Identification of a Vitamin D Response Element in the Proximal Promoter of the Chicken Carbonic Anhydrase II Gene*

The carbonic anhydrase II gene, whose transcription is enhanced by 1,25-dihydroxyvitamin D3(1,25-(OH)2D3), encodes an important enzyme in bone-resorbing cells derived from the fusion of monocytic progenitors. We analyzed the 1,25-(OH)2D3-mediated activation of the avian gene by transient transfection assays with promoter/reporter constructs into HD11 chicken macrophages and by DNA mobility shift assays. Deletion and mobility shift analyses indicated that the −62/−29 region confers 1,25-(OH)2D3responsiveness and forms DNA-protein complexes. The addition of an anti-vitamin D receptor (VDR) antibody inhibited binding to this sequence, whereas anti-retinoid X receptor (RXR) antibody generated a lower mobility complex. Therefore, we concluded that this element binds a VDR·RXR heterodimer, but the addition of extra 1,25-(OH)2D3 had no effect on the formation of this complex. Moreover, the use of nuclear extracts from 1,25-(OH)2D3-treated macrophages led to the formation of an additional high mobility complex also composed of VDR·RXR heterodimer. Mutations provided evidence that the 1,25-(OH)2D3-mediated activation of the carbonic anhydrase II gene is mediated by VDR·RXR heterodimers bound to a DR3-type vitamin D response element with sequence AGGGCAtggAGTTCG. This vitamin D response element is also functional in the ROS 17/2.8 osteoblasts.

Recent work points to the complexity of the molecular mechanisms involved in the vitamin D 3 signaling pathway. It has been known for some time that in addition to the binding of the vitamin D receptor (VDR) 1 to certain vitamin D response elements (VDREs) as homodimer (3)(4)(5)(6)(7), the in vitro binding affinity of the VDR is enhanced by dimerization with accessory factors such as RXRs (8 -11). The response elements for these receptors differ from one another by the number of base pairs (bp) spacing the hexameric repeats according to the so-called 1 to 5 rule (1,2). Following the 1 to 5 rule, optimal VDREs for VDR⅐RXR heterodimers should be direct repeats of two hexameric core binding sites spaced by three nucleotides (DR3) (1,3). The high specificity of these DR3-type VDREs was confirmed by identification of some natural and synthetic VDREs (10). The mouse osteopontin VDRE has been shown to bind VDR homodimers with low affinity (3,6,12,13) but VDR⅐RXR heterodimers with high affinity (3,9,14,15). DR3-type elements have also been found in numerous promoters such as the rat osteocalcin gene (16,17), the rat calbindin D-9k (18), the avian integrin ␤3 subunit gene (19), the rat 24-hydroxylase gene (20 -22), the avian carbonic anhydrase II (CAII) gene (23), and mouse p21 (24). However, the binding sites of the VDREs characterized so far vary considerably in their sequences, preventing definition of a real VDRE consensus sequence (7,(25)(26)(27).
The differentiating effects of 1,25-(OH) 2 D 3 have been studied extensively in a large number of in vitro systems using cultures of leukemia cells (28,29), keratinocytes (30 -32), or bone cells (33)(34)(35). Despite this attention, understanding of the mechanisms that lead to the diverse forms of cell differentiation mediated by 1,25-(OH) 2 D 3 is still fragmentary.
Osteoclasts are large, multinucleated and highly polarized bone-resorbing cells. They belong to the monocytic/macrophage lineage (34,36), and their differentiation pathway is partly under the control of 1,25-(OH) 2 D 3 (37)(38)(39)(40). These cells express some characteristic markers such as tartrate-resistant acid phosphatase, vitronectin receptor, calcitonin receptor, and CAII. CAII, expressed at high levels in osteoclasts (41,42), plays an important role in the extracellular acidification required for bone resorption and therefore bone remodeling. In particular, CAII deficiency is one of the factors responsible for the osteopetrosis characterized in humans by a renal tubular acidosis and a cerebral calcification (43).
We have shown previously that 1,25-(OH) 2 D 3 enhances the expression of CAII in chicken primary blood-derived macrophages (38). Furthermore, 1,25-(OH) 2 D 3 activates the CAII gene expression at the transcriptional level in the chicken monocytic BM2 cells induced to differentiate into macrophages by lipopolysaccharides and phorbol 12-myristate 13-acetate (44) as well as in the human promyelocytic leukemia cells HL60 after phorbol 12-myristate 13-acetate stimulation (45). CAII gene expression is also activated transcriptionally by thyroid hormone in normal erythrocytic cells (46), and several domains in the avian CAII promoter have been shown to control the thyroid hormone regulation of transcription (47,48). In previous work, we identified a VDRE between positions Ϫ1203 and Ϫ1187 of the CAII promoter which mediates 1,25-(OH) 2 D 3 responsiveness to the herpes simplex virus thymidine kinase (tk) minimal promoter in the Drosophila SL3 cell line and in human MCF-7 cells (23). This VDRE, bound by a VDR⅐RXR heterodimer, is however not functional in an avian macrophage cell line.
