Analysis of the Molecular Mechanism for the Antagonistic Action of a Novel 1α,25-Dihydroxyvitamin D3 Analogue toward Vitamin D Receptor Function*

We have recently reported that 23(S)-25-dehydro-1α-hydroxyvitamin D3-26,23-lactone (TEI-9647) efficiently blocks the differentiation of HL-60 cells induced by 1α,25-dihydroxyvitamin D3 (1α,25(OH)2D3) (Miura, D., Manabe, K., Ozono, K., Saito, M., Gao, Q., Norman, A. W., and Ishizuka, S. (1999) J. Biol. Chem. 274, 16392–16399). To clarify the molecular mechanisms of this antagonism, we examined whether TEI-9647 antagonizes the genomic effects of 1α,25(OH)2D3. 10−7 to 10−9 m TEI-9647 inhibited the transactivation effect of 10−8 m1α,25(OH)2D3 in a dose-dependent manner, while TEI-9647 alone did not activate the reporter activity driven by SV40 promoter containing two vitamin D response elements in Saos-2 cells. The antagonistic effect of TEI-9647 was also observed using the rat 24-hydroxylase gene promoter, but the effect was weaker in HeLa and COS-7 cells than in Saos-2 cells. TEI-9647 also exhibited antagonism in an assay system where the VDR fused to the GAL4 DNA-binding domain and the reporter plasmid containing the GAL4 binding site were used in Saos-2 cells, but did not in HeLa cells. TEI-9647 reduced the interaction between VDR and RXRα according to the results obtained from the mammalian two-hybrid system in Saos-2 cells, but did not in HeLa cells. The two-hybrid system also revealed that the interaction between VDR and SRC-1 was reduced by TEI-9647 in Saos-2 cells. These results demonstrate that the novel 1α,25(OH)2D3 analogue, TEI-9647, is the first synthetic ligand for the VDR that efficiently antagonizes the action of 1α,25(OH)2D3, although the extent of its antagonism depends on cell type.

1␣,25-Dihydroxyvitamin D 3 (1␣,25(OH) 2 D 3 ) 1 regulates various biological events including bone and calcium metabolism and cell differentiation (1,2). The vitamin D receptor (VDR), a member of the nuclear hormone receptor superfamily, mediates 1␣,25(OH) 2 D 3 action by modulating the transcription of target genes (1)(2)(3)(4)(5). The effects exerted via the VDR are called genomic effects, in contrast to non-genomic effects, which means that 1␣,25(OH) 2 D 3 acts within a short period without transcription of target genes (6). The VDR-mediated effects of 1␣,25(OH) 2 D 3 in vivo are well known, since there is a human disease associated with the abnormality of the VDR gene, which is called vitamin D dependence type II (7). In addition, a mouse model of the disease was generated by targeting the VDR gene and has been used the analysis of the VDR function in vivo (8,9).
The VDR consists of several functional domains such as DNA-binding and ligand-binding domains (1,2). Heterodimerization of the VDR with retinoid X receptor (RXR) is also important for the binding to a vitamin D-responsive element (VDRE) with high affinity (1,2,5,10). Ligand-dependent activation of transcription requires the extreme C-terminal portion of the VDR, which is designated as the AF-2 core domain (1,11,12). The interaction of these domains is required for the VDR to regulate the transcription of target genes.
A large number of 1␣,25(OH) 2 D 3 analogues have been synthesized in an attempt to dissociate various biological activities including hypercalcemic effects, bone-forming effects, and promotion of cell differentiation (for review, see Ref. 13). Generally, these analogues have been screened in terms of the affinity for the VDR and vitamin D-binding protein. Certain synthetic analogues like 22-oxa-1␣,25(OH) 2 D 3 appear to succeed in the dissection of various effects of 1␣,25(OH) 2 D 3 by taking advantage of the difference in binding to the VDR and vitamin D-binding protein and metabolic pathway (14). To date, however, no anti-vitamin D compounds that oppose the genomic actions of the natural ligand 1␣,25(OH) 2 D 3 , have been reported (13). We have recently synthesized a novel analogue 23(S)-25-dehydro-1␣-hydroxyvitamin D 3 -26,23-lactone (TEI-9647) and found that it efficiently blocks the differentiation of HL-60 cells induced by 1␣,25(OH) 2 D 3 (15). Although this differentiation is believed to be mediated by VDR, more direct evidence is needed to establish whether TEI-9647 is an antagonist of the genomic action of vitamin D. Antagonists, if obtained, would be useful for determining whether each of the effects of 1␣,25(OH) 2 D 3 is mediated by the VDR. In addition, a potentially promising way to achieve the differential effect of vitamin D is to use analogues that antagonize particular effects elicited by vitamin D.
