Antagonistic action of a 25-carboxylic ester analogue of 1alpha, 25-dihydroxyvitamin D3 is mediated by a lack of ligand-induced vitamin D receptor interaction with coactivators.

A 25-carboxylic ester analogue of 1alpha,25-dihydroxyvitamin D(3) (1alpha,25-(OH)(2)D(3)), ZK159222, was described as a novel type of antagonist of 1alpha,25-(OH)(2)D(3) signaling. The ligand sensitivity of ZK159222, in facilitating complex formation between 1alpha,25-(OH)(2)D(3) receptor (VDR) and the retinoid X receptor (RXR) on a 1alpha,25-(OH)(2)D(3) response element (VDRE), was approximately 7-fold lower when compared with 1alpha,25-(OH)(2)D(3). However, ZK159222 was not able to promote a ligand-dependent interaction of the VDR with the coactivator proteins SRC-1, TIF2, and RAC3, neither in solution nor in a complex with RXR on DNA. Functional analysis in HeLa and COS-7 cells demonstrated a 10-100-fold lower ligand sensitivity for ZK159222 than for 1alpha, 25-(OH)(2)D(3) and, most interestingly, a potency that was drastically reduced compared with 1alpha,25-(OH)(2)D(3). A cotreatment of 1alpha,25-(OH)(2)D(3) with a 100-fold higher concentration of ZK159222 resulted in a prominent antagonistic effect both in functional in vivo and in in vitro assays. These data suggest that the antagonistic action of ZK159222 is due to a lack of ligand-induced interaction of the VDR with coactivators with a parallel ligand sensitivity, which is sufficient for competition with the natural hormone for VDR binding.

The main physiological role of 1␣,25-dihydroxyvitamin D 3 (1␣,25-(OH) 2 D 3 ) 1 -mediated gene transcription is the regulation of calcium homeostasis and bone mineralization (1), but the nuclear hormone also plays a role in controlling cellular growth, differentiation, and apoptosis (2). Various analogues of 1␣,25-(OH) 2 D 3 , which mainly contain modifications of the side chain, have been developed with the goal to improve the biological profile of the natural hormone for a potential ther-apeutic application (3). The genomic effects of 1␣,25-(OH) 2 D 3 and its analogues are principally mediated through the 1␣,25-(OH) 2 D 3 receptor (VDR) (4), which is a member of the nuclear receptor transcription factor superfamily (5). VDR binds as a heterodimer with the retinoid X receptor (RXR) (6) to specific sequences in promoter regions of 1␣,25-(OH) 2 D 3 target genes, referred to as 1␣,25-(OH) 2 D 3 response elements (VDREs) (7). Simple VDREs consist of two hexameric nuclear receptor binding sites, which are commonly arranged as a direct repeat with 3 spacing nucleotides (DR3-type VDREs) (4). The VDR consists of several functional domains, which includes the DNA-binding domain (DBD) and the ligandbinding domain (LBD). The LBD of the VDR is formed by 12 ␣-helical structures, of which the last one, helix 12, contains a short transactivation function 2 (AF-2) domain. A critical step in 1␣,25-(OH) 2 D 3 signaling is the specific ligand-triggered induction of a conformational change within the LBD of the VDR. This conformational change induces the dissociation of corepressors, such as NCoR (8) or Alien (9), and facilitates the interaction with coactivator proteins with members of the p160 family, such as SRC-1/ERAP160/NCoA1 (10,11), TIF2/Grip-1/NCoA2 (12,13), and RAC3/AIB1/ACTR/ pCIP (14 -16). This VDR-coactivator interaction then further facilitates recruitment of other factors to form a larger complex that modulates chromatin structure and initiates transcription (17). This also involves the recently described DRIP/ ARC cofactor complexes (18,19), which appear to contact the VDR and other nuclear receptors preceding their interaction with p160 family of coactivators (20).