In the present study, we have looked for specific 1,25-(OH) 2 D 3 regulation of the CAII gene transcription in macrophages. We have studied the ligand-dependent transactivation of the avian CAII promoter in the chicken HD11 macrophage cell line in which numerous hormonal nuclear receptors and CAII gene are expressed endogenously. 2 We have looked for hormone response elements in this promoter and localized an element conferring 1,25-(OH) 2 D 3 -mediated activation (VDRE) to a 34-base pair region located Ϫ62/Ϫ29 upstream the transcriptional start site. The VDRE was defined precisely after methylation interference assays and by using mutated forms of this sequence. This VDRE, functional in HD11 and in ROS 17/2.8 cell lines, has a DR3 structure with sequence AGGGCA tgg AGTTCG and is specifically bound by a heterodimer formed by VDR and the ␣ or ␥ isoform of RXR.

EXPERIMENTAL PROCEDURES
Cells, Medium, and Hormones-The HD11 avian macrophage cells (50) were grown in BT88 complete medium described in Ref. 38. The ROS 17/2.8 rat osteoblast-like osteosarcoma cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 1% L-glutamine (Life Technologies, Inc.), and antibiotics. When the medium was supplemented with hormones, we first depleted sera of small lipophilic compounds by treatment with charcoal. Charcoaltreated sera were prepared as follows. 250 ml of serum was added to the activated charcoal/dextran T70 (100:1; Sigma) and incubated at 55°C for 30 min with frequent agitation. Serum was then spun at 2,000 ϫ g for 25 min. After two incubations, the serum was filtered through 0.45-m filters and stored at Ϫ20°C. 1,25-(OH) 2 D 3 was a generous gift from M. Uskokovic (Hoffman-Laroche).
Reporter Plasmids and Expression Vectors-Functional analysis of the CAII promoter was performed using a CAII-chloramphenicol acetyltransferase (CAT) reporter construct, pHHcaCat (Ϫ1321ϩ31). The pHHcaCat plasmid and 5Ј-end deletions of pHHcaCat were described elsewhere (48). Briefly, constructs were obtained by restriction enzymes that cut once within the promoter at positions Ϫ932, Ϫ619, and Ϫ178 to generate pPstCat, pPvuCat, and pApaCat plasmids, respectively. For functional analysis of the proximal CAII promoter, three fragments excluding the TATA box were obtained from pApaCat construct and cloned between HindIII and XbaI sites of pBLCAT2. These constructs contain the fragments comprised between residues Ϫ177 and Ϫ101 (pϪ177Ϫ101tkCat), Ϫ120 and Ϫ31 (pϪ120Ϫ31tkCat), and Ϫ24 and ϩ31 (pϪ24ϩ31tkCat), respectively. For analysis of the 1,25-(OH) 2 D 3 response domain, three overlapping oligonucleotides of the Ϫ120Ϫ31 fragment (pϪ120 -83tkCat, pϪ90 -55tkCat, and pϪ62Ϫ29tkCat) were synthesized and cloned into the XbaI site of pBLCAT2.
Cell Transfection and CAT Assays-5 ϫ 10 6 HD11 cells were transfected by electroporation (290 V, 500 microfarads; Bio-Rad) with 10 g of CAT reporter constructs and 1 g of CMV-␤-galactosidase plasmid as an internal control to normalize for variations in transfection efficiency. Cells were cultured in BT88 complete medium with 10% charcoalstripped sera in 60-mm dishes (Falcon). When appropriate, the medium was supplemented with 10 Ϫ8 M 1,25-(OH) 2 D 3 or vehicle (0.01% ethanol) 2 h after transfection. After 24 h of treatment, the medium was removed, and cells were incubated for an additional 24 h in 2 ml of fresh medium with the appropriate added factors. Transient transfections with the ROS 17/2.8 cells were performed by the calcium phosphate coprecipitation method in 60-mm dishes with 4.5 g of reporter construct and 0.5 g of CMV-␤-galactosidase plasmid. 10 Ϫ8 M 1,25-(OH) 2 D 3 was added 48 h before harvesting the cells, and vehicle was added at the same concentration to the control. After 48 h of treatment, the HD11 or ROS 17/2.8 cells were lysed in Tris-SDS buffer (0.25 M Tris, pH 8, and 0.05% SDS), and extracts were used to measure ␤-galactosidase and CAT activities. CAT activity was measured by the CAT assay on TLC silica plates (Sigma), quantified by PhosphoImager (Molecular Dynamics), and normalized to ␤-galactosidase activity.
