Evidence for 1,25-Dihydroxyvitamin D3-independent Transactivation by the Vitamin D Receptor

The vitamin D endocrine system plays critical although poorly understood roles in skin. Vitamin D receptor (VDR) knock-out (VDRKO) mice have defects in hair follicle cycling and keratinocyte proliferation leading to epidermal thickening, dermal cyst formation, and alopecia. Surprisingly, skin defects are not apparent in mice lacking 25-hydroxyvitamin D 1α-hydroxylase, the enzyme required for 1,25-dihydroxyvitamin D3 (1,25(OH)2D3) hormone biosynthesis. These disparate phenotypes indicate that VDR effects in skin are independent of the 1,25(OH)2D3 ligand. However, cellular or molecular data supporting this hypothesis are lacking. Here, we show transcriptional activation of the vitamin D-responsive 24-hydroxylase promoter by VDR in primary keratinocytes that is independent of the 1,25(OH)2D3 ligand. This activity required functional vitamin D-responsive promoter elements as well as an intact VDR DNA binding domain and thus could not be distinguished from 1,25(OH)2D3-dependent VDR transactivation. The 1,25(OH)2D3-independent activation of VDR was also observed in keratinocytes from 1α-hydroxylase knock-out mice, indicating that it is not due to endogenous 1,25(OH)2D3 production. Mammalian two-hybrid studies showed strong, 1,25(OH)2D3-independent interaction between VDR and retinoid X receptors in primary keratinocytes, indicating that enhanced heterodimerization of these receptors was involved. Indeed, this 1,25(OH)2D3-independent VDR-RXR heterodimerization was sufficient to drive transactivation by VDR(L233S), an inactive ligand binding mutant of VDR that was previously shown to rescue the skin phenotype of VDR null mice. Cumulatively, these studies support the concept that transactivation by VDR in keratinocytes may be uncoupled from the 1,25(OH)2D3 ligand.

min D endocrine system, are mediated through the vitamin D receptor (VDR). VDR is a member of the steroid hormone receptor/nuclear receptor superfamily. VDR and other steroid hormone receptors are classic, ligand-activated transcriptional regulatory proteins that control RNA polymerase II-directed transcription in response to a steroidal or other lipid ligand. The binding of 1,25(OH) 2 D 3 induces a conformational change in VDR that is essential for high affinity interaction between VDR and its obligate heterodimeric partner, the retinoid X receptor (RXR) (1,2). The liganded VDR-RXR dimer binds with high affinity and selectivity to vitamin D response elements (VDREs) in the promoter regions of target genes (3). Coactivators, such as steroid receptor coactivator family members (4), VDR-interacting proteins (5), and NCoA-62/Ski-interacting protein (6) interact with and enhance the transcriptional activity of the VDR/RXR dimer. Steroid receptor coactivator proteins enhance VDR/RXR activity by recruiting histone acetyltransferases that remodel chromatin (7,8). In addition, steroid receptor coactivators have intrinsic histone acetyltransferase activities (9). VDR-interacting proteins interact with RNA polymerase II and recruit other general transcription factors (10), and NCoA-62/Ski-interacting protein couples transcription to RNA splicing machinery (11). Cumulatively, these coactivators function through distinct mechanisms to promote VDR/RXRmediated transcription.
Synthesis of 1,25(OH) 2 D 3 is a tightly controlled, multistep process involving several cytochrome P450 enzymes. The skin is a major source of vitamin D, where it is synthesized from its 7-dehydrocholesterol precursor upon exposure to UV light (12). Vitamin D is also obtained from the diet. Bioactivation of vitamin D is accomplished through sequential hydroxylations (12). The first takes place in the liver, where vitamin D-25-hydroxylase catalyzes C-25 hydroxylation of vitamin D 3 , producing 25-hydroxyvitamin D 3 , the major circulating form of vitamin D in mammals (13). The second hydroxylation at C-1 of 25-hydroxyvitamin D 3 occurs primarily in renal tissues and is catalyzed by 25-hydroxyvitamin D 1␣-hydroxylase (1␣OHase) (14). The product is 1,25(OH) 2 D 3 , the hormonally active form of vitamin D. One of the best characterized target genes of 1,25(OH) 2 D 3 -and VDR-induced transcription is 25-hydroxyvitamin D 3 -24-hydroxylase (24OHase), a ubiquitously expressed enzyme that targets vitamin D metabolites for degradation and elimination (15). 1,25(OH) 2 D 3 -induced expression of 24OHase and the resulting catabolism of the ligand is an important feedback component of the vitamin D endocrine system.
Considerable insight into the physiological roles of the vitamin D endocrine system has come from the ongoing analysis of knock-out mouse models in which either the receptor (VDR) or the ligand (1,25(OH) 2 D 3 ) has been eliminated. Four independent strains of mice have been generated with targeted deletions of portions of the VDR DNA binding domain (VDRKO mice) (16 -19). As expected, these mice recapitulate the major symptoms of vitamin D deficiency and those of humans with inactivating mutations in VDR (hereditary vitamin D-resistant rickets), namely hypocalcemia, hyperparathyroidism, and the consequent skeletal anomalies, including rickets and osteomalacia. Mice lacking 1␣OHase (i.e. the ligand knock-out mouse) exhibit mineral homeostasis phenotypes that are similar to VDRKO mice and that are also seen in a human genetic disorder termed pseudovitamin D deficiency rickets, a rare disease caused by mutations in the 1␣OHase enzyme (20 -22). Interestingly, a diet containing high amounts of calcium, phosphorus, and lactose corrects the undermineralized skeletal phenotype in both the VDRKO and 1␣OHaseKO mice (23,24), suggesting that the skeletal effects of 1,25(OH) 2 D 3 and VDR may be indirect.
