones of TR␣, VDR, or DBD mut described above were used in the co-transfection experiments. Reporter plasmids containing the F2, DR3, DR4, DR5, or osteopontin HREs and the luciferase cDNA in PT109 described above were used.
CV-1 cells were grown in Dulbecco's modified Eagle's medium, 10% fetal calf serum. The serum was stripped of T 3 and vitamin D by incubating with charcoal for 12 h at 4°C, and constant mixing with 5% (w/v) AG1-X8 resin (Bio-Rad) twice for 12 h at 4°C before ultrafiltration. The cells were transfected with expression (0.1 g) and reporter (2 g) plasmids as well as a Rous sarcoma virus-␤-galactosidase control plasmid (1 g) (27) in 3.5-cm plates using the calcium-phosphate precipitation method (28). Cells were grown for 48 h in the absence or presence of 10 Ϫ6 M T 3 (Sigma) or vitamin D (Biomol), and harvested. Cell extracts then were analyzed for both luciferase (29) and ␤-galactosidase (27) activity in order to correct for transfection efficiency. Except where indicated, the corrected luciferase activities of untreated samples were normalized to the luciferase activities of samples containing vector alone in the absence of ligand (1-fold basal).
VDR Expression in CV-1 Cells-VDR and DBD mut expressions in CV-1 cells were determined using previously described protocols with minor changes (30). Briefly, CV-1 cells in 10-cm dishes were co-transfected with 10 g of expression plasmid and 1 g of ␤-galactosidase cDNA by the calcium phosphate method and grown in stripped media. After 48 h, cells were treated with 0.5 nM [ 3 H]vitamin D (DuPont NEN; specific activity ϭ 175 Ci/mmol) Ϯ 1000-fold excess vitamin D for 2 h. Cells were harvested by scraping in 2 ml of cold phosphate-buffered saline. Nuclear pellets and binding assay then were performed as described previously (30). Total incorporated specific counts per minute (cpm) were normalized to ␤-galactosidase activity to adjust for any differences in transfection efficiency of samples. RESULTS We first examined VDR homodimer and heterodimer binding on DR3, DR4, and DR5 by electrophoretic mobility shift assay. VDR/RXR heterodimer bound well to labeled DR3, DR4, and DR5 oligonucleotides (Fig. 1, lanes 5, 11, and 17). Two VDR/ RXR bands were observed as in vitro translated RXR␤ has two major translation products. The rank order of binding was DR3 Ͼ DR4 Ͼ DR5. Interestingly, addition of vitamin D slightly decreased VDR/RXR heterodimer binding to these elements. No VDR homodimer or monomer binding was observed on DR3 or DR5; however, a faint band was observed on DR4 with the sample containing VDR alone (Fig. 1, lane 9). This band disappeared after addition of vitamin D (Fig. 1, lane 10), similar to recent reports of ligand-induced dissociation of TR and VDR homodimers (26,(31)(32)(33)(34). Additionally, anti-VDR antibody could block the formation of this band (data not shown). These results suggest that VDR weakly formed a homodimer on the DR4 oligonucleotide. In contrast, no VDR/TR dimers were observed on these elements ( Fig. 3 and data not shown).
We next examined the transcriptional activity of VDR on these elements by transfecting CV-1 cells with VDR expression plasmid and reporter plasmids containing the DRs (Fig. 2). In the absence of vitamin D, we observed repression of basal transcription on DR3-and the mouse osteopontin VDRE-containing reporter (this osteopontin VDRE contains an imperfect DR3). Little or no basal repression was observed on DR4 and DR5. In the presence of vitamin D, transcriptional activation was observed on DR3 and osteopontin VDRE, but not on DR4 and DR5. These findings suggest that VDR can repress basal repression in the absence of ligand on DR3 and osteopontin VDRE similar to TR on certain TREs (6,35,36). Additionally, despite binding as VDR/RXR heterodimers to DR4 and DR5, VDR had minimal or no effect on basal repression or liganddependent transactivation.