For some members of the nuclear hormone receptor superfamily, such as the estrogen receptor (ER) and the progesterone receptor, antagonists in addition to agonists have been known for some time (21). Recently, 25-carboxylic ester and 26,23lactone 1␣,25-(OH) 2 D 3 analogues have been described as the first types of VDR antagonists (22,23). The molecular mechanisms of action of ER antagonists have been explained by incorrect positioning and blocking of the AF-2 domain (24,25). In contrast, the molecular mechanisms of the antagonistic action of the 26,23-lactone analogue have been explained by reduced VDR-RXR heterodimer complex formation (26).
In this study, a 25-carboxylic ester analogue of 1␣,25-(OH) 2 D 3 , ZK159222, was characterized as a novel type of 1␣,25-(OH) 2 D 3 antagonist. ZK159222 was not able to promote a ligand-dependent interaction of the VDR with coactivator proteins of the p160 family neither in solution nor in a complex with RXR on DNA. Moreover, functional in vivo assays in HeLa, COS-7, and MCF-7 cells demonstrated that a cotreatment of 1␣,25-(OH) 2 D 3 with a 100-fold higher concentration of ZK159222 showed a prominent antagonistic effect, which was confirmed with in vitro coactivator interaction assays.

DNA Constructs
Mammalian Expression Constructs-The full-length cDNAs for human VDR (27) and human RXR␣ (28) were subcloned into the SV40 promoter-driven pSG5 expression vector (Stratagene, Heidelberg, Germany). These constructs are also suitable for T 7 RNA polymerasedriven in vitro transcription/translation of the respective cDNAs.
VDRE-driven Reporter Gene Construct-Four copies of the DR3-type VDRE from the rat atrial natriuretic factor (ANF) gene promoter (29) were fused with the thymidine kinase (tk) minimal promoter driving the luciferase reporter gene.
GAL4 Fusion Construct-The DBD of the yeast transcription factor GAL4 (amino acids 1-147) was fused with the cDNA of the human VDR LBD (amino acids 109 -427). For the mammalian one-hybrid assays the luciferase reporter gene was driven by three copies of the GAL4 binding site fused to the tk promoter (8).

Limited Protease Digestion Assay
In vitro translated, 35 S-labeled VDR protein (2.5 l) alone or together with 2.5 l in vitro translated RXR and 1 ng of unlabeled rat ANF DR3-type VDRE were incubated with ligand for 15 min at room temperature in 20 l of binding buffer (10 mM Hepes (pH 7.9), 1 mM dithiothreitol, 0.2 g/l poly(dI-C), and 5% glycerol). The buffer was adjusted to 150 mM of monovalent cations by addition of KCl. Trypsin (Promega, final concentration 8.3 ng/l) was then added, and the mixtures were further incubated for 15 min at room temperature. The digestion reactions were stopped by adding 25 l of protein gel loading buffer (0.25 M Tris (pH 6.8), 20% glycerol, 5% mercaptoethanol, 2% SDS, 0.025% bromphenol blue). The samples were denatured at 85°C for 3 min and electrophoresed through a 15% SDS-polyacrylamide gel. The gels were dried and exposed to a Fuji MP2040S imager screen. The individual protease-sensitive VDR fragments were quantified on a Fuji FLA2000 reader (Tokyo, Japan) using Image Gauge software (Raytest, Sprockhövel, Germany).

Gel Shift and Supershift Assay
In vitro translated VDR-RXR heterodimers were incubated with graded or saturating concentrations of 1␣,25-(OH) 2 D 3 and ZK159222 for 15 min at room temperature in a total volume of 20 l of binding buffer. The buffer had been adjusted to 150 mM by addition of KCl. For supershift assays, approximately 3 g of bacterially expressed GST-SRC-1 596 -790 , GST-TIF2 646 -926 , or GST-RAC3 673-1106 fusion protein were included in the incubation. Approximately 1 ng of the 32 P-labeled DR3-type VDRE from the rat ANF promoter (50,000 cpm) was added to the protein-ligand mixture, and incubation was continued for 20 min. Protein-DNA complexes were resolved through 6 or 8% nondenaturing polyacrylamide gels in 0.5 ϫ TBE (45 mM Tris, 45 mM boric acid, 1 mM EDTA (pH 8.3)) and were quantified by phosphorimaging.