In contrast to the case of VDR, several antagonists have been reported for other members of the nuclear hormone receptor superfamily, and have contributed to the understanding of the functions of nuclear hormones (16 -18). For example, RXRselective antagonist acted differently against retinoic acid receptor and RXRs (19). Although the molecular mechanism of the antagonism is still under investigation, the structural basis of the antagonism was reported; a different conformation in the transactivation domain of the estrogen receptor liganded by 17␤-estradiol and raloxifene was revealed by a crystal structure study (20). Interestingly, some antagonists exhibit agonistic effects in selective tissues. For example, tamoxifen exerts a breast-selective and bone-sparing antagonistic action on estrogen (21). The mechanism of the tissue specificity is unclear. Since the structures of receptors liganded with agonist and antagonist are distinct, the interaction with different coactivators or corepressors in each tissue may play a role in the tissue-selective action of the partial antagonist (22).
In this report, we have characterized the effects of the novel vitamin D analogue TEI-9647 on the various functions of the vitamin D receptor including activation of promoter activity, interaction with RXR and coactivator SRC-1, and binding to VDRE in several types of cells and have detected strong antagonistic activity in cells with an osteoblastic phenotype.
Plasmid Construct and Transcription Activation Assay-The polymerase chain reaction product corresponding to the region (Ϫ367/Ϫ57) regulating expression of the human 24-hydroxylase gene, which contains two VDREs (23), was inserted into a luciferase reporter vector pGVP2 containing SV40 promoter (Toyo Ink Co. Ltd., Tokyo, Japan). The promoter region of the rat 24-hydroxylase gene (Ϫ291/ϩ9), which also contains two VDREs (a gift from Dr. Y. Ohyama, Hiroshima University, Japan) (24), was cloned into a luciferase reporter vector pGVB2 (Toyo Ink). The DNA sequences of these plasmids were confirmed using an ABI 373A DNA sequencer (PE Applied Biosystems, Tokyo, Japan).
Each plasmid together with the hVDR expression vector, pSG5-hVDR (Ref. vector which contains the GAL4 DNA-binding domain (pM) was purchased from CLONTECH. The human VDR cDNA was released by EcoRI from pSG5 vector and fused in-frame to pM (pM-VDR). The cDNA of the VDR lacking its own DNA-binding domain (DBD) was also generated by digestion with NaeI and EcoRI, and cloned into pM vector (pM-VDRdDBD). The luciferase reporter plasmid was generated by insertion of 5 copies of consensus GAL4 binding sites into pGVP2 vector (Toyo Ink) and designated as pGVP2-GAL4BS.
Modified Mammalian Two-hybrid Assay-Whole cDNA of human RXR␣ (a gift from Dr. M. R. Haussler) was released by EcoRI from pSG5 vector and fused in frame to pM vector (pM-hRXR␣). Steroid receptor coactivator-1 (SRC-1) cDNA (Ref. 26; a gift from Dr. M. J. Tsai, Baylor College of Medicine) was released by EcoRI, and the mutation was introduced at the site just before the translation initiation site to be cut by SmaI. Then, the SRC-1 cDNA was digested by SmaI and SalI and inserted in pM vector (pM-SRC-1). Whole cDNA of human VDR was released by EcoRI from pSG5 vector and fused in-frame to pVP16 vector, which contains the activation domain of a herpesvirus (CLON-TECH) (pVP16-hVDR). To examine the interaction of RXR and VDR, 0.5 g of pM-RXR␣, 0.5 g of pSG5-hVDR, and 0.5 g of pGVP2-GAL4BS together with 0.25 g of plasmids containing ␤-galactosidase cDNA were transfected into HeLa cells or Saos-2 cells using Lipo-fectAMINE. Next, to investigate the interaction of SRC-1 and VDR, 0.5 g of pM-SRC-1, 0.5 g of pVP16-hVDR, and 0.5 g of pGVP2-GAL4BS together with 0.25 g of plasmids containing ␤-galactosidase cDNA were transfected into Saos-2 cells also by LipofectAMINE. Twenty four hours later, vehicle, 10 Ϫ8 M 1␣,25(OH) 2 D 3 , 10 Ϫ7 M TEI-9637, or both were added. After 24 h of incubation, the luciferase activity of the cell lysate was examined and adjusted using ␤-gal activity.