GAL4 fusion protein systems have been used to characterize roles of nuclear hormone receptor heterodimers on transcriptional activation by minimizing potential contributions of endogenous RXR on transcription (17,37). In order to further characterize basal repression by VDR, we performed co-transfection studies using GAL4, GAL4-RXR, and VDR expression plasmids and a GAL4 binding site-containing reporter in CV-1 cells. Both VDR and GAL4, or GAL4-RXR alone, slightly decreased transcription both in the presence or absence of vita- min D when compared to GAL4 alone. However, when cells were co-transfected with VDR and GAL4-RXR in the absence of vitamin D, there was greater than 90% repression of basal transcription compared to GAL4 alone (fold basal luciferase activity ϭ 0.07 Ϯ 0.01 S.D.) and 90% repression of basal transcription compared to GAL4-RXR alone (fold basal luciferase activity ϭ 0.10 Ϯ 0.01 S.D.). This repression was reversed with addition of vitamin D, as ligand stimulated transcription greater than 3-fold above basal levels (fold basal luciferase activity ϭ 3.45 Ϯ 0.36 S.D.). These data strongly support an important functional role for VDR/RXR heterodimer in mediating basal repression and transcriptional activation.
We then compared VDR and TR␣ binding to the consensus TRE, DR4, and the chicken lysozyme TRE, F2, which contains half-sites arranged as an inverted palindrome. On DR4, TR␣ alone bound mostly as a monomer (Fig. 3A, lane 3) whereas both TR␣ and VDR bound poorly as a homodimer ( Fig. 1 and Fig. 3A, lanes 3 and 4). Both TR/RXR and VDR/RXR heterodimers bound well to DR4, but no VDR/TR dimers were observed even after a 48-h exposure of the film to the gel. Surprisingly, when we examined TR and VDR binding to the inverted palindrome, F2, we also observed strong VDR/RXR heterodimer binding comparable to that observed for TR/RXR (Fig. 3B, lanes 5 and 6). As observed for the DRs, vitamin D slightly decreased VDR/RXR heterodimer binding to F2 (Fig.  3B, lanes 6 and 9). Anti-RXR␣ antibody partially supershifted the VDR/RXR bands, whereas anti-VDR antibody blocked DNA binding by these complexes (Fig. 3B, lanes 11 and 12). No VDR homodimer or VDR/TR dimers were observed even after a 48-h exposure of the gel (Fig. 3B, lanes 4 and 7). These latter findings suggest that VDR not only can bind promiscuously to DRs with variable spacing between half-sites, but also can bind to HREs containing different half-site orientations, e.g. inverted palindromes.
Studies comparing the transcriptional activities on TR␣ and VDR using the DR4-and F2-containing reporters also were performed. On DR4, basal repression by unliganded TR␣ and T 3 -mediated transcriptional activation were observed (Fig. 4A) as noted previously (18). VDR appeared to have weak repression of basal transcription on DR4, although this effect did not reach statistical significance (n ϭ 5 experiments and Figs. 1 and 4A). VDR also did not stimulate transcription on DR4 in the presence of vitamin D. On F2, similar basal repression and ligand-mediated transcriptional activation for TR␣ also was observed. Interestingly, VDR repressed basal transcription in the absence of vitamin D and derepressed basal transcription In these experiments, treated cells then were harvested and luciferase was measured. Luciferase activity was normalized to ␤-galactosidase activity and then calculated as fold basal luciferase activity with 1-fold basal activity defined as the luciferase activity with control pSG vector alone in the absence of ligand. Each point represents the mean of four samples, and bars denote S.E. * denotes significant difference from basal luciferase activity.

FIG. 3. Formation of VDR/RXR heterodimers on DR4 and the inverted palindrome F2 TRE.
In vitro translated TR␣, VDR, and RXR␤ (2-5 l) were incubated with 32 P-labeled DR4 or F2 and then analyzed by EMSA as in Fig. 1. Reticulocyte lysate was added to some samples so that the total volume of reticulocyte lysate was the same for each sample. Similar amounts of TR␣ and VDR, as quantitated by SDS-polyacrylamide gel electrophoresis analyses described under "Materials and Methods," were added. A, DR4. Note that weak TR␣ and VDR homodimer bands were seen (lanes 3 and 4) after longer film exposure (48 h) to gel. B, F2. Lanes 10 -12, preimmune, anti-RXR␤ or anti-VDR antibodies were added after samples incubated with probe. ␣M, TR␣ monomer; ␣D, TR␣ homodimer; HD, TR/RXR or VDR/RXR heterodimer; and SS, supershifted complex.
in the presence of vitamin D on F2 (Fig. 4B). However, there was no significant ligand-dependent transactivation by VDR on F2-reporter activity.