GST Pull-down Assays
GST pull-down assays were performed by coincubation of a 50% GST-TIF2 646 -926 -Sepharose bead slurry with in vitro translated 35 Slabeled VDR and indicated concentrations of 1␣,25-(OH) 2 D 3 and ZK159222 in PPI buffer (20 mM Hepes (pH 7.9), 200 mM KCl, 1 mM EDTA, 4 mM MgCl 2 , 1 mM dithiothreitol, 0.1% Nonidet P-40, and 10% glycerol) for 20 min at 30°C. GST fusion protein-Sepharose slurries were routinely preblocked in PPI buffer containing bovine serum albumin (1 g/l) prior to use in pull-down assays. In vitro translated proteins, that were not bound to GST fusion proteins, were washed away with PPI buffer. GST fusion protein bound VDR was detected by electrophoresis through 10% SDS-polyacrylamide gels and quantified by phosphorimaging.

Transfection and Luciferase Assays
COS-7 SV40-transformed African green monkey kidney cells or MCF-7 human breast cancer cells were seeded into six-well plates (10 5 cells/ml) and grown overnight in phenol red-free Dulbecco's modified Eagle's medium supplemented with 10% charcoal-treated fetal bovine serum. Liposomes were formed by incubating 1 g of the reporter plasmid, 1 g of each pSG5-based receptor expression vectors for VDR and RXR, and 1 g of the reference plasmid pCH110 (Amersham Pharmacia Biotech) with 15 g N- [1-(2,3-dioleoyloxy)propyl]-N,N,Ntrimethylammoniummethylsulfate (DOTAP, Roth, Karlsruhe, Germany) for 15 min at room temperature in a total volume of 100 l. After dilution with 900 l of phenol red-free Dulbecco's modified Eagle's medium, the liposomes were added to the cells. Phenol red-free Dulbecco's modified Eagle's medium supplemented with 30% charcoal-treated fetal bovine serum (500 l) was added 4 h after transfection. At this time, graded concentrations of 1␣,25-(OH) 2 D 3 or ZK159222 were also added. HeLa human cervix carcinoma cells were cultured, seeded, and transfected under the same conditions as COS-7 and MCF-7 cells, but for the mammalian one-hybrid assay, the expression vector for the GAL4 DBD VDR LBD fusion protein and a GAL4 binding site-driven luciferase reporter gene construct were used in transfections. The cells were lysed 16 h after onset of stimulation using the reporter gene lysis buffer (Roche Diagnostics, Mannheim, Germany) for both types of assays and the constant light signal luciferase reporter gene assay was performed as recommended by the supplier (Roche Diagnostics). The luciferase activities were normalized with respect to ␤-galactosidase activity, and induction factors were calculated as the ratio of luciferase activity of ligand-stimulated cells to that of solvent controls.