Electrophoretic Mobility Shift Assays (EMSAs)-The macrophage nuclear extracts were prepared from the HD11 cells treated for 2 days with 10 Ϫ8 M 1,25-(OH) 2 D 3 or vehicle (ethanol 0.01%) in charcoal-treated medium. The nuclear extracts were prepared as described in Ref. 48. The human recombinant proteins, VDR, RXR␣, RAR␣, ␤, and ␥, and chicken RXR␥ were synthesized in TNT reticulocyte lysate (Promega) from their cDNAs using T7 RNA polymerase. Unprogrammed lysate and recombinant luciferase were used as negative controls. The rat anti-chick VDR monoclonal antibody used in EMSAs was purchased from Chemicon. The mouse anti-RXR (recognizing all RXR subspecies) and the mouse anti-RAR (recognizing all RAR subtypes) monoclonal antibodies were generous gifts from P. Chambon (IGBMC, Illkirch, France). The anti-TR antibody was described elsewhere (48). The anti-Mi antibody was provided by S. Saule (Institut Pasteur, Lille, France) and the anti-P19, directed against the P19 gag retroviral protein, by J. Samarut (ENS, Lyon, France).
EMSAs and antibody supershifts were performed as described elsewhere (48). Briefly, 2 l of in vitro translated proteins or 4 l of nuclear extracts was incubated with ␥-32 P-labeled probe (0.5 ng) in 15 l of binding buffer containing 1 g of poly(dI-dC) (Amersham Pharmacia Biotech) for 20 min at 4°C. The antibodies and 1,25-(OH) 2 D 3 were incubated for 10 min at 4°C before the addition of the labeled probe. Unlabeled probes used in competition (100-fold excess) were added together with labeled probes. The complexes were resolved on 4% nondenaturing polyacrylamide gel in 0.2 ϫ TBE.
Methylation Interference Experiments-Dimethyl sulfate interference assays (51,52) were performed using the Ϫ62/Ϫ29 fragment of the pϪ62Ϫ29tkCat construct as a probe. pϪ62Ϫ29tkCat was linearized by HindIII (or BamHI), dephosphorylated, and the 5Ј-end was labeled with [␥-32 P]ATP. The Ϫ62/Ϫ29-labeled fragment was obtained after BamHI digestion for sense-labeled probe to generate an HindIII-BamHI fragment (or HindIII for antisense-labeled probe) and purified on an 8% polyacrylamide nondenaturing gel. Both Ϫ62/Ϫ29 probes (10 6 cpm) were methylated with dimethyl sulfate for 3 min at 18°C, then used for EMSAs with 1,25-(OH) 2 D 3 -treated nuclear extracts or recombinant VDR and RXR␣ proteins. Free and retarded probes were eluted and sequenced using the Maxam-Gilbert technique. The sequences were analyzed on a 9% polyacrylamide denaturating gel and compared with GϩA sequences obtained with nonmethylated-labeled probes.

The CAII Proximal Promoter Contains a 1,25-(OH) 2 D 3 -positive Regulatory Domain-We have verified previously that the HD11 cells expressed the VDR and CAII mRNAs and proteins.
We have also checked the presence of RXR, RAR, and TR mRNAs by using the reverse transcriptase-polymerase chain reaction technique. 2 Then, to investigate the mechanisms involved in the transcriptional regulation of the CAII gene expression, the 5Ј-flanking region of the chicken CAII gene was examined and searched for putative 1,25-(OH) 2 D 3 response domains. The fragment of the CAII gene, from Ϫ1321 bp upstream and ϩ31 bp downstream of the transcription start site, as well as deleted fragments from positions Ϫ932, Ϫ619, and Ϫ178 were cloned into a CAT reporter plasmid (Fig. 1). Each different construct was transfected together with the CMV-␤galactosidase internal control into HD11 cells treated with either 10 Ϫ8 M 1,25-(OH) 2 D 3 or vehicle (0.01% ethanol) and assayed 48 h later for CAT and ␤-galactosidase activities. The transient expression analyses demonstrated that the fulllength promoter and all of the deleted mutants still responded to 1,25-(OH) 2 D 3 but with various efficiencies (Fig. 1). Indeed, a 5-fold level of induction was seen with pHHcaCat construct, whereas it was 16-fold with the fragment containing the Ϫ178 bp upstream initiation codon (hereafter referred to as the proximal promoter) (Fig. 1). This proximal promoter was activated by 1,25-(OH) 2 D 3 more efficiently than the full-length promoter and all of the intermediate promoters, indicating that an efficient 1,25-(OH) 2 D 3 -positive regulatory domain must be present between positions Ϫ178 and ϩ31 of the CAII gene and that repressor elements must have been deleted.
Identification of a Putative 1,25-(OH) 2 D 3 Response Domain-To characterize further the DNA sequence necessary for 1,25-(OH) 2 D 3 response, we constructed different plasmids containing overlapping fragments of the proximal promoter but excluding the TATA box. Those fragments were fused to the herpes simplex virus tk minimal promoter and the CAT gene to generate pϪ177Ϫ101tkCat, pϪ120Ϫ31tkCat, pϪ24ϩ31tkCat, pϪ120 -83tkCAT, pϪ90Ϫ55ptkCAT, and pϪ62Ϫ29ptkCAT constructs (Fig. 2). These plasmids were transfected into HD11 cells and tested for 1,25-(OH) 2 D 3 activation. The basal activity of the tk promoter in pBLCAT2 vector was not modified significantly by the 1,25-(OH) 2 D 3 treatment (Fig. 2, row 1). Under similar conditions, only pϪ120Ϫ31tkCat and pϪ62Ϫ29ptkCAT construct have retained the ability to respond to 1,25-(OH) 2 D 3 (Fig. 2, rows 4 and 8 to compare with rows 3, 5, and 8). These analyses showed that the only proximal regulatory region from position Ϫ62 to Ϫ29 can confer strong 1,25-(OH) 2 D 3 responsiveness when attached to a heterologous tk promoter. Then, the regulating domain mediating the 1,25-(OH) 2 D 3 -dependent activation of the CAII promoter is located between positions Ϫ62 and Ϫ29. This sequence must contain a potential VDRE conferring 1,25-(OH) 2 D 3 responsiveness to the tk promoter.