In Vitro VDR/VDRE Binding Assay-An electrophoretic mobility shift assay (EMSA) was carried out as previously reported (27), with slight modification. Briefly, vehicle, 10 Ϫ8 M 1␣,25(OH) 2 D 3 , 10 Ϫ7 M TEI-9647, or both were administered to the COS-7 or Saos-2 cells transfected with the hVDR expression vector. Five hours after the addition, cellular extracts were obtained. Five g of each cell extract was incubated with 32 P-labeled synthetic DR3-type VDRE (sense strand: CTAGCAGGTCAAGGAGGTCAG).
Localization of hVDR-Green Fluorescent Protein Fusion Protein-The expression vector of green fluorescent protein (GFP), pGreen Lantern, is driven by the cytomegalovirus promoter (Life Technologies, Inc.). The termination codon (TGA) was mutated to CTT, generating a unique HindIII site. cDNA encoding hVDR was released by digestion with EcoRI from pSG5-hVDR and subcloned into pGreen Lantern after the blunting of both ends. As a result, the expression vector of the fusion protein (GFP N-terminal to hVDR), pGreen Lantern-hVDR was generated. pGreen Lantern-hVDR was introduced into COS-7 cells or Saos-2 cells by the lipofection method. Twenty four hours after the transfection, vehicle, 10 Ϫ8 M 1␣,25(OH) 2 D 3 , 10 Ϫ7 M TEI-9647, or both were administered to the COS-7 or Saos-2 cells and the subcellular distribution of the VDR was observed by fluorescent microscopy (BH-2, Olympus, Tokyo, Japan) 3 h after the addition.
Statistical Analysis-The data were analyzed by analysis of variance using StatView software (SAS Institute Inc., Cary, NC).
The Antagonistic Effect of TEI-9647 Was Stronger in Osteoblastic Cells-10 Ϫ7 M TEI-9647 also exhibited a suppressive effect on the function of 10 Ϫ8 M 1␣,25(OH) 2 D 3 in the assay using the homologous promoter of the rat 24-hydroxylase gene, which contains two VDREs in Saos-2 cells (Fig. 2A). In this sense, the suppressive effect of TEI-9647 did not depend on the promoter context. In MG-63 cells, 10 Ϫ7 M TEI-9647 also suppressed the effect of 10 Ϫ8 M 1␣,25(OH) 2 D 3 to 27% and did not act as an agonist, when no expression vector of VDR was introduced. 10 Ϫ7 M TEI-9647 inhibited the promoter activity induced by 10 Ϫ8 M 1␣,25(OH) 2 D 3 in HeLa cells, although the suppressive effect was weaker in HeLa cells than in Saos-2 and MG-63 cells (Fig. 2B). In COS-7 cells, 10 Ϫ7 M TEI-9647 also suppressed the effect of 10 Ϫ8 M 1␣,25(OH) 2 D 3 , to 54%.

TEI-9647 Exhibited an Antagonistic Effect via the VDR Fused to GAL4 DNA-binding Domain and GAL4
Binding Element-To examine the independence of the antagonistic effect on the interaction between the DNA-binding domain (DBD) of the VDR and a VDRE, GAL4 DBD that was fused to VDR (pM-VDR) or substituted for the DBD of the VDR (pM-VDRd-DBD) was generated and assayed for its activity of gene regulation together with the SV40 promoter-driven reporter plasmid containing the GAL4 binding site and luciferase gene (pGVP2-GAL4BS) (Fig. 3A). 10 Ϫ7 M TEI-9647 inhibited the promoter activity induced by 10 Ϫ8 M 1␣,25(OH) 2 D 3 via the interaction between GAL4-DBD and the GAL4 binding site in Saos-2 cells (Fig. 3B). The results were very similar to those using pM-VDRdDBD (data not shown). In contrast, TEI-9647 did not inhibit the promoter activity induced by 10 Ϫ8 M 1␣,25(OH) 2 D 3 via either pM-VDR or pM-VDRdDBD in HeLa cells (Fig. 3C).

TEI-9647 Inhibited the VDR and RXR Heterodimer Formation Detected by the Modified Mammalian Two-hybrid System
in Saos-2 Cells-As the next step in the activation of the gene transcription by VDR, the protein-protein interaction between the VDR and RXR was examined by the modified mammalian two-hybrid system using the expression vectors of the cDNA of the human RXR␣ fused to GAL4DBD (pM-RXR␣) and the human VDR (pSG5-hVDR). 10 Ϫ8 M 1␣,25(OH) 2 D 3 induced the heterodimer formation between the VDR and RXR␣ as indicated by the reporter activity and 10 Ϫ7 M TEI-9647 significantly inhibited the heterodimer formation in Saos-2 cells (Fig.  4A). In contrast, no suppression against the effect of 10 Ϫ8 M 1␣,25(OH) 2 D 3 was exerted by 10 Ϫ7 M TEI-9647 in HeLa cells (Fig. 4B).