An intriguing difference between the VDRKO and 1␣OHaseKO mice is apparent in skin tissue. VDRKO mice develop their first coat of hair after birth, but hair follicle cycling ceases thereafter, leading to alopecia. Hair follicles degenerate into dermal cysts and utriculi (23). In addition, the in vivo interfollicular keratinocyte proliferation rate of VDRKO mice is approximately twice the rate of wild type mice, and consequently, the epidermal thickness of VDRKO mice is double that of wild type controls (25). These anomalies are not corrected by a high calcium diet, indicating that they are due to direct actions of VDR in keratinocytes. In contrast, the skin and hair of 1␣OHaseKO mice appears to be morphologically normal and exhibits only minor differences in expression of skin differentiation markers compared with wild type controls (26). Thus, in animals lacking the 1,25(OH) 2 D 3 ligand, hair follicle cycling and keratinocyte proliferation are normal. These mouse models accurately mimic human skin disorders resulting from vitamin D endocrine system defects. Many, but not all, hereditary vitamin D-resistant rickets patients that have inactivating mutations in VDR exhibit degenerated hair follicles and alopecia (27), whereas vitamin D deficiency and pseudovitamin D deficiency rickets cause no obvious skin or hair defects.
The striking phenotypic differences in the skin and hair of VDRKO and 1␣OHaseKO mice strongly suggest that a novel, 1,25(OH) 2 D 3 -independent mechanism regulates VDR activity in the skin. This hypothesis is also supported by dietary deficiency studies in which long term maintenance of animals on vitamin D-deficient diets does not impact hair follicle cycling (28). Experiments in transgenic mice have localized the critical actions of VDR to epidermal keratinocytes, since targeted expression of VDR in this cell type is sufficient to rescue the skin phenotype of VDRKO mice (29). In addition, expression of a ligand binding mutant of VDR (L233S) rescues hair follicle cycling, further indicating that maintenance of skin homeostasis requires VDR but not 1,25(OH) 2 D 3 (30). These in vivo studies have led to the hypothesis that the actions of VDR in skin are independent of the 1,25(OH) 2 D 3 ligand (30). However, there is little experimental evidence directly supporting this hypothesis or illuminating the molecular mechanisms governing VDR transcriptional activity in the skin. In this report, we provide evidence for VDR-dependent and 1,25(OH) 2 D 3 -independent expression of 24-hydroxylase, a well characterized VDR target gene in primary keratinocytes. The ligand-independent activity of VDR in keratinocytes appears to stem from strong interactions between VDR and RXR, which occur in the absence of 1,25(OH) 2 D 3 .
Cell Culture and Transient Reporter Assays-Primary human keratinocytes and primary human fibroblasts were isolated from the foreskins of newborn donors. Skin was floated in keratinocyte-SFM (Invitrogen) with 5 units/ml Dispase II (Roche Applied Science), 10 units/ml penicillin G, 10 g/ml streptomycin, and 1.0 g/ml Fungizone (Invitrogen) at 4°C. Keratinocytes were isolated from the epidermal layer as described previously (34). Fibroblasts were isolated from the dermis. The dermis was minced with sterile scissors and digested in 0.35% collagenase I (Worthington) dissolved in DMEM (Invitrogen), shaking at 37°C for 45 min. The cell suspension was centrifuged and resuspended in DMEM supplemented with 10% fetal bovine serum, 10 units/ml penicillin G, and 10 g/ml streptomycin (Invitrogen) and then maintained at 37°C and 5% CO 2 .
HaCaT keratinocytes were maintained in DMEM supplemented with 10% fetal bovine serum, 10 units/ml penicillin G, and 10 g/ml streptomycin (Invitrogen) at 37°C and 5% CO 2 . For transient transfection assays, primary human fibroblasts and HaCaT cells were grown in DMEM with 5% charcoalstripped calf bovine serum, 10 units/ml penicillin G, and 10 g/ml streptomycin. Primary human keratinocytes were grown in keratinocyte-SFM with 5 ng/ml epidermal growth factor, 30 g/ml bovine pituitary extract, 10 units/ml penicillin G, and 10 g/ml streptomycin (Invitrogen) at 37°C and 5% CO 2 . HaCaT cells, primary human fibroblasts in passages 2-4, or primary human keratinocytes in passages 3-5 were seeded for reporter assays at a density of 3.2 ϫ 10 4 cells/well in 12-well plates and transfected the next day with Fugene 6 (Roche Applied Science) diluted in Opti-MEM (Invitrogen). 18 h later, the cells were washed with phosphate-buffered saline, and media were replenished and supplemented with 1,25(OH) 2 D 3 , 9-cis-retinoic acid, ethanol vehicle, or no additive as indicated. Following a 24-h incubation, cells were harvested in 150 l of passive lysis buffer (Promega). Cell extracts were analyzed for luciferase activity using the dual luciferase assay system (Promega) according to the manufacturer's instructions and a LMax luminometer (Molecular Devices, Sunnyvale, CA).