VDR and RXR Bind Directly to the Ϫ62/Ϫ29 Fragment as a Heterodimer-The 1,25-(OH) 2 D 3 DR3-type activation may result from the binding of nuclear hormonal receptors to this region. To test this hypothesis, a double-stranded oligonucleotide with a sequence from positions Ϫ62 to Ϫ29 was used as a probe for EMSAs and was incubated with several in vitro translated nuclear hormonal receptors (Fig. 3A). The synthetic consensus VDRE DR3, described in Ref. 1 and known to bind VDR⅐RXR heterodimer, was used as a positive control and a sequence corresponding to an AP-1 binding site as a negative control. No specific binding was ever observed when the Ϫ62/ Ϫ29 probe was incubated with unprogrammed lysate or recombinant luciferase (data not shown). Moreover, VDR, RXR (␣ or ␥ isoforms), or RAR (␣, ␤, ␥ isoforms) alone did not generate any specific complexes with the Ϫ62/Ϫ29 region (Fig. 3A, lanes 2  and 8 -12) or RAR in combination with VDR (Fig. 3A, lanes  5-7). The only specific signal was obtained when the Ϫ62/Ϫ29 probe was incubated with a mixture of VDR and either RXR␣ or ␥ isoforms (Fig. 3A, lanes 3 and 4). Mobility shift experiments clearly demonstrated the RXR requirement for the binding of the VDR and the cooperative binding of the VDR⅐RXR heterodimer to theϪ62/Ϫ29 region of the CAII gene. This signal was not affected by the addition of 10 Ϫ6 M 1,25-(OH) 2 D 3 to the binding assay (data not shown). This VDR⅐RXR complex, named C1, was competed specifically by a 100-fold excess of unlabeled DR3 or Ϫ62/Ϫ29 oligonucleotides (Fig. 3B, lanes 3 and 7 to compare with lane 2). In contrast, unlabeled oligonucleotides containing an AP-1 site or corresponding to the Ϫ120/ Ϫ83 and Ϫ90/Ϫ55 regions of the CAII promoter failed to compete (Fig. 3B, lanes 4 -6). Most importantly, the specific retarded band C1 is disrupted by the antibody directed against the DNA binding domain of VDR (Fig. 3B, lane 10) and supershifted by the anti-RXR antibody (Fig. 3, lane 12), but this complex is not affected by an unrelated anti-Mi antibody (Fig.  3B, lane 11). These data confirmed the binding specificity of the VDR⅐RXR heterodimer to the Ϫ62/Ϫ29 region of the CAII promoter. Therefore VDR and RXR proteins are necessary and sufficient to form a C1-binding complex on Ϫ62/Ϫ29 sequence.
Ϫ62/Ϫ29 Fragment Binds to VDR⅐RXR Heterodimers Present in HD11 Nuclear Extracts-Nuclear extracts of HD11 cells treated for 2 days with 1,25-(OH) 2 D 3 were incubated with labeled probes corresponding to the Ϫ62/Ϫ29, Ϫ120/Ϫ83, or Ϫ90/Ϫ55 DNA fragments of the CAII promoter, and the resulting complexes were analyzed by EMSA. No specific binding was observed on either Ϫ120/Ϫ83 or Ϫ90/Ϫ55 probes (data not shown), but two specific complexes, C1 and C2, were revealed with the Ϫ62/Ϫ29 sequence (Fig. 4, lanes 2 and 8). As for the complex obtained with recombinant proteins (Fig. 3B), these two complexes were both specifically competed by an excess of unlabeled Ϫ62/Ϫ29 probe (Fig. 4, lane 3) or DR3 oligonucleotide but not by an excess of AP-1, Ϫ120/Ϫ83, or Ϫ90/Ϫ55 oligonucleotides (data not shown). To identify the proteins implicated in these complexes, antibodies directed against different transcription factors were used to interfere with the protein-DNA interactions. The anti-VDR antibody induced a disappearance of the two binding complexes, and the antibody against RXR further retarded both C1 and C2 complexes to lower mobility complexes (Fig. 4, lanes 4 and 5). Thus, RXR is present in both C1 and C2 complexes formed with the VDR on this VDRE. No significant binding interference was observed with the antibodies directed against P19 (anti-gag antibody), RAR, TR␣, or Mi (Fig. 4, lanes 6 and 9 -11). Thus the C1 and C2 complexes are both composed of VDR⅐RXR heterodimers bound to DNA. Taken together, these results indicated that the nuclear extracts contained the proteins essential for 1,25-(OH) 2 D 3 activation of the CAII promoter and for the binding to the The deleted fragments of the CAII proximal promoter were obtained from the pApaCat plasmid (Ϫ178/ϩ31) and cloned into pBLCAT2. 10 g of the reporter constructs was transfected into HD11 cells. pBLCAT2 and pApaCat constructs were used as negative and positive controls, respectively. After transfection, HD11 cells were cultured in the presence or absence of 10 Ϫ8 M 1,25-(OH) 2 D 3 for 48 h.
