Physical and Functional Interaction between the Vitamin D Receptor and Hairless Corepressor, Two Proteins Required for Hair Cycling*

Both the vitamin D receptor (VDR) and hairless (hr) genes play a role in the mammalian hair cycle, as inactivating mutations in either result in total alopecia. VDR is a nuclear receptor that functions as a ligand-activated transcription factor, whereas the hairless gene product (Hr) acts as a corepressor of both the thyroid hormone receptor (TR) and the orphan nuclear receptor, RORα. In the present study, we show that VDR-mediated transactivation is strikingly inhibited by coexpression of rat Hr. The repressive effect of Hr is observed on both synthetic and naturally occurring VDR-responsive promoters and also when VDR-mediated transactivation is augmented by overexpression of its heterodimeric partner, retinoid X receptor. Utilizing in vitro pull down methods, we find that Hr binds directly to VDR but insignificantly to nuclear receptors that are not functionally repressed by Hr. Coimmunoprecipitation data demonstrate that Hr and VDR associate in a cellular milieu, suggesting in vivo interaction. The Hr contact site in human VDR is localized to the central portion of the ligand binding domain, a known corepressor docking region in other nuclear receptors separate from the activation function-2 domain. Coimmunoprecipitation and functional studies of Hr deletants reveal that VDR contacts a C-terminal region of Hr that includes motifs required for TR and RORα binding. Finally, in situ hybridization analysis of hr and VDR mRNAs in mouse skin demonstrates colocalization in cells of the hair follicle, consistent with a hypothesized intracellular interaction between these proteins to repress VDR target gene expression, in vivo.

Both the vitamin D receptor (VDR) and hairless (hr) genes play a role in the mammalian hair cycle, as inactivating mutations in either result in total alopecia. VDR is a nuclear receptor that functions as a ligand-activated transcription factor, whereas the hairless gene product (Hr) acts as a corepressor of both the thyroid hormone receptor (TR) and the orphan nuclear receptor, ROR␣. In the present study, we show that VDR-mediated transactivation is strikingly inhibited by coexpression of rat Hr. The repressive effect of Hr is observed on both synthetic and naturally occurring VDR-responsive promoters and also when VDR-mediated transactivation is augmented by overexpression of its heterodimeric partner, retinoid X receptor. Utilizing in vitro pull down methods, we find that Hr binds directly to VDR but insignificantly to nuclear receptors that are not functionally repressed by Hr. Coimmunoprecipitation data demonstrate that Hr and VDR associate in a cellular milieu, suggesting in vivo interaction. The Hr contact site in human VDR is localized to the central portion of the ligand binding domain, a known corepressor docking region in other nuclear receptors separate from the activation function-2 domain. Coimmunoprecipitation and functional studies of Hr deletants reveal that VDR contacts a C-terminal region of Hr that includes motifs required for TR and ROR␣ binding. Finally, in situ hybridization analysis of hr and VDR mRNAs in mouse skin demonstrates colocalization in cells of the hair follicle, consistent with a hypothesized intracellular interaction between these proteins to repress VDR target gene expression, in vivo.
Nuclear receptors comprise a family of ligand-activated transcription factors that coordinate physiological and developmental processes by regulating specific changes in gene expression (1,2). The vitamin D receptor (VDR) 1 mediates signaling by 1,25-dihydroxyvitamin D 3 (1,25(OH) 2 D 3 ) and is a member of the thyroid hormone/retinoic acid receptor subfamily of nuclear receptors that heterodimerize with the retinoid X receptor (RXR) on direct repeat hormone-responsive elements in the promoters of regulated genes (3,4). Binding of liganded VDR⅐RXR to a vitamin D-responsive element (VDRE) in target genes such as osteocalcin, osteopontin, and vitamin D 24-hydroxylase (CYP24) is accompanied by the recruitment of coactivator proteins (5). VDR coactivators such as steroid receptor coactivator-1 (6), NCoA-62 (7), and vitamin D receptor interacting protein 205 (8) stimulate transcriptional activation by facilitating chromatin remodeling and/or attraction of RNA polymerase II. Although some unliganded nuclear receptors (thyroid hormone and retinoic acid receptors) can mediate repression through association with corepressors such as nuclear receptor corepressor (N-CoR) (9) and silencing mediator for retinoic acid and thyroid hormone receptors (SMRT) (10), evidence suggests that VDR does not associate strongly with these corepressors (9 -12).
Targeted gene deletion studies in mice (13,14) and inactivating mutations in humans (15) have revealed multiple biological consequences of VDR signaling. VDR is required primarily for the following: (i) stimulation of calcium and phosphate absorption from the intestine to prevent rickets, (ii) induction of the CYP24 enzyme that initiates the catabolism of 1,25(OH) 2 D 3 , and (iii) progression of the normal hair cycle in mammalian skin. Although VDR gene ablation in mice elicits both rickets and hair loss (13,14), point mutations in human VDR that specifically compromise either 1,25(OH) 2 D 3 ligand (16,17) or coactivator (18) contacts confer rickets without disruption of the hair cycle. Loss of function mutations in human VDR that do result in both rickets and congenital hair loss (alopecia or atrichia) abolish either VDR DNA binding (19 -21) or VDR⅐RXR heterodimerization (17). Interaction of VDR with its heterodimeric partner RXR is likely relevant to hair cycling, as the conditional inactivation of RXR␣ in mouse skin (22) results in alopecia resembling that in VDR-null mice (14).