Primary mouse keratinocytes were isolated from 1-3-dayold pups essentially as described (35). Briefly, the skin was removed from pups and floated overnight in keratinocyte-SFM (Invitrogen) with 5 units/ml Dispase II (Roche Applied Science), 10 units/ml penicillin G, 10 g/ml streptomycin, and 1.0 g/ml Fungizone (Invitrogen) at 4°C. The epidermis was separated from the dermis and trypsinized in 0.05% trypsin (Invitrogen) in phosphate-buffered saline containing 0.5 mM EDTA at 37°C for 4 min. Trypsin was inactivated by adding DMEM with 10% fetal bovine serum, and cells were centrifuged at 500 ϫ g for 5 min at 25°C. Cells were resuspended in growth media (keratinocyte-SFM with 5 ng/ml epidermal growth factor, 30 g/ml bovine pituitary extract, 10 units/ml penicillin G, and 10 g/ml streptomycin (Invitrogen)) and seeded in collagen I-coated plates. Keratinocytes were allowed to attach to plates overnight at 37°C and 5% CO 2 , and then cells were washed and media were replaced. Some keratinocytes were transfected with Fugene and Opti-MEM, as noted, and treated the next morning with 1,25(OH) 2 D 3 or ethanol as indicated. Mouse dermal fibroblasts were isolated and cultured using the same conditions as primary human fibroblasts.
Reverse Transcription-PCR-Total RNA was isolated from mouse keratinocyte and fibroblast cultures using RNA Bee (Tel-test, Friendswood, TX) according to the manufacturer's instructions. RNA was purified using RNeasy Plus Mini columns (Qiagen, Valencia, CA). Five g of purified RNA was reverse-transcribed using the Superscript III First-Strand synthesis system (Invitrogen) with the oligo(dT) 20 primer. 24OHase transcripts were amplified by PCR using PCR Master Mix (Promega) and the primers 5Ј-ggagtccatgaggcttacc-3Ј and 5Ј-gtcttcgctagagcccagc-3Ј, producing a product of 136 nucleotides. ␤-Actin transcripts were amplified with the primers 5Јaagatcctcaccgagcgcg-3Ј and 5Ј-tggatgccacaggactccat-3Ј, producing a product of 254 nucleotides. Cycling conditions were 94°C for 45 s, 58°C for 45 s, and 72°C for 45 s. PCR products were visualized on 2% agarose gels stained with ethidium bromide.

Evidence for 1,25(OH) 2 D 3 -independent Activation of the 24-Hydroxylase Promoter in Primary Human Keratinocytes-
Similarities in the undermineralized phenotypes of the VDRKO and 1␣OHaseKO mice highlight the essential roles that VDR and 1,25(OH) 2 D 3 play in maintaining appropriate calcium and phosphate absorption in the intestine, but their prominent phenotypic differences in the skin provide compelling evidence for an uncoupling of VDR and its 1,25(OH) 2 D 3 ligand in keratinocyte biology. However, direct evidence for 1,25(OH) 2 D 3 -independent actions of VDR in keratinocytes has not been forthcoming. To determine whether keratinocyte-derived VDR exhibits transcriptional activity that is independent of 1,25(OH) 2 D 3 , we transfected two highly responsive, VDR-activated promoter-reporter constructs into primary human keratinocytes, HaCaT immortalized keratinocytes, and primary human dermal fibroblasts, three cell types that have been previously characterized as VDR-and 1,25(OH) 2 D 3 -responsive. The VDRE 4 -TATA-Luc reporter contains four tandem copies of the VDRE from the rat osteocalcin gene fused to the minimal TATA box from the same gene, and it represents one of the most powerful and sensitive VDR-activated constructs of its kind in transfection assays. As shown in Fig. 1A, it was activated 3-fold by 1,25(OH) 2 D 3 in the HaCaT cell line, but ectopic expression of VDR led to a profound 100-fold activation by 1,25(OH) 2 D 3 . In primary human fibroblasts, 1,25(OH) 2 D 3 activated the consensus reporter 11-fold, and expression of VDR increased activation to 45-fold (Fig. 1B). In contrast, 1,25(OH) 2 D 3 activated this reporter 1.6-fold in primary human keratinocytes, and surprisingly, ectopic expression of VDR increased the 1,25(OH) 2 D 3 response by a comparatively modest factor of 3 (Fig. 1C). Western blot analysis of VDR protein levels in primary human and HaCaT keratinocytes showed comparable expression of the SG5-VDR vector that could not account for these observed differences in reporter gene activation (data not shown). These data indicate that 1,25(OH) 2 D 3dependent activation of VDR, as assessed with the synthetic VDRE 4 -TATA-luc reporter, is suppressed in primary human keratinocytes when compared with primary dermal fibroblasts and HaCaT keratinocytes.