Since VDR could bind to the DR4 and F2 TREs but was unable to mediate ligand-dependent transcriptional activation, we examined whether VDR might instead exert dominant negative activity on T 3 -mediated transcription. Previously, it had been shown the natural mutant TR␤s from patients with resistance to thyroid hormone as well as the viral oncogene homolog of TR, v-erbA, possess dominant negative activity on wild-type TRs. Moreover, this dominant negative activity depended on mutant TR or v-erbA binding to TREs. Accordingly, we transfected increasing amounts of VDR expression plasmid and a fixed amount of TR␣ expression plasmid with DR4 reporter into CV-1 cells (Fig. 5A). In the absence or presence of vitamin D, increasing VDR:TR expression plasmid ratio pro-gressively decreased T 3 -mediated transcriptional activation. In fact, a 3:1 ratio of VDR and TR␣ expression plasmids completely blocked T 3 -mediated transcriptional activation to basal levels. Similar findings were observed for VDR blockade of T 3 -mediated transcription on the F2 reporter (Fig. 5B). We also observed similar VDR blockade of T 3 -mediated transcription when TR␤ expression plasmid was used, indicating VDR did not exhibit TR isoform specificity in its dominant negative activity (data not shown).
In order to investigate whether the dominant negative activity depended on VDR binding to TREs, we created a mutant VDR, DBD mut , in which the fourth cysteine of the first zinc finger was mutated to a serine. We and others have shown that similar mutations in TR abrogate or markedly reduce TR homodimer and TR/RXR heterodimer binding to TREs without affecting ligand binding (24,25). DBD mut was unable to bind as FIG. 5. VDR blocks T 3 -mediated transcriptional activity on DR4 and F2 reporter. Expression plasmids encoding TR␣, VDR, or pSG control vector (0.1 g) were co-transfected with DR4-or F2-containing reporter plasmid (1.7 g) and Rous sarcoma virus-␤-galactosidase control vector (1 g). Cells then were treated with or without 10 Ϫ6 M T 3 Ϯ vitamin D for 48 h and analyzed for luciferase activity. Luciferase activity was normalized to ␤-galactosidase activity and then calculated as fold basal luciferase activity with 1-fold basal activity defined as the luciferase activity with pSG alone in the presence of T 3 alone or T 3 ϩ vitamin D. Each point represents the mean of four samples, and bars denote S.D. A, DR4-containing reporter; B, F2-containing reporter. a heterodimer to F2, DR3, and DR4 ( Fig. 6A and data not shown). We then examined the dominant negative activity of DBD mut on T 3 -mediated transcriptional activity on F2-containing reporter (Fig. 6B). At a 1:1 expression plasmid ratio, DBD mut had little dominant negative activity in contrast to wild type VDR. At increasing expression plasmid ratios, however, DBD mut had some dominant negative activity which was less than wild type VDR dominant negative activity at the same expression plasmid ratio. Addition of vitamin D did not significantly affect the dominant activity of DBD mut or wild type VDR, and similar results were observed with the DR4-containing reporter (Fig. 5, A and B, and data not shown). Additionally, we performed [ 3 H]vitamin D binding studies on nuclear extracts from CV-1 cells co-transfected with VDR or DBD mut and observed similar binding, suggesting similar expression of these receptors in CV-1 cells (data not shown). Collectively, these results suggest that DNA binding is not absolutely required for dominant negative activity by VDR, particularly at high expression plasmid ratios. However, VDR/ RXR heterodimer binding to TREs may be a major contributor to dominant negative activity at low expression plasmid ratios (Figs. 6 and 7). DISCUSSION These studies provide new information on transcriptional mechanisms mediated by VDR and TR and suggest cross-talk between these receptors. In our studies of transcriptional activity by VDR on different reporter plasmids, we observed that unliganded VDR can repress basal transcription on DR3, osteopontin VDRE, and F2. This phenomenon has not been reported for VDR previously. However, basal repression or silencing has been well described for TRs (6,35,36). One mechanism proposed for basal repression involves the binding of unliganded TR homodimer and TR/RXR heterodimer to TREs, possibly in conjunction with co-repressor(s) to TREs. These complexes may then interact with the basal transcriptional machinery to repress basal transcription (24,38). In the presence of T 3 , the TR homodimer dissociates from the TRE (inducing derepression), whereas the liganded TR/RXR assumes an "active" conformation such that it now interacts with co-activator(s) or the basal transcriptional machinery to support ligand-dependent transcriptional activation. It is likely that VDR may utilize similar repression and derepression mechanisms as liganded VDR homodimer dissociates from HREs whereas liganded VDR/RXR heterodimers remain bound to HREs (34). However, VDR formed homodimers weakly on DR4 and poorly on all other elements studied (including DR3 and osteopontin VDRE) suggesting that unliganded VDR/RXR may be the major complex in VDR-mediated repression of basal transcription. Our studies with GAL4-RXR and VDR further support the notion that VDR/RXR heterodimers can mediate basal repression.