Like mutations in the gene encoding VDR, mutations in the mammalian hairless (hr) gene result in congenital hair loss in both mice (23) and humans (24,25). Remarkably, the hair loss phenotype caused by specific mutations in the human VDR gene resembles the generalized atrichia caused by mutations in the hr gene (26). The shared hair loss phenotype of hr and VDR mutant animals and humans suggests that both proteins impact a common signaling pathway.
In addition to genetic evidence of a potential relationship between hr and VDR, the function of the hr gene product (Hr) may be relevant to VDR signaling as well. Hr is a nuclear protein with a molecular mass of 130 kDa that is expressed primarily in skin and brain (27)(28)(29) and was shown recently (30) to function as a potent nuclear receptor corepressor. Despite a lack of sequence identity with other corepressors (N-CoR and SMRT), Hr functions in a similar manner; Hr was shown to mediate repression via unliganded thyroid hormone receptor (TR), bind to TR through conserved hydrophobic motifs, and interact with histone deacetylases (29,30). Unlike N-CoR and SMRT, Hr interacts with TR but not retinoic acid receptor (RAR) and can also inhibit transcriptional activation by the RAR-related orphan receptor, ROR (31).
Based on genetic and biochemical evidence, we postulate that a network of interactions between Hr and VDR exists in the skin to drive the progression of the hair cycle. Because VDR and TR are similar in structure, we tested the hypothesis that Hr interacts with VDR as a corepressor. In the present report we show the following: (i) VDR interacts physically with Hr, (ii) this association dramatically represses VDR-mediated transactivation, and (iii) Hr and VDR are coexpressed in cells of the hair follicle. Our results are consistent with the presence of VDR and Hr in a signal transduction cascade in the hair follicle that represses gene expression, possibly silencing a gene that codes for an inhibitor of the hair cycle.
Hormone-responsive element (HRE) reporter plasmids were constructed as follows. Synthetic oligonucleotides containing four copies of the rat osteocalcin direct repeat 3 (DR3) VDRE (GGGTGAATGAGG-ACA) (42), four copies of the rat CRBPII direct repeat 1 (DR1) RXRresponsive element (AGGTCACAGGTCA) (43), or three copies of the rat tyrosine aminotransferase indirect repeat 3 glucocorticoid-responsive element (TGTACAGGATGTTCT) (44) were each cloned into the HindIII site of ptkGH. Each construct includes a herpes simplex virus thymidine kinase (tk) promoter directing basal transcription of a human growth hormone (hGH) reporter gene. DR3x2 tk-luc was constructed by insertion of synthetic oligonucleotides containing two copies of the consensus AGGTCA DR3 VDRE upstream of the minimal tk promoter in tk-luc. p24-OHaseLuc was constructed by subcloning 5.5 kb of the promoter region (45) of the human CYP24 gene (kindly provided by Drs. S. Christakos and J. W. Pike, New Jersey Medical School and University of Wisconsin, respectively) into a firefly luciferase plasmid, lucp1 (45).
Transfection of COS Cells/Transcription Assays-African green monkey kidney cells (COS-1, COS-7) obtained from ATCC were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (Omega Scientific, Tarzana, CA or Invitrogen). For experiments using DR3x2 tk-luc, COS-1 cells were plated in 12-well plates and transfected the following day using LipofectAMINE 2000 (Invitrogen) with 200 ng of reporter plasmid, 75 ng of receptor and/or Hr expression plasmid, and 300 ng of CMV-␤-galactosidase. After 36 h, cells were harvested in passive lysis buffer (Promega, Madison, WI), and extracts were assayed for ␤-galactosidase and luciferase activity. Luciferase activity was divided by ␤-galactosidase activity to normalize for transfection efficiency. Experiments were done in duplicate and repeated at least three times with similar results.
For transfection assays with p24-OHaseLuc, COS-7 cells were transfected with 37.5 ng/well of p24-OHaseLuc, 0.1 ng/well of pRL-CMV (non-regulated Renilla luciferase control), 25 ng/well of pSG5hVDR, and where indicated, 10 ng/well of pRK5myc-rhr, either in the absence or presence of 10 nM 1,25(OH) 2 D 3 . After 48 h, cells were harvested with passive lysis buffer. Firefly and Renilla luciferase activities were measured sequentially from each well using a Sirius Luminometer (Pforzheim, Germany) and dual luciferase reporter assay reagents (Promega, Madison, WI) per the manufacturer's instructions. The ratio of firefly to Renilla luciferase activity was calculated to normalize for transfection efficiency.