A second reporter tested contains the natural promoter and 1200 base pairs of upstream flanking sequence from the 24OHase gene, one of the most highly induced genes that is stimulated by 1,25(OH) 2 D 3 and VDR in a variety of cell types. 1,25(OH) 2 D 3 activated this reporter 23-fold in HaCaT cells (Fig. 1A), 8-fold in primary human fibroblasts (Fig. 1B), and 6-fold in primary human keratinocytes (Fig. 1C). Expression of additional VDR increased -fold activation by 1,25(OH) 2 D 3 to 150-fold in HaCaT keratinocytes and 35-fold in primary human fibroblasts (Fig. 1, A and B, respectively). Surprisingly, when VDR was expressed in the primary human keratinocytes in the absence of the 1,25(OH) 2 D 3 ligand, we observed a 7-fold increase in 24OHase promoter reporter gene expression compared with cells transfected with the SG5 vector control (Fig.  1C, open bars). This 1,25(OH) 2 D 3 -independent activity of VDR was not evident in the HaCaT immortalized keratinocytes and primary human fibroblasts (Fig. 1, A and B, open bars). These data indicate that in primary human keratinocytes, VDR possesses significant transactivation potential in the absence of its requisite 1,25(OH) 2 D 3 ligand.

VDR-induced Activation of the 24-Hydroxylase Promoter Is Not Due to Endogenous Production of 1,25(OH) 2 D 3 by the 1␣-Hydroxylase Enzyme-
The skin is the only organ of the human body capable of synthesizing 1,25(OH) 2 D 3 from its 7-dehydrocholesterol precursor, and 1,25(OH) 2 D 3 is known to be produced from 25-OHD 3 in cultured primary human keratinocytes (36). 1␣-Hydroxylase catalyzes 1,25(OH) 2 D 3 synthesis, and some evidence exists for its expression in keratinocytes (37). To test the possibility that VDR activation may result from endogenous production of 1,25(OH) 2 D 3 , the 24-hydroxylase promoter was transfected into wild type and 1␣OHaseKO primary mouse keratinocytes. VDR activated the 24-hydroxylase reporter to similar extents in both wild type and 1␣OHaseKO keratinocytes ( Fig. 2A). These data indicate that local production of 1,25(OH) 2 D 3 by the CYP27 gene product is not the basis for the activity observed in these studies.
1␣-Hydroxylase is the only known enzyme that catalyzes 1,25(OH) 2 D 3 synthesis (38), but the possibility exists that 1,25(OH) 2 D 3 is synthesized by an alternate pathway in keratinocytes. However, mammalian one-hybrid analysis using Gal4-VDR and a Gal4-responsive reporter (Gal4 5 -TATA-Luc) gene argues against this possibility. Gal4-VDR is a highly sensitive 1,25(OH) 2 D 3 -activated fusion protein that consists of the VDR ligand binding domain (amino acids 93-427) and the heterologous Gal4 DNA binding domain. 1,25(OH) 2 D 3 strongly activated this Gal4-VDR fusion protein in primary human keratinocytes (Fig. 2B). However, the Gal4-VDR fusion displayed no inherent transactivation potential in the absence of ligand (compare Gal4 and Gal4-VDR; open bars in Fig. 2B), indicating that little or no endogenous 1,25(OH) 2 D 3 was present in this system. Although the keratinocyte medium is serum-free and lacks vitamin D derivatives, these observations formally rule out the contribution of active vitamin D metabolites in the media and in the cells. Cumulatively, these data indicate that endogenous 1,25(OH) 2 D 3 is unlikely to account for the ligandindependent activity observed in these studies.
VDR Promotes Basal Expression of 24OHase Transcripts in Wild Type and 1␣OHaseKO Keratinocytes-To determine whether the native 24OHase gene was similarly regulated in a 1,25(OH) 2 D 3 -independent fashion by VDR, steady-state mRNA levels for 24OHase were examined in primary keratinocytes obtained from wild type, VDRKO, and 1␣OHaseKO mice. As illustrated in Fig. 3A, wild type keratinocytes expressed detectable levels of 24OHase transcripts in the absence of ligand, and 1,25(OH) 2 D 3 increased 24OHase transcript levels. In contrast, 24OHase expression was not detected in VDRKO mouse keratinocytes in either the absence or presence of 10 nM 1,25(OH) 2 D 3 (compare lanes 3 and 1 in Fig. 3A). Thus, VDR ablation leads to a loss of basal 24OHase gene expression in the absence of 1,25(OH) 2 D 3 . To confirm that endogenous production of 1,25(OH) 2 D 3 was not responsible for basal 24OHase expression in wild type keratinocytes, 24OHase transcripts were measured in 1␣OHaseKO keratinocytes. Importantly, basal expression of 24OHase transcripts was unaffected in 1␣OHaseKO keratinocytes compared with wild type keratinocytes (Fig. 3A, lanes 1 and 5). These data indicate that VDR, but not the 1,25(OH) 2 D 3 ligand, is required for basal expression of the 24OHase gene.
This effect was selective for keratinocytes, since no differences in basal expression of 24OHase transcripts were evident in primary dermal fibroblasts obtained from wild type and VDRKO mice (Fig. 3B, lanes 1 and 3). Cumulatively, these data, together with the transfection data in Figs. 1 and 2, provide strong evidence for 1,25(OH) 2 D 3 -independent transactivation of the 24OHase promoter by VDR that is selective for primary keratinocytes obtained from humans and mice.