Ϫ62/Ϫ29 region. This binding complex is at least composed of RXR and VDR nuclear proteins.
Role of 1,25-(OH) 2 D 3 on Protein Binding to DNA-Mobility shift experiments were also performed with nuclear extracts from cells treated for 2 days with vehicle. When the Ϫ62/Ϫ29 probe was incubated with those nuclear extracts only the C1 low mobility complex was detected (Fig. 5, lane 6), and this complex was competed by an excess of cold probe (Fig. 5, lane  9). The high mobility C2 complex, absent from vehicle-treated nuclear extracts (Fig. 5, lane 6), was detected only in cells treated with 1,25-(OH) 2 D 3 (Fig. 5, lane 2). Those different results could be explained by the cell pretreatment with 1,25-(OH) 2 D 3 before the preparation of the nuclear extracts. Thus, we hypothesized that the C2 complex could be due to 1,25-(OH) 2 D 3 binding onto its receptor and inducing a change in complex conformation and in gel mobility. Therefore we have tested the 1,25-(OH) 2 D 3 impact on DNA binding by the addition of 1,25-(OH) 2 D 3 to the EMSA reaction mixture. Adding 1,25-(OH) 2 D 3 did not modify the complex formation observed with 1,25-(OH) 2 D 3 or vehicle-treated nuclear extracts (Fig. 5,  lanes 3 and 4; 7 and 8). These results clearly show that the C2 complex is not caused by the immediate binding of the ligand to the VDR. Nevertheless, this C2 complex could be the consequence of a conformational change of the VDR⅐RXR heterodimer bound to the VDRE compared with the C1 complex. The in vivo bound ligand could induce this conformational change and allow the recruiting of a third protein. Thus C2 appears to be the physiological complex since it corresponds to the specific complex formed in 1,25-(OH) 2 D 3 -treated cells. These experiments demonstrated that a VDRE is located in the region between residues Ϫ62 and Ϫ29 of the CAII promoter which can bind specifically the VDR and RXR present in the nuclear extracts of the HD11 macrophages.
Identification of the Putative VDRE by Mutational Analysis-To define more precisely the VDR⅐RXR binding site on the CAII proximal promoter, the Ϫ62/Ϫ29 region obtained from the pϪ62Ϫ29tkCat construct was used as a probe for methylation interference assays (Fig. 6A). The guanine residues of this domain involved in the DNA-protein interactions are shown in Fig. 6B. In this area we could easily point out an AGGGCA 5Ј-half site, between positions Ϫ53 and Ϫ48 (framed sequence in Fig. 6B), quite homologous to the published consensus VDRE hexamer and including some of the protected residues (underlined nucleotides). On the other hand, the 3Ј-putative hexameric half-site could consist of either a GGAGTT between residues Ϫ46 and Ϫ41 in the case of a DR1 element or an AGTTCG between residues Ϫ44 and Ϫ39 for a DR3 element. The only detectable specific complexes were composed of VDR⅐RXR␣ or VDR⅐RXR␥ heterodimers. Panel B, the VDR⅐RXR␣ binding to the Ϫ62/ Ϫ29 DNA fragment is specific. EMSAs were performed using recombinant VDR and RXR␣ incubated with the Ϫ62/Ϫ29-labeled probe. Left, DR3 is an artificial VDRE containing two directly repeated AGGTCA motifs spaced by 3 bp. AP1 is the DNA binding site for the AP-1 complex. Ϫ62, Ϫ90, and Ϫ120 correspond to the Ϫ62/Ϫ29, Ϫ90/Ϫ55, and Ϫ120/Ϫ83 regions of the CAII promoter, respectively. The unlabeled competitors ( To discriminate between these two possibilities, we cloned oligonucleotides containing 3-bp mutations between positions Ϫ56 and Ϫ36 (Fig. 7) upstream from the tk promoter fused to the CAT gene. These constructs were transfected into HD11 cells and tested for putative activation by 1,25-(OH) 2 D 3 . Compared with the pϪ62Ϫ29tkCat construct (Fig. 7, row 1), we observed that the 1,25-(OH) 2 D 3 stimulation obtained with the m1tkCat and m7tkCat constructs with mutations localized outside the interfering area was not affected (Fig. 7, rows 2 and 8). Stimulations with m2, m3, m5, and m6 constructs carrying mutations within the interfering domain were abolished completely (Fig. 7, rows 3, 4, 6, and 7, respectively). Stimulation obtained with the m4tkCat construct was only slightly decreased in response to 1,25-(OH) 2 D 3 (Fig. 7, row 5). These results indicate that only mutations between positions Ϫ53 and Ϫ39 altered the 1,25-(OH) 2 D 3 responsiveness, defining a more precise putative VDRE. Among the mutants showing a loss of induction, m4tkCat is the less effective. It appears that the mutations fall upon the three spacing nucleotides of the putative DR3 element.