GST Pull Down Assays-Wild-type hVDR (32), hRXR␣ (34), and rat hr (28) cDNAs were cloned into the EcoRI site of the GST fusion protein vector, pGEX-4T (Amersham Biosciences) to create GST-hVDR, GST-hRXR␣, and GST-rHr (residues 31-1207). pGEX fusion constructs were transformed into Escherichia coli (strain BL21 for VDR and Hr; strain DH5␣ for RXR␣). The detailed procedure for overexpression of GST fusion proteins has been described previously (46). GST alone was expressed from pGEX-4T in E. coli strain DH5␣ and linked to glutathione-Sepharose beads to serve as a control for background protein association. For GST pull down assays, expression plasmids (1.0 g) were used to generate [ 35  Next, the desired 35 S-labeled protein(s) was incubated with the beads for 30 min at 4°C. When Hr beads were employed in pull down assays, the ligand (1,25(OH) 2 D 3 (10 Ϫ7 M) or dexamethasone (10 Ϫ6 M)) was included in the in vitro transcription/translation reaction. The liganded, synthesized protein was then incubated with GST-Hr beads for 30 min at 4°C as above. In all cases, beads were washed four times with KETZD-0.15 to remove unbound protein(s). The bound proteins were extracted from the beads into loading buffer (4% SDS, 10% ␤-mercaptoethanol, 125 mM Tris-Cl, pH 6.8, 20% glycerol), boiled 3 min, and separated by gradient (5-20%) SDS-PAGE and visualized by autoradiography.
In Situ Hybridization-For the hr-specific probe, a fragment corresponding to nucleotides 1004 -3928 of the mouse hr cDNA (27) was subcloned into pBluescript KSϩ (Stratagene), and the resulting plasmid was linearized with XhoI and NotI to generate templates for sense and antisense probes, respectively. The VDR-specific probe corresponds to nucleotides 71-1371 of mouse VDR cDNA and was made using reverse transcription of mouse brain total RNA (Thermoscript RT-PCR System; Invitrogen) followed by PCR amplification with specific primers (5Ј-TCAGGAGATCTCATTGCCAAAC-3Ј and 5Ј-CAGACCAGAGT-TCTTTTGGTTG-3Ј). The mVDR cDNA was ligated into pCR2.1 (Invitrogen) in both orientations, and the resulting plasmids were linearized with BamHI and used as templates to produce sense and antisense probes. Digoxigenin-labeled cRNA probes were made as described (29).
Serial sections (5 m) of dorsal skin from postnatal day 15 mice were used for in situ hybridization. Sections were fixed in 4% paraformaldehyde and then acetylated/dehydrated in 0.25% acetic anhydride in 0.1 M triethanolamine. Sections were prehybridized with hybridization solution (5ϫ Denhardt's solution, 12.5 mg of yeast tRNA, 5ϫ SSC, 50% formamide) and then incubated with hybridization solution containing ϳ200 ng of probe per slide in a humidified chamber overnight at 60°C. The following day sections were washed at 60°C with 5ϫ SSC, followed by 2ϫ SSC, 0.2ϫ SSC/50% formamide, and 0.2ϫ SSC. To detect the hybridized probe, sections were incubated with anti-digoxigenin-alkaline phosphatase Fab fragments (Roche Applied Science) diluted 1:5000 in Buffer 1 (0.1 M Tris-Cl, pH 7.5, 0.15 M NaCl). After washing, sections were incubated with color solution (0.34 mg/ml nitro blue tetrazolium, 0.175 mg/ml 5-bromo-4-chloro-3-indolyl phosphate, 4-toluidine salt, 0.24 mg/ml Levamisole in 0.1 M Tris-Cl, pH 9.5, 0.1 M NaCl, 0.005 M MgCl 2 ) overnight in a humidified chamber. The sections were dehydrated and cleared overnight in Citrisolv (Fisher). Digital images of sections were obtained as described previously (29).

RESULTS AND DISCUSSION
Hr Inhibits Transactivation by Liganded VDR-The potential effect of Hr on VDR activity was evaluated in a mammalian cell line (COS) that expresses low levels (Յ500 copies/cell) of endogenous VDR. Cells were cotransfected with a hVDR expression plasmid and VDR-responsive reporter constructs, and 1,25(OH) 2 D 3 -induced transcriptional activity was measured in the absence and presence of cotransfected Hr. When the rat osteocalcin VDRE reporter construct was employed, as expected, the 1,25(OH) 2 D 3 ligand markedly enhances transcription via VDR (Fig. 1A). Cotransfection of Hr results in a moderate decrease in basal transcription (40%) and a dramatic reduction in transcriptional activation by 1,25(OH) 2 D 3 -liganded VDR (ϳ7-fold). In addition, Hr sharply represses (11-fold) VDR-mediated transactivation even in the presence of overexpressed RXR␤, which increases 1,25(OH) 2 D 3 -induced transcriptional activation by VDR to 70-fold (Fig. 1A). The observed Hr repression is not because of squelching, as cotransfection of identical amounts of vectors with the same cytomegalovirus promoter used to drive Hr expression does not significantly reduce VDR activity (data not shown). Hr-elicited VDR repression is also specific, because cotransfection of SMRT (41) produced no repression of basal or 1,25(OH) 2 D 3 -stimulated transcription (data not shown).
Because RXR␣ is the receptor isoform considered important in hair cycling (22), we tested the effect of Hr on VDR in the presence of excess RXR␣ (Fig. 1B). Like RXR␤, RXR␣ increases VDR-mediated transactivation, and Hr dramatically curtails transactivation by the 1,25(OH) 2 D 3 -liganded VDR⅐RXR␣ complex (18-fold). The experiment illustrated in Fig. 1B also demonstrates that Hr can repress VDR-mediated transcriptional activation on different promoters, as an alternative VDRE-

FIG. 1. Hr represses ligand-stimulated transcription by VDR-RXR.