VDR/RXR Binds to a VDRE with TEI-9647-In vitro binding of the VDR/RXR heterodimer to the VDRE was examined by EMSA. The VDR/RXR/VDRE complex was observed in lanes using extracts from cells transfected with the hVDR expression vector (Fig. 5). The addition of 10 Ϫ8 M 1␣,25(OH) 2 D 3 or 10 Ϫ7 M TEI-9647 increased the intensity of the complex, and 10 Ϫ7 M TEI-9647 did not impair this effect of 10 Ϫ8 M 1␣,25(OH) 2 D 3 in COS-7 cells. Essentially the same results were observed in Saos-2 cells, although the enhancing effect of 1␣,25(OH) 2 D 3 was not as high as in COS-7 cells, making it difficult to examine the antagonistic effect of TEI-9647.
Interaction of VDR Liganded by TEI-9647 with Coactivator SRC-1 Is Impaired-As another critical step in the activation of the gene transcription by VDR, the protein-protein interaction between the VDR and SRC-1 was examined by the mammalian two-hybrid system using the expression vectors of the human SRC-1 fused to GAL4DBD (pM-SRC-1) and of the human VDR (pVP16-hVDR). 10 Ϫ8 M 1␣,25(OH) 2 D 3 induced the interaction between the VDR and SRC-1, as indicated by the reporter activity and 10 Ϫ7 M TEI-9647 significantly inhibited the heterodimer formation in Saos-2 cells (Fig. 6).
Localization of hVDR-Green Fluorescent Protein Fusion Protein-The GFP indicated that the VDR localized predominantly in nucleus rather than cytoplasm, and that 10 Ϫ8 M 1␣,25(OH) 2 D 3 shifted the VDR to the nucleus of COS-7 cells (Fig. 7, A and B). TEI-9647 also facilitated the localization of the VDR to nuclei, but did not block the effect of 1␣,25(OH) 2 D 3 (Fig. 7, C and D). The results were the same when pGreen Lantern-hVDR was introduced into Saos-2 cells. The fusion of GFP and the VDR was confirmed by Western blotting using monoclonal antibody 9A7␥ against the VDR and anti-GFP antibody (Roche Molecular Biochemicals; data not shown). DISCUSSION TEI-9647 is the first synthetic ligand for the VDR that efficiently blocks the differentiation of HL-60 cells, human promy- elocytic leukemia cells, induced by 1␣,25(OH) 2 D 3 (15). Although such differentiation of HL-60 cells was speculated to be mediated by the VDR, we have presented direct evidence that TEI-9647 antagonizes the action of the VDR induced by 1␣,25(OH) 2 D 3 , and further studied the mechanism of this antagonistic effect of TEI-9647 in other types of cells.
10 Ϫ7 to 10 Ϫ9 M TEI-9647 inhibited the transactivation effect of 10 Ϫ8 M 1␣,25(OH) 2 D 3 dose-dependently in the reporter system using the VDREs of the human 24-hydroxylase gene in Saos-2 cells. The binding affinity of TEI-9647 for the VDR was one-tenth of that of 1␣,25(OH) 2 D 3 , and 10 Ϫ9 M and 10 Ϫ7 M TEI-9647 significantly inhibited the function of 1␣,25(OH) 2 D 3 by 41% and 74%, respectively. These results suggest that the antagonistic effect of TEI-9647 was more efficient than the simple competitive inhibition of binding to the ligand-binding domain of the VDR, and confirm that TEI-9647 is the first ligand for the VDR to have an antagonistic effect on the VDRmediated action of 1␣,25(OH) 2 D 3 . Although 1␤,25(OH) 2 D 3 was reported to be a potent antagonist of the non-genomic action of 1␣,25(OH) 2 D 3 such as transcaltachia and 45 Ca 2ϩ uptake (28), no antagonists of the genomic action of 1␣,25(OH) 2 D 3 have been reported to date.