Activation of the 24OHase Promoter by VDR Requires a Functional DNA Binding Domain and Functional Vitamin D Response Elements in the Promoter-1,25(OH) 2 D 3 -liganded VDR activates the 24OHase promoter via direct binding to two vitamin D response elements (VDREs), located at nucleotides Ϫ293 to Ϫ273 (the distal element) and nucleotides Ϫ172 to Ϫ143 (the proximal element), relative to the transcriptional start site. It is unclear whether these VDREs also mediate the atypical, 1,25(OH) 2 D 3 -independent induction of this promoter by VDR or if other regions are involved. The role of these VDREs in 1,25(OH) 2 D 3 -independent activation of the 24-hydroxylase promoter by VDR was examined by mutating one or both of the response elements and measuring the response of the mutant promoter to VDR expression. As shown in Fig. 4A, treatment with the 1,25(OH) 2 D 3 ligand stimulated the wild FIGURE 2. VDR-induced activation of the 24-hydroxylase promoter is not due to endogenous production of 1,25(OH) 2 D 3 by the 1␣-hydroxylase enzyme. A, primary mouse keratinocytes were isolated from newborn 1␣-hydroxylase wild type (1␣OHaseWT) and knock-out (1␣OHaseKO) mice. One day after plating, cells were transfected with the human 24OHase promoter (Ϫ1200 to ϩ120 bp)-Luc, phRG-TK, and SG5-VDR or SG5 vector control. Following a 24-h treatment with ethanol vehicle, cell extracts were prepared and analyzed for luciferase expression. B, primary human keratinocytes were transfected with Gal4 5 -TATA-Luc, phRG-TK, and the indicated SG5-Gal4 plasmid. Following a 24-h treatment with ethanol vehicle or 10 Ϫ8 M 1,25(OH) 2 D 3 , cell extracts were prepared and analyzed for luciferase expression. Data are graphed as the mean of duplicate samples Ϯ S.D. type 24OHase promoter 14-fold. As expected, mutating either response element significantly reduced the 1,25(OH) 2 D 3 responsiveness to 2-3-fold (Fig. 4A), and mutation of both elements abolished all responsiveness to 1,25(OH) 2 D 3 . In the absence of added 1,25(OH) 2 D 3 , VDR stimulated the wild type 24OHase promoter 8-fold, whereas mutation of the proximal and distal response elements reduced promoter responsiveness to 4-and 2-fold, respectively. Inactivation of both elements rendered the promoter virtually unresponsive to 1,25(OH) 2 D 3 -independent VDR activity in this assay (Fig. 4A). We also compared the abilities of wild type VDR and DNA binding domain mutants of the VDR (H35Q and R73Q) to activate the 24OHase promoter in the absence or presence of 1,25(OH) 2 D 3 in primary human keratinocytes (Fig. 4B). Wild type VDR activated the reporter 7-fold in the absence of ligand, whereas the H35Q and R73Q were unable to activate the 24-hydroxylase promoter in the absence of ligand. Expression of the H35Q and R73Q VDR mutants reduced the 1,25(OH) 2 D 3 responsiveness due to endogenous VDR in this cell type (Fig. 4B). Together, these data show that both vitamin D response elements of the 24-hydroxylase promoter and an intact DNA binding domain on the VDR are required for 1,25(OH) 2 D 3 -dependent and 1,25(OH) 2 D 3 -independent activity of VDR. Furthermore, these data suggest that both canonical 1,25(OH) 2 D 3 -induced and 1,25(OH) 2 D 3independent actions of VDR probably involve direct binding of VDR to VDREs in the promoters of target genes.
Cell-selective, 1,25(OH) 2 D 3 -independent Heterodimerization of VDR and RXRs in Primary Human Keratinocytes-The requirement of VDR heterodimerization with RXR in 1,25(OH) 2 D 3 -activated transcription is well accepted, but the role of RXR and its 9-cis-retinoic acid ligand in 1,25(OH) 2 D 3independent transactivation by VDR is untested. Therefore, a mammalian two-hybrid system was used to monitor VDR-RXR heterodimer formation in primary human keratinocytes, primary human fibroblasts, and HaCaT immortalized keratinocytes. The ligand binding domain of VDR was expressed as a fusion protein with the yeast VP16 activation domain, and the ligand binding domains of three RXR isoforms were expressed as fusion proteins with the yeast Gal4 DNA domain. None of the Gal4 fusion proteins activated the Gal4-responsive reporter gene alone (data not shown). However, coexpression of VP16-VDR and Gal4-RXR␣ in primary human keratinocytes resulted in a strong activation of the reporter (more than 1500-fold) in the absence of 1,25(OH) 2 D 3 (Fig. 5A). In contrast, significant VDR-RXR heterodimerization was not apparent in the absence of 1,25(OH) 2 D 3 in HaCaT keratinocytes (Fig. 5A) or in primary human dermal fibroblasts (Fig. 5B). Indeed, 1,25(OH) 2 D 3 was required to promote strong interactions between VDR and RXR in HaCaT keratinocytes and in primary fibroblasts. These data show an increased propensity of VDR to heterodimerize with RXR in the absence of 1,25(OH) 2 D 3 selectively in primary keratinocytes compared with primary fibroblasts and HaCaT immortalized keratinocytes.