A, Hr expression represses VDR-RXR␤-mediated transactivation. COS-7 cells were transfected with pSG5hVDR, pSG5mRXR␤, and pRK5myc-rhr where indicated, together with a rat osteocalcin VDRE-linked reporter gene. Cells were exposed to ethanol vehicle or 10 nM 1,25(OH) 2 D 3 for 24 h, and transcriptional activity was quantitated by growth hormone radioimmunoassay; transcription with hVDR in the presence of 1,25(OH) 2 D 3 was set at 100%. Error bars represent the standard deviation for triplicate analysis. Mock transfections were with empty hVDR vector. B, Hr represses VDR-mediated transactivation in the presence of overexpressed hRXR␣. COS-1 cells were cotransfected with a VDR-responsive reporter gene (two copies of a consensus DR3 VDRE linked to a minimal tk promoter and luciferase gene) together with pCMX-hVDR, pCMX-hRXR␣, and pRK5myc-rhr as indicated. Hormone treatment was with 10 nM 1,25(OH) 2 D 3 and transcriptional activity for hVDR in the presence of 1,25(OH) 2 D 3 was set at 100%. Error bars represent the standard deviation for three experiments done in duplicate. C, Hr represses VDR transactivation induced by LCA. Conditions were as in A, except that LCA was added at 10 Ϫ4 M where designated. D, Hr overexpression does not affect either RXR␣ or GR homodimeric transcriptional activation. Conditions were similar to those in A, except that the RXR-specific ligand was LG5004 (5 M), and dexamethasone (Dex; 1 M) was the GR ligand, used in conjunction with respective RXR-responsive element and glucocorticoid-responsive element reporter constructs. A representative example of three independent experiments is shown in A, C, and D.
reporter system (two copies of a consensus DR3 VDRE) was employed. Hr also effectively repressed VDR⅐RXR␣-mediated transcriptional activation from the rat osteocalcin VDRE that was utilized to generate the results shown in Fig. 1A (data not shown). Thus, the data in Fig. 1, A and B demonstrate that Hr markedly attenuates 1,25(OH) 2 D 3 -stimulated transcription mediated by VDR complexed with its RXR partners, revealing that Hr is a novel corepressor for 1,25(OH) 2 D 3 -activated VDR⅐RXR.
An alternative, lower affinity VDR ligand, the bile acid LCA, was reported recently (50) to trigger VDR-mediated transactivation. As shown in Fig. 1C, LCA yields ϳ50% of the VDRmediated transactivation compared with 1,25(OH) 2 D 3 . Although Hr expression again strongly represses 1,25(OH) 2 D 3stimulated VDR, cotransfection of Hr completely suppresses LCA-induced activation of VDR (Fig. 1C). This observation suggests that VDR bound to LCA is exquisitely sensitive to repression by Hr and demonstrates that Hr can inhibit transcriptional activation induced in response to multiple VDR ligands.
To evaluate the specificity of Hr corepressor activity on VDR, two other nuclear receptors were tested (Fig. 1D). Transcriptional activation by both RXR␣ and GR homodimers in response to their cognate ligands was not affected by coexpression of Hr. Thus, Hrmediated repression is specific for VDR⅐RXR heterodimeric action and does not influence either RXR␣ or GR homodimeric signaling. VDR thereby joins a subset of nuclear receptors that includes TR and the orphan receptor ROR␣ as targets of Hr repressor activity.
Hr Binds Directly to VDR-To determine whether repression by Hr is mediated through the direct interaction of Hr and VDR, GST pull down experiments were utilized. A GST-VDR fusion protein was used to test for interaction with 35 S-labeled Hr. Fig. 2A shows that Hr binds specifically to GST-VDR, as GST control beads show no signal. In contrast to RXR␣-VDR dimerization, which is enhanced by 1,25(OH) 2 D 3 , Hr⅐VDR interaction is unaffected by the 1,25(OH) 2 D 3 ligand. Coincubation with RXR␣ did not increase the apparent Hr⅐VDR association, and Hr bound only weakly to RXR␣ both in the absence and presence of a RXR ligand (rexinoid), LG5009. Thus, Hr binds preferentially to VDR. Interestingly, in the presence of RXR␣, the alternative VDR ligand LCA (Fig. 2, A and B) stimulates VDR⅐Hr binding better than either 1,25(OH) 2 D 3 or a bile acid that does not function as a VDR ligand, CA. Thus, LCA may be superior to 1,25(OH) 2 D 3 in facilitating VDR⅐Hr interaction, consistent with the potent Hr-mediated repression of LCA-liganded VDR activity (Fig. 1C). The ability of Hr to repress transcriptional activation by LCA-bound VDR may be physiologically significant, as VDR knockout (13,14) but not 1,25(OH) 2 D 3 ligand deficiency (51, 52) leads to alopecia. Thus, a hypothesized action of VDR⅐RXR␣ to drive the hair cycle via Hr-mediated corepression may be triggered by a novel bile acid-like VDR ligand in skin.