RESULTS
Conformations of the VDR, bound by saturating concentrations (10 M) of 1␣,25-(OH) 2 D 3 or the 25-carboxylic ester analogue ZK159222 (for structures see Fig. 1), were analyzed by limited protease digestion assays, which provided two digestion products, c1 LPD and c3 LPD, for 1␣,25-(OH) 2 D 3 and an additional digestion product, c2 LPD , for ZK159222 ( Fig. 2A). The VDR fragments c1 LPD , c2 LPD , and c3 LPD have been previously characterized to contain major parts of the LBD and its carboxylterminal truncations, which represent the functional VDR conformations 1, 2, and 3, respectively (30 -33). With monomeric VDR, 1␣,25-(OH) 2 D 3 stabilized 50% and ZK159222 33% of all VDR molecules in c1 LPD , which is known to be the most ligandsensitive VDR conformation (30). In the presence of RXR and DNA, i.e. when VDR-RXR heterodimers are formed on a specific VDRE, the amount of c1 LPD stabilization is increased to 75 and 48% of VDR input, respectively. The amount of VDR molecules that were in conformations c2 LPD and c3 LPD did not exceed 20%, both under DNA-independent and DNA-dependent conditions. However, due to the fact that ZK159222 stabilized three VDR conformations, the total amount of ligandstabilized VDR molecules was equal for both compounds at saturating ligand concentrations. Ligand-dependent gel shift assays were performed with in vitro translated VDR-RXR heterodimers bound to the rat ANF DR3-type VDRE and graded concentrations of 1␣,25-(OH) 2 D 3 or ZK159222 (Fig. 2B). A comparable intensity (approximately 30% shifted probe) of dose-dependent VDR-RXR heterodimer complex formation on DNA was observed for both compounds providing EC 50 (half-maximal activation) values of 0.14 and 1.0 nM for 1␣,25-(OH) 2 D 3 and GST pull-down assays were performed with bacterial produced GST-TIF 646 -926 fusion protein (containing the nuclear receptor interaction domains) and in vitro translated 35 S-labeled VDR protein at a saturating concentration (10 M) of 1␣,25-(OH) 2 D 3 and ZK159222 (Fig. 3A). 1␣,25-(OH) 2 D 3 mediated a precipitation of up to 18% of VDR input, whereas in the presence of ZK159222 the precipitation of VDR protein was not significantly higher than solvent control. Gel shift assays were performed with in vitro translated VDR-RXR heterodimers bound to the rat ANF DR3-type VDRE in the presence of approximately 3 g of GST-SRC-1 596 -790 , GST-TIF2 646 -926 , or GST-RAC3 673-1106 fusion proteins and saturating concentrations (10 M) of 1␣,25-(OH) 2 D 3 and ZK159222 (Fig. 3B). In the presence of 1␣,25-(OH) 2 D 3 , VDR-RXR-VDRE coactivator complexes were observed with all three p160 family members, whereas in the presence of ZK159222 a supershift could not be detected with any of the tested coactivators. The quantification of the relative intensities of these VDR-RXR and VDR-RXR coactivator complexes indicated that ZK159222 stabilized the same amount of VDR-RXR heterodimers compared with 1␣,25-(OH) 2 D 3 , but these heterodimers did not demonstrate interaction with coactivator proteins.
Mammalian one-hybrid assays were performed in HeLa cells that were transiently transfected with an expression vector for a fusion protein containing the DBD of the yeast transcription factor GAL4 and the LBD of the VDR together with a reporter gene construct containing a GAL4 binding site-driven luciferase gene in the presence of graded concentrations of 1␣,25-(OH) 2 D 3 and ZK159222 (Fig. 4A). In this assay system, 1␣,25-(OH) 2 D 3 induced reporter gene activity in a typical dose response (potency of 14-fold induction and ligand sensitivity, i.e. EC 50 value, of 1.0 nM), whereas ZK159222 showed very weak potency (2-fold induction at saturating concentrations) and low ligand sensitivity (EC 50 value of 120 nM). Reporter gene assays in COS-7 cells, that were transiently transfected with expression vectors for VDR and RXR and a DR3-type VDRE-driven luciferase reporter construct, showed similar results (Fig. 4B). 1␣,25-(OH) 2 D 3 provided a typical dose response with an EC 50 value of 1.7 nM, whereas in this assay system, ZK159222 also showed clearly weaker potency (2-fold induction at saturating concentrations) and lower ligand sensitivity (EC 50 value of 10 nM).