Recently, two groups also have shown that TRs and RARs can interact with co-repressors that may participate in basal repression of transcription (39,40). This new class of proteins are called thyroid hormone and retinoic acid receptor associated co-repressors (TRACs). Interestingly, these proteins bind well to TRs or RARs in the absence, but not in the presence, of ligand. The co-repressor interacts with the hinge region of TR and RAR, and a three-amino acid substitution of amino acids 223, 224, and 227 of TR␤ abrogates binding with co-repressor (40). Interestingly, these amino acids are conserved in the vitamin D receptor. It remains to be seen whether VDR interacts with these recently described co-repressors or its own distinct co-repressor.
Our findings also demonstrated that VDR/RXR heterodimers can bind to DRs with gaps greater than three nucleotides. Although these heterodimers can bind to DR4 and DR5, reporter plasmids containing these HREs exhibit little vitamin D-dependent transcriptional activity. This discordance between DNA binding and transcriptional activity likely is due to spacing between VDRE half-sites modulating VDR/RXR heterodimer conformation, which, in turn, may affect its interaction with other associated proteins involved in transcriptional activation. In this connection, recent studies have shown similar discordance between TR/RXR heterodimer binding to certain elements and transcriptional activity (21,41,42). Moreover, binding to different TREs can affect the trypsin sensitivity of TR/RXR heterodimers (42). Likewise, het-FIG. 6. DNA binding and dominant negative activity by DNAbinding mutant, DBD mut , on F2. A, in vitro translated VDR, RXR, and DBD mut in the presence and absence of vitamin D were incubated with 32 P-labeled F2 probe, similar to Figs. 1 and 3. Samples then were analyzed by EMSA. B, increasing amounts of VDR or DBD mut expression vector (0.1, 0.3, 0.5 g) and a fixed amount of TR␣ expression vector (0.1 g) were co-transfected with F2-containing reporter plasmid (1.7 g) and ␤-galactosidase control vector (1 g) in the presence of 10 Ϫ6 M T 3 for 48 h and analyzed for luciferase activity. In some samples, control pSG vector was added so that the total amount of expression vector was identical for each sample. Luciferase activity was normalized to ␤-galactosidase activity and then calculated as fold basal luciferase activity with 1-fold basal activity defined as the luciferase activity with control pSG vector alone in the absence of ligand. Each point represents the mean of triplicate samples. Similar results were obtained with DR4-containing reporter. erodimer binding was not sufficient for mediating basal repression as we observed basal repression on DR3, osteopontin, and F2 HREs, but not on DR4 or DR5. This selective ability of unliganded VDR to repress and derepress basal transcription on certain HREs is reminiscent of recent studies by Kurokawa et al. (43) in which they observed similar effects for RAR on DR1 and DR5. Thus, DNA sequences as well as spacing between half-sites may modulate receptor complex interactions with associated proteins for both transcriptional activation and basal repression.
We also observed VDR/RXR heterodimer binding on the chicken lysozyme TRE, F2, suggesting that VDRs can bind to HREs in which half-sites are arranged as inverted palindromes. Again, as previously observed for TR/RXR heterodimers (6 -8), the formation of VDR/RXR heterodimers allows a more flexible and permissive binding to HREs containing different half-site spacing and orientations. Recently, Schrader et al. (44) have reported VDR/RXR heterodimer binding on two other inverted palindromes with a 9-nucleotide spacing, human calbindin and rat osteocalcin HREs. Both vitamin D and 9-cis-retinoic acid can stimulate transcription via these elements. However, in the case of F2, which contains an inverted palindrome with a gap of 6 nucleotides, no vitamin D-stimulated transcription was observed. These results again suggest that spacing between half-sites may dictate transcriptional activity by VDR/RXR heterodimers.
