1α,25(OH)2-3-Epi-Vitamin D3, a Natural Physiological Metabolite of Vitamin D3: Its Synthesis, Biological Activity and Crystal Structure with Its Receptor

Background The 1α,25-dihydroxy-3-epi-vitamin-D3 (1α,25(OH)2-3-epi-D3), a natural metabolite of the seco-steroid vitamin D3, exerts its biological activity through binding to its cognate vitamin D nuclear receptor (VDR), a ligand dependent transcription regulator. In vivo action of 1α,25(OH)2-3-epi-D3 is tissue-specific and exhibits lowest calcemic effect compared to that induced by 1α,25(OH)2D3. To further unveil the structural mechanism and structure-activity relationships of 1α,25(OH)2-3-epi-D3 and its receptor complex, we characterized some of its in vitro biological properties and solved its crystal structure complexed with human VDR ligand-binding domain (LBD). Methodology/Principal Findings In the present study, we report the more effective synthesis with fewer steps that provides higher yield of the 3-epimer of the 1α,25(OH)2D3. We solved the crystal structure of its complex with the human VDR-LBD and found that this natural metabolite displays specific adaptation of the ligand-binding pocket, as the 3-epimer maintains the number of hydrogen bonds by an alternative water-mediated interaction to compensate the abolished interaction with Ser278. In addition, the biological activity of the 1α,25(OH)2-3-epi-D3 in primary human keratinocytes and biochemical properties are comparable to 1α,25(OH)2D3. Conclusions/Significance The physiological role of this pathway as the specific biological action of the 3-epimer remains unclear. However, its high metabolic stability together with its significant biologic activity makes this natural metabolite an interesting ligand for clinical applications. Our new findings contribute to a better understanding at molecular level how natural metabolites of 1α,25(OH)2D3 lead to significant activity in biological systems and we conclude that the C3-epimerization pathway produces an active metabolite with similar biochemical and biological properties to those of the 1α,25(OH)2D3.