To determine whether an intact activation function-2 (AF-2)/helix 12 is necessary for interaction of Hr with VDR, we tested a transcriptionally disabled AF-2 mutant of hVDR (E420A) (40). The E420A VDR mutant yields the same pattern of Hr association as does the wild-type hVDR (Fig. 2B), demonstrating that VDR⅐Hr interaction does not require the functional AF-2/helix 12 in hVDR that is known to contact coactivators (40,53). Intriguingly, there exists a natural AF-2 mutation (E420K) in hVDR that causes vitamin D-resistant rickets but does not elicit alopecia (18). The phenotype of this mutant implies that the transactivation function (calcium absorption) and repression actions (hair cycling) of VDR are separable, perhaps with the former mediated by AF-2-coactivator association and the latter effected by Hr contact with a novel domain in VDR.
To investigate whether the apparent nuclear receptor specificity of Hr corepressor activity is reflected at the level of direct contact, we performed pull down assays using GST-Hr and Analysis was performed as in A, using GST-hVDR or GST-E420A hVDR. Ligands were as in A, including CA, a bile acid that does not bind to VDR. C, immobilized Hr binds to VDR (lanes 1, 2, 5, and 6) but does not bind GR (lanes 9, 10). Radiolabeled VDR and GR were produced by in vitro transcription/translation in the absence (ethanol vehicle) or presence of 10 Ϫ7 M 1,25(OH) 2 D 3 and 10 Ϫ6 M dexamethasone, respectively, followed by incubation with immobilized GST alone (negative control; lanes 3, 4, 7, and 8) or immobilized GST-Hr fusion protein. Input was 2.85% of the amount used in the coprecipitation reactions. radiolabeled GR. As illustrated in Fig. 2C, hVDR association with immobilized Hr is readily detected and is not affected, on average, by the presence of 1,25(OH) 2 D 3 . TR, a known corepressor target for Hr (30), also binds significantly to immobilized Hr under these conditions (data not shown). In contrast, no association is observed between GR and Hr (Fig. 2C). Thus, Hr appears to be relatively selective in its affinity for VDR compared with RXR␣ ( Fig. 2A) and GR (Fig. 2C), consistent with the functional results in Fig. 1.
Having demonstrated in vitro interaction of Hr and VDR, we next assessed whether Hr can interact with VDR in a more physiologic context using protein extracts from mammalian cells. Extracts prepared from cells transfected with VDR and Hr expression plasmids, both individually and together (Hr ϩ VDR), were used for immunoprecipitation with a VDR-specific monoclonal antibody, followed by Western analysis with Hrspecific antiserum. When VDR is immunoprecipitated from the extract containing both VDR and Hr, Hr is detected by Western analysis (Fig. 3, right side of upper panel). Therefore, Hr coprecipitates with VDR, suggesting that these proteins interact in the cell. Consistent with the results of GST pull down assays (Fig. 2), coimmunoprecipitation of Hr⅐VDR was not influenced by 1,25(OH) 2 D 3 (Fig. 3). Western analysis of the same blots with the VDR-specific antibody verified that VDR is expressed and immunoprecipitated (Fig. 3, lower panel).
Hr Interacts with the VDR Ligand Binding Domain-To map the region of VDR required for interaction with Hr, the binding of VDR truncation mutants was compared with full-length hVDR using CoIP. Fig. 4A reveals that both full-length hVDR and the VDR ligand binding domain (LBD; ⌬1-88) coimmunoprecipitate with Hr, whereas the hVDR DNA binding domain (DBD; ⌬134) does not associate with Hr under these conditions. Further dissection of the hVDR LBD was performed by CoIP of a series of C-terminally truncated hVDRs (Fig. 4B). CoIP of the ⌬403 and ⌬304 hVDR mutants is similar to full-length VDR, whereas C-terminal deletion to amino acid 201 severely compromises interaction between hVDR and Hr. Independent investigation of the potential VDR docking site for Hr using GST-Hr pull down experiments yielded similar results (Fig.  4C). Association with Hr is slightly reduced in the ⌬403 hVDR truncation lacking both helices 11 and 12 but intensified in the ⌬304 hVDR truncation that exposes helices 3, 4, 5, and 6 at the new C terminus. Interaction with Hr is dramatically reduced in the ⌬202 truncation, and the ⌬134 truncation (DBD) lacking the entire LBD except for helix 1 displays little or no Hr association (Fig. 4C).
The results of VDR deletion analyses (Fig. 4) are consistent with GST pull down data (Fig. 2B) showing that the AF-2/helix 12 C terminus is not required for VDR⅐Hr interaction. Based on both CoIP and GST pull down analyses (summarized in schematic form in Fig. 4D), the Hr docking site on hVDR lies between residues 202 and 303 in the central portion of the LBD, although the domain between hVDR amino acids 134 and 201 cannot be completely excluded. Analogous to the targeting of the SMRT corepressor to the helix 3-4 region of peroxisome proliferator activated receptor ␣ (54), Hr may interact with a surface of hVDR corresponding to repressive interfaces of helices 3 and 4 (55). Additional mutagenesis studies will be required to pinpoint the key hVDR amino acids in the helix 3-6 region that contact Hr and to determine whether they coincide positionally with SMRT docking sites in peroxisome proliferator activated receptor ␣ or differ in the case of the novel Hr corepressor.
VDR Interacts with a C-terminal Domain of Hr-To map the region of Hr that is required for interaction with VDR, a series of rHr deletion mutants was tested by CoIP with VDR (Fig. 5A). VDR bound to the C-terminal half of Hr (amino acids 568 -1207), the region shown previously to interact with TR (30) and ROR␣ (31). A series of smaller deletion mutants within this region were tested, and the minimal domain of Hr that was able to interact with VDR consisted of amino acids 750 -864. This region includes hydrophobic motifs involved in both TR and ROR␣ interaction (see Fig. 5C).