Antagonistic effects of ZK159222 on 1␣,25-(OH) 2 D 3 signaling were tested in vivo and in vitro by applying a saturating concentration of 1␣,25-(OH) 2 D 3 with a up to 1000-fold higher con-

FIG. 2. 1␣,25-(OH) 2 D 3 and ZK159222
can act as VDR agonists. Limited protease digestion assays (A) were performed by preincubating in vitro translated 35 Slabeled VDR alone or in combination with the unlabeled RXR and the unlabeled DR3-type VDRE from the rat ANF gene promoter in the presence of saturating concentrations (10 M) of 1␣,25-(OH) 2 D 3 or ZK159222. After digestion with trypsin, samples were electrophoresed through 15% SDS-polyacrylamide gels. The amount of ligand-stabilized VDR conformations 1 (c1 LPD ), 2 (c2 LPD , only in case of ZK159222), and 3 (c3 LPD ) in relation to VDR input was quantified by phosphorimaging. Gel shift experiments (B) were performed with in vitro translated VDR-RXR heterodimers that were preincubated at room temperature with graded concentrations of 1␣,25-(OH) 2 D 3 or ZK159222 and the 32 P-labeled DR3-type VDRE from the rat ANF gene promoter. Protein-DNA complexes were separated from free probe through 8% nondenaturing polyacrylamide gels. The amount of VDR-RXR-VDRE complexes in relation to free probe was quantified by phosphorimaging. Representative experiments are shown. Columns (A) or data points (B) represent the mean of triplicates, and the bars indicate standard deviation. The EC 50 values for VDR-RXR-VDRE complex formation were determined from the respective dose-response curves (B). centration of ZK159222. In reporter gene assays in COS-7 and MCF-7 cells (Fig. 5, A and C) and in mammalian one-hybrid assays in HeLa cells (Fig. 5B) Fig. 2B). In GST pulldown assays (Fig. 5D) and supershift assays (Fig. 5E) with GST-TIF2 646 -926 , 100 nM 1␣,25-(OH) 2 D 3 mediated precipitation of up to 18% of in vitro translated 35 S-labeled VDR protein and 10 nM 1␣,25-(OH) 2 D 3 provided a shift of nearly all VDREcomplexed VDR-RXR heterodimers into complexes with TIF2, respectively. In contrast, 10 M ZK159222 did not provide a significant precipitation of VDR or a supershift of DNA-complexed VDR-RXR heterodimers (as already shown in Fig. 3, A  and B). However, the combination of 100 nM 1␣,25-(OH) 2 D 3 with 10 M ZK159222 provided a reduction of VDR precipitation by 40% (Fig. 5D). Moreover, a combination of 10 nM 1␣,25-(OH) 2 D 3 with 10 M ZK159222 resulted in an equal amount of VDR-RXR-VDRE and VDR-RXR-VDRE-TIF2 complexes (Fig.  5E). The supershift experiments (Fig. 5E) have been performed at different doses of ZK159222 and indicated the dose dependence of the antagonistic action of ZK159222. 1␣,25-(OH) 2 D 3 -mediated transactivation is a multistep event, in which the binding of VDR-RXR heterodimers to a VDRE is one of the first critical steps. Interestingly, on this level the 1␣,25-(OH) 2 D 3 analogue that was characterized in this study, ZK159222, displayed the profile of a weak VDR agonists that requires an approximate 7-fold higher concentration than of the natural hormone 1␣,25-(OH) 2 D 3 to stabilize VDR-RXR heterodimer complex formation on a DR3-type VDRE. ZK159222 was found to belong to the category of 1␣,25-(OH) 2 D 3 analogues that stabilize an additional third functional VDR conformation, which has also been described for some agonistic 20-epi analogues (32,34). However, under the conditions used in this study agonistic analogues were not found to stabilize conformation c2 LPD (data not shown). Like 1␣,25-(OH) 2 D 3 and most other potent VDR agonists, and in particular in DNA-bound heterodimeric complexes with RXR, ZK159222 also stabilizes the high affinity VDR conformation c1 LPD most prominently (30,33).