DNA Binding Analysis of the Ϫ62/Ϫ29 Region Mutants-Oligonucleotides containing mutations described in Fig. 7 were tested in EMSAs to compete for the formation of DNA-protein complexes on the Ϫ62/Ϫ29 sequence. Ϫ62/Ϫ29-labeled probe was incubated with nuclear extracts from 1,25-(OH) 2 D 3treated HD11 cells, and mutated oligonucleotides were used as competitors; the EMSAs are shown in Fig. 8A. The mutants can be classified within three groups according to their ability to bind specific protein complexes. The first one, with the m1 and m7 mutants, localized outside the interfering area, can compete for the DNA-protein binding (Fig. 8A, lanes 4 and 10) equally as well as the unlabeled Ϫ62/Ϫ29 oligonucleotide (Fig.  8A, lane 3). The m3 and m4 mutants, in the second group, were also able to displace the bound complexes but to a lesser extend than the native sequence (Fig. 8A, lanes 6 and 7). Third, Fig. 8A also shows that the m5 mutant (lane 8) has completely lost the ability to bind, whereas m2 and m6 mutants seem to have retained a slight binding (lanes 5 and 9).
Therefore, the ability of these different mutated sequences to bind nuclear proteins was tested directly. To identify the complexes generated, we used these mutants as probes for EMSAs with 1,25-(OH) 2 D 3 -treated nuclear extracts. One probe representative of each of the three groups (i.e. m1, m2, m3) was used in the EMSAs that are shown in Fig. 8B. As expected from the previous results, m1 and m3 probes (Fig. 8A) as well as m1 and m7 (data not shown) were able to form complexes with VDR and RXR as shown by treatment with both anti-VDR and anti-RXR antibodies, and this binding is competed by an excess of cold-related probes (Fig. 8B, lanes 1-6 and 13-18, respectively). The m2 mutant, in one element of the two putative hexameric motifs of the VDRE, failed to generate any significant complex with HD11-derived nuclear proteins (Fig. 8B,  lanes 7-12); the same results were observed using m5 and m6 as probes (data not shown). These results correlate with the abrogation of the functional response obtained with these mutants. So those mutations (m2, m5, and m6) inhibit the transcription activation by preventing the DNA binding of the  2-4), whereas only the C1 complex was observed with untreated extracts (lanes 6 -8). Both C1 and C2 complexes were competed by an excess of cold probe (lanes 5 and 9). No change in the protein-DNA binding pattern was observed in the presence of in vitro added 1,25-(OH) 2 D 3 .

FIG. 6. Methylation interference experiments.
Panel A, methylation interference assays were performed with methylated Ϫ62/Ϫ29 probes radiolabeled on the 5Ј-end of the sense or antisense strands. The probes were incubated with VDR and RXR␣ recombinant proteins or 1,25-(OH) 2 D 3 -treated HD11 cell nuclear extracts for EMSAs. The free and retarded probes were then sequenced and compared with the GϩA sequence obtained with the nonmethylated Ϫ62/Ϫ29-labeled probes. The G residues implicated in protein-DNA interactions were identified by a disappearance of the signal or a weaker signal. Panel B, summary of the interactions between VDR⅐RXR proteins and the Ϫ62Ϫ29 domain of the CAII promoter. The guanines identified by methylation interference assays are presented on a double-stranded DNA fragment. transcription factors. In conclusion, the oligonucleotides containing mutations in the second hexameric core motif of the putative DR3-type element (m5 and m6 mutants) were not able to bind nuclear proteins nor to respond to 1,25-(OH) 2 D 3 . In contrast, mutations in the 5Ј-binding site AGGGCA differ in their effects: the first three nucleotides (m2 mutant) are essential for both binding and transactivation, whereas the last three (m3 mutant) are essential for transactivation but not for binding.