We next investigated the ability of wild-type Hr and Hr deletion mutants to repress transcriptional activation by VDR via VDREs in their biological context. The CYP24 gene is normally induced by 1,25(OH) 2 D 3 in keratinocytes (56); therefore we used a reporter gene with the human CYP24 promoter linked to a luciferase gene (45). The results (Fig. 5B) demonstrate that Hr is a potent repressor of VDR transactivation via a natural VDRE, with transcriptional activity in the presence of 1,25(OH) 2 D 3 reduced 9.5-fold by Hr. Thus, Hr can repress VDR-induced transcription from the promoter of a gene that is actually regulated by VDR in cells that participate in hair cycle regulation. Although the present experiments (Fig. 5B) were performed in COS cells, preliminary studies indicate that Hr also suppresses CYP24 promoter activity in cotransfected human keratinocytes (data not shown).
The ability of Hr deletion mutants to repress VDR-mediated transactivation was then determined using the natural VDRresponsive reporter gene (Fig. 5B). The C-terminal half of Hr (568 -1207) was able to repress transcription by VDR, consistent with its ability to bind VDR and the presence of two previously mapped repression domains (30). The N-terminal half of the protein (31-568) and a central fragment (450 -730), which do not bind specifically to VDR via CoIP analysis (data not shown), do not significantly repress transcription (Fig. 5B). Smaller C-terminal Hr fragments were investigated, and the FIG. 3. Coimmunoprecipitation of Hr and VDR. Protein extracts prepared from COS-1 cells transfected with expression vectors for Hr, VDR, or both (Hr ϩ VDR) were used for immunoprecipitation with VDR-specific (␣-VDR) or nonspecific (IgG) antibodies. Interaction was analyzed in the absence and presence (ϩvitD) of 10 Ϫ6 M 1,25(OH) 2 D 3 . Upper panel, Hr detected by Western analysis with Hrspecific antisera. Hr is detected only when VDR is immunoprecipitated from an extract containing both Hr and VDR. Lower panel, Western analysis of the same blot detecting VDR with monoclonal antibody 9A7␥. In, 3% of extract used for immunoprecipitation. Sizes of molecular mass markers (kDa) are indicated. data in Fig. 5B show that within this region only the 750 -864 sequence of Hr is capable of significant VDR repression. Strikingly, this minimal 750 -864 region of rHr that retains repressor activity corresponds to the smallest domain shown to interact with VDR.
Functional analysis of Hr deletion mutants shows a clear correlation between VDR binding and relative repressive activity (summarized in Fig. 5C). This may be explained in part by the functional organization of Hr, in which repression and receptor interaction domains are closely linked (30). Wild-type Hr was the most potent repressor (9.5-fold), whereas mutants 568 -1207 (3.0-fold), 750 -1084 (1.6-fold), and 750 -864 (2.6fold) repress to a lesser extent, possibly because they do not contain the full complement of repression domains. Repression may also be linked to the number and avidity of receptor interaction domains, as the minimal 750 -864 region of Hr that displays suppressive activity also possesses two motifs that mediate interaction with TR or ROR␣ (Fig. 5C). Interestingly, the same two ROR␣ and TR docking sites, in concert with repressive domain RD2, occur in all Hr deletants that repress VDR. Further analysis using site-directed mutagenesis will be required to determine whether the interaction of VDR with Hr is mediated by a combination of hydrophobic motifs responsible for ROR␣ and TR interaction or involves other residues in and outside the 750 -864 region of rHr.
Endogenous Hr and VDR Colocalize in Hair Follicles-VDR expression has been demonstrated in human and mouse skin, most intensely in hair root sheaths (57), keratinocytes (58), and the outer shaft of the hair follicle (59). Similarly, hr expression has been reported in hair follicles (including inner root sheath and matrix cells) and in the interfollicular epidermis (27,60). To determine whether the interaction between Hr and VDR that we have defined biochemically potentially occurs in vivo, we assessed whether hr and VDR are coexpressed in skin. Expression of hr and VDR mRNAs was detected by in situ hybridization of neonatal mouse skin with hr-and VDR-specific cRNA probes. Fig. 6A (upper left panel) shows that hr expression is detected throughout the hair follicle, including matrix cells, as well as inner and outer root sheath cells. VDR mRNA is also significantly expressed throughout the hair follicle (Fig.  6A, lower left panel), with expression detected in a subset of matrix cells and the most concentrated signal in peripheral, likely outer root sheath, cells. Comparison of hr and VDR expression in adjacent sections reveals significant cellular overlap between the mRNAs for these two nuclear proteins (Fig. 6B). Thus, hr and VDR mRNAs are often present in the same cells, suggesting that the translation and subsequent interaction of Hr and VDR proteins likely occurs in vivo.