VDR did not display preferential interaction with the various RXR isoforms, since all three isoforms of RXR interacted with VDR to a similar degree in primary human keratinocytes (Fig.  5C, open bars). Moreover, treatment with 9-cis-retinoic acid, the ligand for RXRs, strongly reduced VDR interaction with RXR in this system (Fig. 5C). In addition, 9-cis-retinoic acid reduced the ability of VDR to activate the 24-hydroxylase reporter in the absence of 1,25(OH) 2 D 3 (data not shown), further supporting the hypothesis that strong VDR/RXR heterodimerization drives the activity of "unliganded" VDR. Together, these data suggest that the 1,25(OH) 2 D 3 -independent activity of VDR in primary keratinocytes may arise, in part, from enhanced VDR/RXR heterodimerization.

RXR Heterodimerization Drives Transactivation by the VDR L233S Ligand Binding Mutant in Primary Human
Keratinocytes-An obvious correlation exists between ligandindependent VDR/RXR heterodimerization and 1,25(OH) 2 D 3 - independent 24OHase promoter activation by VDR in primary keratinocytes. These transcriptional data suggest a possible mechanism to explain the in vivo activity of the VDR L233S ligand binding mutant to rescue the skin and hair phenotype of the VDRKO mouse; enhanced heterodimerization of this ligand binding mutant with RXR may impart a transactivation potential to VDR L233S in the absence of 1,25(OH) 2 D 3 binding. To test this possibility, we examined the ability of VDR L233S to interact with RXR and to drive gene activation in the absence of 1,25(OH) 2 D 3 . As shown in Fig. 6A, VDR L233S retained the ability to interact with RXR in the absence of added ligand although to a lesser extent than wild type VDR. As expected, interaction of VDR L233S with RXR was not affected by 1,25(OH) 2 D 3 treatment. Thus, to determine if the VDR L233S/ RXR mutant heterodimer expressed activity on the 24-hydroxylase promoter, wild type and mutant VDRs were expressed in keratinocytes with and without added RXR. Wild type VDR activated the 24-hydroxylase promoter ϳ4-fold in the absence of ectopic RXR expression, whereas the ligand binding mutant (L233S) and DNA binding mutant (H35Q) had no effect (Fig.  6B). Expression of RXR alone produced a modest response. Coexpression of wild type VDR and RXR synergistically activated the promoter 11-fold compared with the SG5 vector alone, and expression of VDR L233S with RXR resulted in 8-fold activation compared with the empty vector control (Fig.  6B). Importantly, the VDR H35Q DNA-binding mutant displayed little or no activity in the absence or presence of RXR expression. Therefore, DNA binding, but not 1,25(OH) 2 D 3 binding, is required for VDR-induced activation of the 24-hydroxylase promoter in keratinocytes. Collectively, these data show that VDR interacts strongly with its heterodimer transcriptional partner RXR in the absence of added ligand in primary human keratinocytes, and the heterodimer pair drives transcription of the 24-hydroxylase target gene.

DISCUSSION
A functional vitamin D endocrine system is critical for maintaining appropriate skin homeostasis, although the functional roles of VDR and its 1,25(OH) 2 D 3 ligand in skin remain largely unexplored. Genetic mouse models of the vitamin D endocrine system indicate that VDR and 1,25(OH) 2 D 3 exert both overlapping and distinct actions in epidermal keratinocytes. Gene expression profiling of head-and neck-derived squamous cell carcinoma cell lines and primary human keratinocytes has identified a myriad of genes regulated by 1,25(OH) 2 D 3 and its low calcemic synthetic analogs (39,40). Notable among these genes are cell cycle regulators, such as the growth arrest and DNA damage (gadd45␣) gene, cell adhesion proteins, such as integrin ␣7B, and members of the peptidylarginine deiminase, kallikrein, and serine proteinase inhibitor families that regulate expression of genes involved in formation of the cornified enve- lope and desquamation during differentiation. This antiproliferative and prodifferentiative response provides a rational basis for the clinical effectiveness of 1,25(OH) 2 D 3 and synthetic analogs in the treatment of psoriasis, a hyperproliferative disorder of the skin, and in reducing squamous cell carcinoma growth (41). Typically, 1,25(OH) 2 D 3 exerts its physiological actions by binding to VDR, inducing a conformational change that results in the recruitment of RXR and a variety of transcriptional cofactors. The 1,25(OH) 2 D 3 -induced changes in keratinocyte gene expression described above are presumably mediated by VDR in this manner. However, the disparate skin phenotypes of vita-min D receptor null mice (VDRKO) and 1,25(OH) 2 D 3 null mice (1␣OHaseKO) indicate that VDR may also regulate biologically relevant target genes in keratinocytes in the absence of the 1,25(OH) 2 D 3 ligand. Our studies provide direct evidence supporting this concept. Here, we show that VDR activates transcription of the 24-hydroxylase promoter in a 1,25(OH) 2 D 3independent manner in primary human and mouse keratinocytes. 24-hydroxylase is a well characterized 1,25(OH) 2 D 3 -responsive gene in many cell types, but our data show that VDR also activates this promoter independently of 1,25(OH) 2 D 3 in keratinocyte cell culture.