1␣,25(OH) 2 D 3 exerts its effect via the VDR through multistep events; first binding to the VDR, then forming a heterodimer with RXR, binding to the VDRE and interacting with a coactivator. To investigate the interaction of the VDR with RXR␣, we used the modified mammalian two-hybrid system with the VDR expression vector, pSG5-VDR, and the plasmid construct expressing GAL4 DBD fused to RXR␣. In our experiments, 10 Ϫ8 M 1␣,25(OH) 2 D 3 enhanced the direct interaction between the VDR and RXR␣ 5-fold as indicated by the reporter activity, and the result was consistent with data reported previously (29,30). A 10 M excess of TEI-9647 markedly inhibited the interaction between the VDR and RXR␣ induced by 1␣,25(OH) 2 D 3 in Saos-2 cells, suggesting that one of the mechanisms for the antagonistic effect of TEI-9647 on the function of 1␣,25(OH) 2 D 3 is reduced interaction between the VDR and RXR. In contrast, the interaction of the VDR with RXR was not affected in HeLa cells. The interaction of VDR and RXR was previously investigated in a yeast two-hybrid system, and a close correlation was found between the ability to activate transcription and the degree of heterodimerization of VDR-RXR liganded with several vitamin D analogues (31). The binding of VDR-RXR complex to a VDRE was detected in EMSA even with the addition of TEI-9647, suggesting that binding of VDR to RXR was not blocked by the metabolite despite the inhibitory effect of TEI-9647 on the enhancement by 1␣,25(OH) 2 D 3 of the interaction between VDR and RXR.
To investigate the interaction between the VDR and coactivators, we used the mammalian two-hybrid system with a VDR expression vector containing the activation domain of a herpesvirus, pVP16-hVDR, and the plasmid construct expressing GAL4 DBD fused to SRC-1, which was originally found as a representative coactivator of the thyroid hormone and retinoic acid receptor superfamily (26). SRC-1 is now recognized as a general coactivator for the nuclear receptor superfamily including VDR. Recently, deletion-mutation analysis revealed that an activation domain of VDR is required for the interaction between VDR and SRC-1 (32). Although the reason is unclear, this ligand-dependent interaction between SRC-1 and VDR was observed in Saos-2 cells, but not in HeLa cells in our system. In Saos-2 cells, the interaction between VDR and SRC-1 was reduced by TEI-9647, suggesting that the conformational change of VDR elicited by the ligand differs between 1␣,25(OH) 2 D 3 and TEI-9647 as suggested by studies using other vitamin D analogues (22,33,34).
The results described above suggest that the multi-step inhibitory action of TEI-9647 on VDR function confers the strong antagonism of this metabolite in Saos-2 cells. Interestingly, TEI-9647 did not have an antagonistic effect in two distinct experiments using pM-VDR or pM-RXR␣ with pSG5-hVDR in HeLa cells (Figs. 3C and 4B). Thus, TEI-9647 appears to be a weaker antagonist in HeLa cells than in Saos-2 cells because of the difference in the number of inhibitory steps. In general, an antagonist exerts its effect in a manner that is either ubiquitous (a pure antagonist) or tissue-specific (a partial antagonist). Representative analogues of estrogen are tamoxifen and raloxifene, tissue-specific antagonists and agonists referred to as SERMs, or selective estrogen receptor modulators (35,36). TEI-9647 showed tissue-specific antagonism in vitro; the effect was marked in Saos-2 cells, MG-63 cells, and HL-60 cells (this paper and Ref. 15), and weak in HeLa and COS-7 cells where it acts as a weak agonist. The difference in the inhibitory activity may explain the difference in the efficiency of the antagonism in various cells. The precise mechanism of partial antagonism remains to be elucidated, but the balance of coactivator and corepressor regulation or some regulating factors such as L7/ SPA may be involved in the tissue-specific activity of the antagonist (37,38). In addition, the antagonism at various stages of the receptor function required for activation of transcription may contribute to the marked inhibitory effect.
The subcellular distribution of VDR in the absence of the ligand remains controversial (39 -41). We generated chimera proteins in which wild-type human VDR was fused to GFP at the C-terminal position and found that the GFP-tagged wildtype VDRs were located predominantly in nuclei with a significant cytoplasmic distribution, while GFP alone was uniformly distributed in both nuclei and cytoplasm (this paper and Ref. 42). Addition of 10 Ϫ8 M 1␣,25(OH) 2 D 3 accelerated the nuclear transfer of VDR in a few hours. The nuclear translocation of VDR was not affected by addition of TEI-9647 in COS-7 and Saos-2 cells. These results are consistent with the finding that TEI-9647 is not a pure antagonist of 1␣,25(OH) 2 D 3 , because estrogen receptor liganded by pure antagonist is reported to be localized in the cytoplasm (43).
In conclusion, we have found that TEI-9647 antagonizes the function of VDR elicited by 1␣,25(OH) 2 D 3 . The extent of the antagonism of TEI-9647 differed between cells. The inhibitory action of TEI-9647 against the interaction between VDR and SRC-1 as well as between VDR and RXR contributes to the antagonistic effect in Saos-2 cells.