CONCLUSIONS
In summary, a physical and functional interaction between VDR and Hr, two gene products essential for normal hair cycling in mammals, has been demonstrated. The physiological significance of VDR-Hr interaction is not yet clear, but our   FIG. 4. Hr association is mediated by a distinct region within  the VDR ligand binding domain. A, deletion mutants of hVDR consisting of the DNA binding (⌬134) and ligand binding (⌬1-88) domains were tested for interaction with Hr by CoIP. Protein extracts prepared from COS-1 cells cotransfected with expression vectors for Hr, and VDR derivatives were used for immunoprecipitation with VDRspecific (␣-VDR) or nonspecific (IgG) antibodies. Upper panel, Hr detected by Western analysis with Hr-specific antisera. Lower panel, Western analysis of the same blot detecting VDR. In, 3% of extract used for immunoprecipitation. Sizes of molecular mass markers (kDa) are indicated. B, C-terminal truncation mutants of hVDR tested for interaction with Hr by CoIP. Protein extracts prepared from COS-1 cells cotransfected with expression vectors for Hr and the indicated hVDR truncations were used for immunoprecipitation with VDR-specific (␣-VDR) or nonspecific (IgG) antibodies. The designation for hVDR mu-tants indicates the amino acid residue at which protein is truncated by replacement with a stop codon (see schematic in D). Upper panel, Hr detected by Western analysis with Hr-specific antisera. Lower panel, Western analysis of the same blot detecting VDR. In, 3% of extract used for immunoprecipitation. C, GST-Hr pull down assay employing a series of hVDR truncations. None of the hVDR truncated mutants bound significantly to GST-control beads (data not shown). Input (left) was 2.85% of the amount used in the pull down reactions. D, schematic representation of hVDR with the DBD and LBD indicated, along with the position of helices 3-6 (H3-6) and 11-12 (H11/12) based on the hVDR LBD x-ray crystal structure (55). Truncation mutants tested in A-C are illustrated below the full-length hVDR, and the right column summarizes the Hr binding properties of each fragment as assayed by CoIP and/or pull down methods. results reveal the convergence of two nuclear proteins that, as a complex, may impinge upon the molecular circuitry that controls the hair cycle in skin and perhaps biological events in other tissues.
We have shown that Hr is a potent repressor of VDR-mediated transcription (Fig. 1). Repression is manifest on both synthetic VDRE-reporters (Fig. 1) and on a natural VDR-responsive promoter from the CYP24 gene (Fig. 5B). Hr represses both VDR-directed basal transcription and transcriptional activation by VDR liganded with 1,25(OH) 2 D 3 or the bile acid LCA. The ability of Hr to suppress transcriptional activation by ligand-bound VDR demonstrates for the first time that a corepressor (Hr) can antagonize transcriptional activation by a native, ligand-bound nuclear receptor. Hr had been shown previously to mediate transcriptional repression by unliganded TR and to inhibit transcriptional activation by the orphan nuclear receptor ROR. Thus, the present results add VDR to the spectrum of nuclear receptors influenced by Hr and support the proposal that Hr is a member of a novel class of nuclear receptor corepressors (30).
Although only minimal evidence exists to show that unliganded VDR shares the ability of TR and RAR to repress transcription (61,62), the current data suggest that Hr functions to attenuate basal transcription driven by VDR. Hr expression resulted in a suppression of basal transcription mediated by VDR⅐RXR by an average of 46% in ten independent experiments (see Figs. 1 and 5) (data not shown). This observation indicates that Hr-VDR association resembles Hr-TR interaction, as Hr mediates repression by unliganded TR, playing a role similar to that of N-CoR and SMRT. However, VDR is a much weaker basal suppressor than TR (62), likely because unliganded TR also associates strongly with SMRT/N-CoR. The first column on the right summarizes CoIP results in A (and data not shown), with ϩ representing interaction with VDR, and Ϫ representing no specific interaction. The second column on the right lists repressive activity as -fold repression (Fold Repress.) by each Hr fragment relative to VDR in the presence of 1,25(OH) 2 D 3 (Ϫ represents no statistically significant repression).
Hr-VDR interaction also shares some similarities with the association of Hr with ROR. Hr was shown previously (31) to inhibit both constitutive transcriptional activation by ROR and retinoic acid-inducible activation of a hybrid RAR containing the ROR AF-2. Similar to Hr-ROR interaction (31), Hr can suppress transcriptional activation by liganded VDR, but Hr⅐VDR association occurs independently of the AF-2/helix 12 domain (see Figs. 2 and 4), whereas the AF-2 of ROR is an important determinant of Hr interaction specificity (31). Interestingly, a reported natural mutation in the hVDR AF-2/helix that compromises coactivator binding and results in rickets does not elicit alopecia (18), implying that VDR function in hair cycling does not require transcriptional activation via the AF-2/helix 12 coactivator binding motif. This genetic insight is consistent with the potential relevance of Hr⅐VDR association in controlling the hair cycle, because Hr binds to VDR in an AF-2-independent manner.
The robust repression of VDR-mediated transcriptional activation by Hr (see Fig. 1 and Fig. 5B) also suggests that the function of the Hr⅐VDR complex is biologically relevant. The current findings provide biochemical support for the extensive genetic evidence consistent with the role of both proteins in regulating the hair cycle. Mice with mutations in the hr gene have a normal first coat of hair that does not re-grow after it is shed (63). More severe alleles exhibit wrinkling of the skin, in addition to hair loss (63). Similarly, VDR-null mice and patients with loss-of-function mutations in hVDR also have initial hair growth that subsequently fails (13)(14)(15). Although the skin and hair cycling phenotypes of inactivated alleles of hr and VDR are similar, VDR mutants do not exhibit the severe wrinkling phenotype found in some hr alleles. In addition, the role of Hr in hair cycling may include TR. TR is expressed in most cells of the hair follicle, and patients that suffer from thyroid hormone deficiency frequently experience hair thinning or loss (64). Thus, a network of interactions between Hr and VDR, and perhaps TR, may exist in the skin to drive progression of the hair cycle.