The 1,25(OH) 2 D 3 -independent activity of VDR is unique to primary keratinocytes and selective for the 24-hydroxylase promoter. Although 1,25(OH) 2 D 3 activated the 24-hydroxylase promoter in primary keratinocytes, primary fibroblasts, and immortalized (HaCaT) human keratinocytes, "unliganded" VDR strongly activated the 24-hydroxylase promoter only in primary keratinocytes (Fig. 1). These in vitro data are consistent with in vivo evidence that defective hair cycling in VDRKO mice is due to lack of VDR in the keratinocyte, not the dermal fibroblast (29). Importantly, ligand-independent activation of a promoter by VDR has not been observed in other cell types, suggesting that a cell type-specific modulatory protein, alternate ligand, or signaling pathway exists in normal keratinocytes that activates VDR. Primary human keratinocytes proliferate only in a specially formulated low calcium medium, whereas HaCaT keratinocytes and primary fibroblasts must be grown in a normocalcemic medium. However, 1,25(OH) 2 D 3 -independent actions of VDR do not require low calcium conditions. In fact, switching primary keratinocytes to medium containing 2 mM Ca 2ϩ , in addition to initiating terminal differentiation, modestly increases the activation of the 24OHase promoter by VDR compared with low calcium medium (data not shown). In addition to cell type selectivity, the 1,25(OH) 2 D 3 -independent actions of VDR display promoter selectivity. In primary keratinocytes, expression of VDR activates the natural 24-hydroxylase promoter but fails to activate a synthetic 1,25(OH) 2 D 3 -responsive reporter gene composed of tandem VDREs from the osteocalcin gene promoter (Fig. 1). The VDRE nucleotide sequences differ between osteocalcin and 24-hydroxylase. Although both sequences fit the DR-3 consensus sequence known for high affinity VDR-DNA interaction, it is possible that VDR binds the 24OHase VDREs with a higher affinity in the absence of its 1,25(OH) 2 D 3 ligand. Alternatively, the sequences adjacent to the VDREs in the 24-hydroxylase promoter may contain binding sites for other transcription factors that act in concert with VDR to effectively activate the promoter. Additional mechanistic studies will be required to address these possibilities and perhaps lead to identification of additional target genes of "unliganded" VDR.
The skin is the only organ of the body that synthesizes 1,25(OH) 2 D 3 from its precursor, 7-dehydrocholesterol. Therefore, endogenous synthesis of 1,25(OH) 2 D 3 has the potential to activate VDR in keratinocytes in the absence of 1,25(OH) 2 D 3 addition. However, this possibility is minimized, based on recent evidence obtained from reporter gene knock-in mice, substituting lacZ for the 1␣OHase gene, which indicates that 1␣OHase is not expressed in the skin or in keratinocyte cul- tures, even under conditions of severe hypocalcemia (42). Our data indicate that the VDR activity in primary keratinocytes is not due to local production of 1,25(OH) 2 D 3 by the canonical pathway or by an alternate pathway. VDR activated the 24-hydroxylase promoter equally well in wild type and 1␣OHaseKO keratinocytes, indicating that the 1␣OHase enzyme putatively expressed in keratinocytes is dispensable for VDR activity (Fig.  2A). These in vitro studies support in vivo observations in the 1␣OHase null mice. The lack of systemic 1,25(OH) 2 D 3 and the absence of a major skin phenotype in these animals indicates that 1␣OHase expression is unnecessary for select VDR actions in the skin. A large number of cytochrome P450 enzymes are expressed in keratinocytes. It is possible that another enzyme catalyzes 1␣-hydroxylation of vitamin D metabolites, thus constituting an alternate pathway for 1,25(OH) 2 D 3 synthesis. The studies presented here do not formally rule out local production of 1,25(OH) 2 D 3 by an alternate pathway. However, the absence of transactivation by Gal4-VDR in the absence of 1,25(OH) 2 D 3 treatment tends to argue against this possibility (Fig. 2B). Gal-VDR contains the entire ligand binding domain of VDR and is exquisitely sensitive to 1,25(OH) 2 D 3 -induced transactivation. However, in primary keratinocytes, this fusion protein expresses no activity in the absence of ligand treatment, indicating that local 1,25(OH) 2 D 3 production is insufficient to result in VDR activation in the primary human keratinocyte cell system. Collectively, our data argue against local 1,25(OH) 2 D 3 production as being responsible for the ligand-independent, VDR-induced activation of the 24-hydroxylase promoter demonstrated in primary keratinocyte cultures.
At present, we are unable to mechanistically distinguish between 1,25(OH) 2 D 3 -independent, VDR-induced activation of the 24OHase promoter and the 1,25(OH) 2 D 3 -dependent canonical pathway. Both processes need a functional DNA binding domain, intact VDREs in the promoter, and appropriate heterodimerization with RXR. This suggests that the activity of "unliganded" VDR in primary keratinocytes may be due to binding of an alternate ligand or phosphorylation by a signaling pathway that activates VDR in the absence of 1,25(OH) 2 D 3 . Indeed, we use the phrase "unliganded" VDR in quotes to reflect the possibility that an alternate, keratinocyte-selective ligand or activating modification of VDR may be involved in this process. Structural studies show that the VDR ligand binding domain is much larger than most other nuclear receptors (43). Thus, alternate ligands may have a greater propensity to bind VDR, perhaps via distinct residues in the VDR ligand binding domain compared with 1,25(OH) 2 D 3 binding. To date, 1,25(OH) 2 D 3 and lithocholic acid metabolites are the only known endogenous ligands to bind and activate VDR (44). As a secondary bile acid, lithocholic acid is unlikely to be biologically relevant in keratinocytes. Indeed, we observe little or no stimulation of the 24OHase promoter by lithocholic acid in primary human keratinocytes (data not shown). Alternatively, some endogenous metabolites of 1,25(OH) 2 D 3 express bioactivity mediated through VDR (45). As yet, no endogenous keratinocyte-specific ligands have been identified for VDR. However, it is formally possible that an alternate keratinocyte-specific ligand may activate VDR, either wild type or L223S, by binding in a novel way to the receptor.