The Identified VDRE Is a DR3 Element-From the previous experiments it was clear that the sequence comprised between positions Ϫ53 and Ϫ39 of the CAII promoter was a VDRE. At this point we could not completely rule out that it was DR1 element (AGGGCA t GGAGTT cg) rather than a DR3 element (AGGGCA tgg AGTTCG). It has been shown that the second base of the 3Ј-element of a VDRE is the more conserved residue among all VDREs described so far. We then constructed two other mutated Ϫ62/Ϫ29 oligonucleotides, namely m8 and m9 mutants, in which the G residue in position Ϫ45 (for the putative DR1 element) and Ϫ43 (for the putative DR3 element) was mutated to a C residue. Transient transfection experiments have revealed that the 1,25-(OH) 2 D 3 -mediated response obtained with m8tkCat reporter plasmid is similar to the pϪ62Ϫ29tkCat construct, whereas m9tkCat has completely lost the 1,25-(OH) 2 D 3 response capability showing that the complex bound needs to contact this G residue for efficient transactivation (Fig. 9A, lanes 4 and 6, respectively). We then tested their capability to bind nuclear proteins by using EMSAs with radiolabeled Ϫ62/Ϫ29, m8, or m9 probes. As shown in Fig.  9B, the m8 mutant can still compete with the Ϫ62/Ϫ29 oligonucleotide for the binding of HD11 nuclear extracts (lane 4) and bind VDR⅐RXR heterodimer (lanes 6 -10), whereas the m9 mutant is unable to do so ( lanes 5 and 11-15). These experiments have confirmed that the m8 mutant is identical in function to the native VDRE and that the m9 mutant is no more able to bind the VDR⅐RXR heterodimer specifically. Then, we concluded that the G residue in position Ϫ43 is essential for both transactivation and binding of macrophage-derived nuclear extracts. This result clearly confirms that the CAII proximal VDRE is a DR3 with sequence AGGGCA tgg AGTTCG.
CAII Proximal VDRE Is Functional in a Heterologous Cellular System-To confirm further the validity of the CAII proximal VDRE, constructs containing this VDRE were tested in a heterologous cellular system. We undertook transient transfection experiments in the rat osteosarcoma cell line ROS 17/2.8 cells with the different tkCat reporter constructs described above. These cells were transfected with the tkCat reporter constructs and the CMV-␤-galactosidase as an internal control and were then treated for 2 days with 10 Ϫ8 M 1,25-(OH) 2 D 3 . The results presented in Fig. 10 2, 8, and 14) were analyzed by competition with unlabeled probe or Ϫ62/Ϫ29 fragment (Ϫ62) (50 ng) and by using antibodies directed against avian VDR (␣-VDR, dilution 1/15) or human RXR (␣-RXR, dilution 1/40). Representative results obtained with m1, m2, and m3 mutants are shown. and is also able to confer 1,25-(OH) 2 D 3 responsiveness to the tk promoter in a heterologous system. Moreover, by using mutated forms of this VDRE (Fig. 10, lanes 3-9) we have confirmed that, like in HD11 cell line, this VDRE is also of the DR3 type in ROS 17/2.8 cells. DISCUSSION In this work, we showed that the chicken CAII promoter activity is induced in response to 1,25-(OH) 2 D 3 in an avian macrophage cell line. We identified a domain in the CAII promoter responsible for 1,25-(OH) 2 D 3 -mediated transactivation. Transient transfection assays in HD11 macrophages of total or deleted fragments of the CAII promoter allowed us to define a region involved in 1,25-(OH) 2 D 3 responsiveness. In macrophages, most if not all DNA sequences essential for CAII gene basal and 1,25-(OH) 2 D 3 -induced expression are within the first 178 bp upstream of the initiation site, referred to as the proximal promoter. It is noteworthy that the level of 1,25-(OH) 2 D 3 activation increases with the 5Ј-end progressive deletions of the promoter. A repressor domain further upstream may be present and would explain the increasing activation of CAII pro-moter correlated with deletions. Further deletions and mutational analyses of the CAII proximal promoter allowed us to determine a more precise vitamin D-responsive domain, between residues Ϫ53 and Ϫ39 of the CAII promoter. We showed that this VDRE is highly functional in macrophages and is bound specifically by a complex formed by a RXR⅐VDR heterodimer. Lastly, we showed that mutations that abolished protein binding to the VDRE inhibit vitamin D-dependent activation in macrophages.
Many have shown that the VDR has a binding preference for the direct repeat composed of AG(G/T)TCA motifs spaced by 3 bp (2,3,53,54). The CAII proximal VDRE consists of an imperfect tandem of 6 bases spaced by three nucleotides with the sequence AGGGCA for the 5Ј-motif and AGTTCG for the downstream motif. Mutational analysis revealed the validity of the DR3 structure for the CAII proximal VDRE. Oligonucleotides containing mutations within the spacer motif of this DR3 (m4 and m8 mutants) were still able to bind VDR⅐RXR heterodimer and have a very mild effect on transactivation efficiency. In contrast, mutations within the two hexameric motifs of the DR3 element inhibited 1,25-(OH) 2 D 3 -mediated transcriptional activation and most of the protein-DNA interactions. Furthermore, our results indicate that the 3Ј-element integrity of this VDRE is more crucial for binding than the 5Ј-element. This is evidenced by the ability of a mutated 5Јelement VDRE (m3 mutant) to compete still for VDR binding with the native sequence, whereas mutations in the 3Ј-element (m5 and m6 mutants) resulted in a marked decrease in VDR binding to the mutant element. In contrast with others studies showing the important role of the residues in 5Ј-position outside the VDRE, we have shown here that the nucleotides located outside the VDRE were not essential either for the binding of the protein complex on DNA or for the transactivation activity.