Previously, Schrader et al. (1) reported VDR/TR heterodimer binding to calbindin HREs. We were unable to observe formation of VDR/TR heterodimers on any of the different HREs as well as on the mouse and rat calbindin HREs. 2 Although differences in EMSA conditions likely account for these differences, these data nonetheless suggest that RXR preferentially forms heterodimers with VDR in comparison to TR on most HREs. Schrader et al. (1) also observed vitamin D, T 3 , and dual ligand activation via the calbindin HREs. In contrast, we have observed that VDR had no ligand-dependent transcriptional activation on two TREs (DR4 and F2) and a retinoic acid response element (DR5). Moreover, when VDR was co-transfected with TR, we observed dominant negative activity on T 3 -mediated transcription. Similar results also were observed for VDR on RA-mediated transcription using the DR5 reporter. 2 These findings suggest that VDR modulation of transcriptional activity on these HREs are fundamentally different than those reported on the calbindin HREs (1), both in terms of mechanism and the functional consequence of blocking hormone-mediated transcriptional activation.
It previously had been shown that mutation of the DNAbinding domain abrogated dominant negative activity by natural mutant TRs and v-erbA (25,45) suggesting that DNA binding is essential for dominant negative activity on T 3 -mediated transcription by these proteins. Our studies with DBD mut , however, indicate that, although DNA binding is important for full potency of dominant negative activity by VDR, it is not absolutely required as high concentrations of DBD mut also can have dominant negative activity. These findings suggest that both competition for DNA binding to TREs by transcriptionally inactive VDR/RXR heterodimers as well as titration of putative common associated proteins or co-activators may contribute to dominant negative activity by VDR (Fig. 7). It does not appear that titration of RXR is the likely mechanism for the latter effect as co-transfection of RXR␤, the major endogenous T 3receptor auxiliary protein in CV-1 cells (46), does not reverse VDR dominant negative activity on T 3 -mediated transcrip-tion. 2 Moreover, in contrast to TR/RXR heterodimers, VDR/TR heterodimers do not form in solution (47). 3 Interestingly, another dominant negative inhibitor of T 3 action, c-erbA␣-2 (the alternative splice variant of TR␣), may have similar mechanisms of action (48 -50). Several groups have shown DNA binding and/or DNA-dependent blockade of T 3 -mediated transcription by c-erbA␣-2 (48 -50, 52). However, DNA binding is not absolutely required as co-transfection of high amounts of a DNA-binding mutant of c-erbA␣-2 still can mediate dominant negative activity (51). Recently, Juge-Aubry et al. (53) showed that peroxisome proliferator-activated receptor, a member of the nuclear hormone receptor superfamily, also can block T 3 -mediated transcription. These studies with peroxisome proliferator-activated receptor suggest that heterodimerization with RXR as well as titration of a common co-activator may be important in mediating dominant negative activity.
Several human VDR mutations have been described which contain either mutations in the DNA-binding domain or premature stop codons (54,55). Some of these mutants have been shown to bind DNA poorly or are likely to dimerize poorly since the critical distal heptad repeats in the ligand-binding domain which are important for TR and VDR dimerization are eliminated (56). It is possible that these mutants may behave similarly to the DBD mut ; that is, they may exhibit impaired dominant negative activity on T 3 -mediated transcription. It thus is possible that mutations in certain nuclear hormone receptors may have a deleterious effect on the function(s) of other nuclear hormone receptors, which, in turn, may contribute to the phenotype of affected patients.
Currently, little is known about vitamin D regulation of TRs and vice versa. Previously, Kaji  yet another mechanism for VDR:TR cross talk, formation of VDR/RXR heterodimers on TREs that can modulate T 3 -mediated transcription. These findings raise the possibility that VDR can modulate T 3 -mediated transcription in tissues where both receptors are co-expressed (e.g. bone, gut, skin). Although the physiological significance of this cross-talk currently is not known, studies on the thyroid hormone status and function in patients with vitamin D resistance or in VDR knockout animals may shed light on this issue. In summary, our studies of VDR and TR regulation of transcription demonstrate the intricacy and complexity of the signaling pathways and networks that may occur among nuclear hormone receptors. These, in turn, may enable finely tuned regulation of gene transcription in target tissues.