This study confirmed previous reports that VDR interacts in a ligand-dependent fashion with the three members of the p160 family, SRC-1, TIF2, or RAC3 (35,36). 1␣,25-(OH) 2 D 3 stimulated this VDR-coactivator interaction, whereas ZK159222 was found to be unable to induce an interaction between VDR and SRC-1, TIF2, and RAC3, neither with VDR monomers in solution nor with heterodimeric complexes with RXR on DNA. This lack of interaction with coactivators appears to be the reason why the inducibility of ZK159222-stimulated reporter gene activity demonstrated to be very low in comparison to that of the natural hormone 1␣,25-(OH) 2 D 3 . This effect could be observed in mammalian one-hybrid assays in HeLa cells, in VDRE-driven reporter gene assays performed in COS-7 cells that overexpressed VDR-and RXR-overexpressing COS-7 model system, as well as in the natural 1␣,25-(OH) 2 D 3 target cell line MCF-7. A combination of a saturating 1␣,25-(OH) 2 D 3 concentration with a 100-fold higher concentration of ZK159222 resulted in a significant antagonistic effect in all these in vitro and in vivo assay systems: the amount of 1␣,25-(OH) 2 D 3 -induced VDR-coactivator interaction in solution and on DNA was found to be reduced as well as that of 1␣,25-(OH) 2 D 3 -induced reporter gene activity. A 100-fold higher concentration of ZK159222 appears to be sufficient for effectively competing with the natural hormone in occupying the majority of the VDR molecules in these different experimental systems. This would then result in the observed reduced interaction with coactivators and the subsequent decrease of ligand-induced reporter gene activity.
The antagonistic mechanism described here for ZK159222 is also likely to apply to other VDR ligands. If a compound, which does not necessarily have to be a classical 1␣,25-(OH) 2 D 3 analogue, binds with an affinity to the ligand binding cleft of the VDR that is in the order of the EC 50 value of the VDR-1␣,25-(OH) 2 D 3 interaction, i.e. 0.1 nM, but in parallel does not enable the receptor to interact with coactivators, it may act as an antagonist of 1␣,25-(OH) 2 D 3 signaling. This suggests that the transactivation potency of a VDR-binding ligand, i.e. its fold induction, should be taken in relation to its interaction sensitivity with the VDR, i.e. its EC 50 value. Structural relatives of ZK159222 also show a low potency in reporter gene assays, but their affinity for the VDR is even lower than that of ZK159222, such that a 1000-fold or higher excess of these compounds is needed to obtain an antagonistic effect (data not shown). Moreover, tissue-specific differences in coactivator expression as well as in analogue metabolism may cause tissue-specific differences in the extent of the antagonistic effects of VDR ligands. However, there may also be other mechanisms of antagonism in 1␣,25-(OH) 2 D 3 signaling, e.g. a prevention of VDR-RXR complex formation on DNA, as suggested for the 26,23lactone 1␣,25-(OH) 2 D 3 analogue TEI-9647 (26).
The AF-2 domain in helix 12 was described to be repositioned after ligand binding to the LBD (37) and provides an interface for nuclear receptors together with amino acids of helices 3 and 5 for the binding of coactivators (38). A comparison of the crystal structure of agonist-and antagonist-bound ER suggests that antagonists block AF-2 function by disrupting the topography of the AF-2 surface (24). In analogy to this, one would expect that ZK159222 also stabilizes a VDR conformation that differs from the agonistic conformation c1 LPD and in which the AF-2 is functionally blocked (25). This could be c2 LPD , which is stabilized only by ZK159222 and its antagonistic relatives and not by agonists, if physiological ionic strength conditions are used (data not shown). One could speculate that the rather long side chain of ZK159222 results in an alternative packing arrangement of ligand-binding pocket residues. This may then result in a conformation of the LBD, where helix 12 reaches the static region of the AF-2 surface, which was shown in the case of ER to mimic bound coactivator (25).
In conclusion, the 25-carboxylic ester 1␣,25-(OH) 2 D 3 analogue ZK159222 has been characterized in this report as a novel type of VDR antagonist with a partial agonistic character, where the mechanism of antagonistic action is based mainly on a lack of induction of VDR-coactivator interaction.