The sequences of numerous known natural positive VDREs, generally of the DR3 type, identified within the promoter regions of different genes, have been aligned. It is of interest to note that the hexameric core binding sites are rather degenerated, although they all can specifically bind VDR complexes and confer transactivation upon 1,25-(OH) 2 D 3 stimulation. We have observed that the upstream motif of the consensus VDRE is more conserved than the downstream one. The difference between the two half-sites may indicate preferential binding of Ϫ62/Ϫ29 sequences containing point mutations were cloned into pBLCAT2 plasmid, and the constructs were transfected into HD11 cells treated for 2 days with 10 Ϫ8 M 1,25-(OH) 2 D 3 . The pϪ62Ϫ29tkCat construct was used as a positive control. The m9 mutant is unable to respond to 1,25-(OH) 2 D 3 treatment, whereas the m8 mutant acts as the Ϫ62/Ϫ29 sequence. Panel B, mobility shift experiments with radiolabeled point mutant VDREs. Ϫ62/Ϫ29, m8, and m9 probes were radiolabeled and used for EMSAs with nuclear extracts from 1,25-(OH) 2 D 3 -treated HD11 cells. The m8 mutant can bind equivalently to the native Ϫ62/Ϫ29 sequence, whereas the m9 mutant cannot compete or bind to the VDR⅐RXR heterodimer. each receptor of the complex bound to DNA. Previous studies have shown that RXRs bind preferentially to the 5Ј-core binding site in retinoic acid and thyroid hormone response elements as well as in VDREs (1, 7, 25-27, 57, 58).
Haussler et al. (59) have postulated that the guanine in the second position of the 3Ј-hexamer is absolutely conserved, as we found for the CAII proximal VDRE. Indeed, mutation of this guanine abolishes 1,25-(OH) 2 D 3 responsiveness and the binding of transcription factors to the CAII proximal VDRE. Thus, the 5Ј-motif of the CAII VDRE could be a high affinity RXR binding motif, whereas the 3Ј-hexanucleotide sequence could be a VDR binding motif (60).
Consistent with the results obtained with gel retardation experiments, we have demonstrated by transient transfection experiments that the vitamin D stimulation of the CAII gene expression acts, in vivo, through VDR and RXR dimerization. 2 Furthermore, we have also shown that, in vivo, addition of RXR ligand does not interfere with the 1,25-(OH) 2 D 3 response and so, does not affect the stability of VDR⅐RXR heterodimers in the cells nor the transactivation efficiency (49).
Surprisingly, two specific complexes, C1 and C2, were formed on the CAII proximal VDRE with nuclear extracts from 1,25-(OH) 2 D 3 -treated HD11 cells, whereas only one complex (C1) was formed with vehicle-treated extracts or recombinant proteins. However, C1 and C2 complexes bound to this VDRE are both composed at least of VDR and RXR heterodimers. We have speculated that in vivo, 1,25-(OH) 2 D 3 induces a specific conformational change of the VDR⅐RXR heterodimer leading to the formation of the C2 complex, whereas in vitro, only the C1 complex is observed. This C2 higher mobility complex could be caused by the binding of 1,25-(OH) 2 D 3 to the VDR, although the addition of extra 1,25-(OH) 2 D 3 to the EMSA reaction mixture had no effect on the DNA binding affinity of the protein complex. In such conditions no change in mobility or complex composition was observed. This suggests that in vitro, binding of the ligand is not necessary for the binding of proteins to DNA. In contrast, in vivo, cell exposure to 1,25-(OH) 2 D 3 could change the complex conformation on DNA and in doing so could either recruit or displace the binding of either a coactivator or a corepressor. The two complexes, C1 and C2, obtained by EMSAs using nuclear extracts, may reflect the possible involvement of known or unknown factors in addition to the VDR⅐RXR heterodimer. In fact, the transcription factor TFIIB, which was shown to interact directly with VDR (61), has been described as forming part of in these complexes and appears to be required for the VDR to bind the VDREs (62). Other potential unidentified proteins may also be involved in the binding of the VDR to the VDRE such as positive cofactors, including TRIP1 (63), NCoAϪ62 (64), RIP140, RIP160 (65,66), and SRC1 (67). These accessory factors have been shown to contribute to the transcriptional activation mediated by nuclear hormonal receptors through a direct interaction with those receptors (49, 64 -70). The identified bases of the CAII proximal VDRE may therefore bind a transcription factor that promotes 1,25-(OH) 2 D 3 -mediated transactivation by facilitating interactions between the receptor-occupied VDRE and the basal transcription apparatus.
Finally, we tested the activity of this CAII proximal VDRE in a heterologous cellular system, the ROS 17/2.8 cell line. These analyses have demonstrated that the CAII proximal VDRE is also functional as a VDRE in the ROS 17/2.8 cells but less efficiently. Thus, the VDRE identified in this study allows 1,25-(OH) 2 D 3 transactivation of the CAII gene after binding of the VDR⅐RXR heterodimer, is efficient when cloned upstream a heterologous promoter, and is functional in a heterologous cellular system. Although the distal CAII VDRE is not functional in this macrophage cell line, the CAII proximal VDRE is fully active in the cell type (i.e. macrophages) expressing the CAII gene endogenously, and it is indeed the region promoting the 1,25-(OH) 2 D 3 response.