Notably, a subset of natural mutations in hVDR, specifically in the ligand binding domain (including alteration of a 1,25(OH) 2 D 3 contact residue), elicit rickets without causing alopecia (15)(16)(17). In addition, neither vitamin D deficiency nor inactivation of the enzyme that produces 1,25(OH) 2 D 3 in mice (51,52) or humans (65) results in alopecia. Because insufficient vitamin D does not cause hair loss, then presumably it is not simply lack of transcriptional activation by 1,25(OH) 2 D 3 -liganded VDR that results in alopecia. Instead, signaling via unliganded VDR or VDR bound to an uncharacterized alternative ligand analogous to the bile acid LCA, may be required for regulation of the hair cycle. Consequently, Hr-mediated repression by unliganded VDR and/or Hr inhibition of non-vitamin D ligand-induced transcriptional activation by VDR may be critical for proper hair cycling.
An important question raised by the current study is the identity of the downstream target genes that are regulated by the Hr⅐VDR⅐RXR␣ ternary complex. This repressive complex may influence the signal transduction cascade that elicits tonic inhibition of the hair cycle. The hair follicle displays cyclic activity, with periods of resting (telogen), active growth and hair shaft generation (anagen), and apoptosis-driven regression (catagen). A tonic inhibitor of hair growth has been hypothesized to exist in telogen skin (66), and it is neutralization or suppression of this putative tonic inhibitor that is thought to trigger the telogen to anagen transition. Proteins that affect the telogen to anagen transition are candidates for regulation by VDR, as the transition from telogen to anagen is abrogated in VDR-null mice (67). Sonic hedgehog has been shown to stimulate the telogen-to-anagen transition (68) and acts downstream of ␤-catenin. ␤-Catenin is required for fate decisions of stem cells to form follicular rather than epidermal keratinocytes (69), and conditional ablation of ␤-catenin in the skin results in alopecia (69). Therefore, Sonic hedgehog, ␤-catenin, and other genes whose expression is affected during hair cycling are potential candidates for Hr⅐VDR⅐RXR␣ action.
Another potential target for Hr⅐VDR⅐RXR␣ repression of the genetic cascade that inhibits the hair cycle is parathyroid hormone-related peptide (PTHrP). PTHrP regulates the rate of keratinocyte differentiation (70), and PTHrP antagonists accelerate anagen and enhance hair growth in mice (71). Thus, PTHrP may be a tonic inhibitor of hair growth. Interestingly, 1,25(OH) 2 D 3 -VDR is known to repress PTHrP expression via negative VDREs (72,73). The mechanism(s) whereby liganded VDR bound to negative VDREs results in down-regulation of target gene expression by VDR has not been elucidated (3). Hr is a candidate for participation in this function, as it interacts with VDR in the presence of its known ligands and has been shown to mediate repression via interaction with histone deacetylases (29,30).
In addition to its role in hair cycling, the Hr⅐VDR complex may act in mammalian central nervous system development, particularly in the cerebellum. Previous work showing that Hr can regulate the transcriptional activity of TR and ROR␣ led to the hypothesis that development of this brain region is controlled by a dynamic interplay between Hr and these nuclear receptors (31,74). Thyroid hormone, presumably acting through its cognate receptor, has major actions in central nerv-FIG. 6. The hr and VDR genes are coexpressed in cells of the hair follicle. A, in situ hybridization of dorsal mouse skin with hr-and VDR-specific cRNA probes. Longitudinal sections through the hair follicles were hybridized with hr-specific probes (top panels) or VDRspecific probes (lower panels). Antisense (left panels) represents genespecific signal; sense (right panels) shows background with control probe. B, overlapping expression of hr and VDR in hair follicle cells. In situ hybridization of consecutive serial sections with antisense hr-and VDR-specific probes is shown; sense control probe did not produce a significant signal (data not shown). Expression of both hr and VDR is detected in outer root sheath and a subset of matrix cells. Magnification, ϫ200.
ous system development (75,76), and in fact induces expression of Hr in neonatal cerebellum (28,29,75). The orphan receptor ROR␣ (31) is directly involved in regulating cerebellar development, as ROR␣-null (staggerer) mice display Purkinje cell depletion and cerebellar ataxia (74,77). In the cerebellum, both Hr (31) and VDR (78) are abundantly expressed in granule cells; therefore, in granule cells the presumed protein complex may influence cerebellar development and/or function. This proposed action of VDR in cerebellum could complement or parallel the known functions of TR and/or ROR␣ to promote cell migration, differentiation, and synaptogenesis (75,76).
In conclusion, based upon biochemical/functional interaction experiments, insight gleaned from the phenotypes of genetic mutations, and the cellular expression patterns of Hr and various nuclear receptors, the Hr corepressor could represent a significant coregulator that orchestrates the actions of VDR, TR, and ROR in developmental gene expression.