Alternate mechanisms exist to regulate nuclear receptor function in addition to ligand binding. All of the hormone receptors, including the progesterone, estrogen, androgen, and glucocorticoid, and nonsteroidal receptors, such as the retinoic acid receptor, RXR, peroxisome proliferator-activated receptor, and VDR have multiple phosphorylation sites (reviewed in Ref. 46). It is possible that a keratinocyte-selective signaling pathway phosphorylates VDR or RXR, resulting in increased transcriptional activity from the heterodimer. Phosphorylation regulates the activity of VDR in different ways, depending on the kinase and phosphorylation site. Phosphorylation at serine 208 by casein kinase II increases 1,25(OH) 2 D 3 -responsiveness of VDR (47), whereas phosphorylation of serine 51 by protein kinase C inhibits VDR activity (32). Treatment with okadaic acid, a protein phosphatase inhibitor, increases VDR transactivation by 1,25(OH) 2 D 3 and increases "unliganded" transactivation of select promoters as well (48). Notably, both ligand binding and phosphorylation have been shown to activate mutant VDR proteins that are resistant to activation by 1,25(OH) 2 D 3 , although few data exist regarding these events in keratinocytes. Phosphorylation of VDR or RXR may be the impetus for the increased heterodimerization observed in these studies. Indeed, okadaic acid has been shown to increase VDR interaction with VDR-interacting protein 205 coactivators (48). Further research into keratinocyte-selective phosphorylation events may yield insight into the 1,25(OH) 2 D 3 -independent activity of VDR as well as potentially discovering ways of modulating VDR activity in a skin-specific manner. Skin-selective manipulation of VDR activity would prove useful in the treatment of hyperproliferative diseases, such as psoriasis and skin cancer, without the patient incurring the hypercalcemic side effects of systemic vitamin D endocrine system activation.
Strong in vivo support for the 1,25(OH) 2 D 3 -independent actions of VDR in skin comes from the rescue of the VDRKO skin phenotype through keratinocyte-selective, transgenic expression of the VDR L233S ligand binding mutant. The L233S mutant rescues the skin phenotype despite its lack of 1,25(OH) 2 D 3 binding and 1,25(OH) 2 D 3 -induced transactivation (30). The mechanisms are unclear, but our data show that "unliganded" VDR has an enhanced capacity to interact with RXR in primary human keratinocytes compared with immortalized keratinocytes (Fig. 5A) and primary fibroblasts (Fig. 5B). The VDR L233S ligand binding mutant retained the ability to interact with RXR (Fig. 6A). This suggests a potential mechanism in which enhanced heterodimer formation of VDR L233S with RXR drives transactivation of select VDR target genes in the keratinocytes of the transgenically rescued VDRKO mice, thus preserving normal keratinocyte function in this animal model. Indeed, we provide cellular data supporting this mechanism in which expression of RXR rescued the "unliganded" activity of the VDR L233S mutant (Fig. 6B). In addition to its role as a heterodimer partner for VDR, RXR also regulates the transcription of certain genes as a homodimer in the presence of its ligand, 9-cis-retinoic acid (49). VDR/RXR interaction in primary human keratinocytes was disrupted by 9-cis-retinoic acid (Fig. 5C). Thus, retinoid ligands negatively impact VDR transactivation, indicating that endogenous retinoids may not play a major role in the 1,25(OH) 2 D 3 -independent VDR activity observed in these studies. Indeed, 9-cis-retinoic acid has been shown to divert RXR toward homodimer-regulated, rather than 1,25(OH) 2 D 3and VDR-regulated, transcriptional pathways (33).
This paper provides strong evidence that select transcriptional effects of VDR may be uncoupled from the 1,25(OH) 2 D 3 ligand in keratinocytes. We show that VDR activates the 24-hydroxylase promoter independently of 1,25(OH) 2 D 3 in primary keratinocytes. Endogenously produced 1,25(OH) 2 D 3 is not the stimulus for VDR activation. Rather, strong interaction between RXR and "unliganded" VDR, as measured by mammalian two-hybrid assays and shown functionally in reporter gene assays, drives the transcription of the 24-hydroxylase promoter. These data suggest a mechanism for 1,25(OH) 2 D 3 -independent actions of VDR in the skin, and they provide a potential explanation for the rescue of the VDRKO skin phenotype by the VDR L233S ligand binding mutant. A more detailed dissection of the pathways leading to 1,25(OH) 2 D 3 -independent VDR transactivation may lead to the development of new strategies to target skin pathologies such as alopecia, cancer, and psoriasis.