Introduction
The 1a,25-dihydroxyvitamin D 3 (1a,25(OH) 2 D 3 or calcitriol), is the most active form of vitamin D 3 and mediates its pleiotropic effects through VDR activation, which heterodimerizes with retinoid X receptor (RXR).VDR-induced genomic action results in growth inhibition of lymphomas, breast or prostate primary tumor cells, renal osteodystrophy, osteoporosis, psoriasis or autoimmune diseases [1,2].Consequently, VDR is an exquisite therapeutic target to combat human metabolic diseases and uncontrolled cell proliferation in many tissues [3][4][5].In addition 1a,25(OH) 2 D 3 is a key regulator of calcium and phosphate homeostasis and bone metabolism but its intrinsic hypercalcemic effect prevents its use in therapeutical applications [6].
The production of 1a,25(OH) 2 -3-epi-D 3 is initiated via A-ring C3-epimerization (Figure 1), where the C-3 hydroxyl moiety is Figure 1.Proposed pathway of the 1a,25(OH) 2 -3-epi-D 3 production [18].The reaction is initiated via A-ring C3-epimerization, where the C-3 hydroxyl moiety is changed from b to its diastereomer a. Two distinct pathways may be employed by cells to generate 1a,25(OH) 2 -3-epi-D 3 .The first, more likely used pathway, starts with dehydrogenation catalyzed by yet unidentified enzyme leading to a keto-intermediate, which is converted most probably by the same enzyme to the final product 1a,25(OH) 2 -3-epi-D 3 .The second one uses dehydration and a subsequent hydroxylation at C-3 a position.doi:10.1371/journal.pone.0018124.g001changed from position b to its diastereomer a.The enzymes responsible for the C3-epimerization have not been identified to present date.It was also proposed by Reddy et al. that this pathway might be used for metabolites that resist inactivation through C-24 oxidation [18] a phenomenon well characterized in the bile acid metabolism where the reaction is catalyzed by bile acid hydroxysteroid dehydrogenase [27].This pathway plays also a major role in the activation and/or inactivation of steroid hormones such as androgens [28].
Despite a lower binding affinity than calcitriol, 1a,25(OH) 2 -3epi-D 3 possess significant biological activity only in specific tissues where it is produced [29].The transcriptional response of the 1a,25(OH) 2 -3-epi-D 3 compound varies for different VDR-regulated genes in different tissues.For instance, it shows lower activation of osteocalcin gene and lower HL60 differentiation [30] but has almost equipotent activity to 1a,25(OH) 2 D 3 in inhibiting cellular proliferation in keratinocytes [19] and in suppressing parathyroid secretion in bovine parathyroid cells [25].These in vitro properties associated with its low calcemic activity [31,32] assign potential therapeutic interest to this compound.
To further unveil, the structural mechanism and structureactivity relationships of 1a,25(OH) 2 -3-epi-D 3 /hVDR-LBD complex, we describe a more effective synthetic route to the synthesis of 1a,25(OH) 2 -3-epi-D 3 , some of its in vitro biological properties and the crystal structure of its complex with hVDR LBD.
Further, we hypothesized about the absence of the 1a,25(OH) 2 -3-epi-D 3 signaling in HL60 cellular model and thus turned to characterize some of the biological properties of 1a,25(OH) 2 -3epi-D 3 in cells where it is produced [20,21].We first determined the kinetics of CYP27B1and CYP24A1-catalyzed oxidation by monitoring the major lipophilic metabolites arising from a single pulse of 3   S2 and Methods S2.Since the 1a,25(OH) 2 -3-epi-D 3 is present steadily up to 5 h in rather high concentration in this tissue and the fact that the primary genomic effects of hVDR ligands are exerted in first hours suggested that primary keratinocytes may be a good cellular model to investigate the anti-proliferative actions of this metabolite.Therefore we determined the dose-dependent anti-proliferative effects of 1a,25(OH) 2 D 3 and 1a,25(OH) 2 -3-epi-D 3 using 3 H-thymidine incorporation assay (Figure 2D), and found that the IC 50 values for 1a,25(OH) 2 D 3 and 1a,25(OH) 2 -3epi-D 3 were highly similar (41.4 and 66.1 nM, respectively) with no significant statistical difference (using unpaired t-test p = 0.074).In addition, we correlated the course of the anti-proliferation data between the two epimers and find a strong correlation (Pearson r = 0.940**) between them indicating the similar anti-proliferative activity for 1a,25(OH) 2 D 3 and 1a,25(OH) 2 -3-epi-D 3 .The antiproliferative effects of the two metabolites are comparable and they are in close agreement with our coactivator peptide recruitment (Figure 2A) and reporter gene assays (Figure 2B).Although in this assay we cannot totally exclude the possibility that the potential cell type specific difference in the function of the two natural ligands may be partly due to the accumulation of 1a,25(OH) 2 -3-epi-D 3 in 1a,25(OH) 2 D 3 treated cells with C3epimerization ability leading to additive effect, we consider this accumulation process as a naturally occurring in vivo physiological event when 1a,25(OH)2D3 is present in these cells.
Overall structure of the hVDR complexed to 1a,25(OH) 2 -3-epi-D 3 The mechanistic action of analogues of 1a,25(OH) 2 D 3 is unveiled by the determination at high resolution of the crystal structure of their complexes with the VDR LBD [49][50][51][52][53].We solved the crystal structure of the complex formed by 1a,25(OH) 2 -3-epi-D 3 with the hVDR LBD mutant previously used to solve the structures of hVDR LBD in complexes with 1a,25(OH) 2 D 3 or several synthetic agonists [49][50][51][52][53][54].The crystal was isomorphous and the structure of hVDR LBD bound to 1a,25(OH) 2 -3-epi-D 3 determined at a resolution of 1.9 A ˚(PDB ID: 3A78).The crystallographic data are summarized in Table S1.After refinement of the protein alone, the map showed an unambiguous electron density where to fit the ligand (Figure 3B).The complex formed by the hVDR LBD bound to 1a,25(OH) 2 -3-epi-D 3 adopts the canonical active conformation as described in all previously reported agonist-bound nuclear receptor LBDs (Figure 3A).The conformation of the activation helix 12 is strictly maintained.When compared to the structure of hVDR LBD-1a,25(OH) 2 D 3 complex, the atomic coordinates of hVDR LBD bound to 1a,25(OH) 2 -3-epi-D 3 show very small root-mean-square deviation (RMSD) of 0.17 A ˚for all 255 Ca atoms, reflecting its high structural homology.

Conformation of the 3a-epimer in the ligand-binding pocket of hVDR
The 1a,25(OH) 2 -3-epi-D 3 , is buried in the predominantly hydrophobic ligand-binding pocket (LBP) of the VDR.The conformation of the 3-epi-hydroxyl group does not modify the Aring chair conformation of the ligand.Furthermore the seco B-, C-, D-rings, and the aliphatic side chain present conformations similar to those observed with 1a,25(OH) 2 D 3 (Figure 3B and C).
In the complexes of hVDR LBD bound to 1a,25(OH) 2 D 3 versus 1a,25(OH) 2 -3-epi-D 3 , the distance between the C1-OH and the C25-OH groups varies from 13.1 A ˚to 12.7 A ˚and between the C3-OH and the C25-OH groups from 15.3 A ˚to 16.0A ˚, respectively.The adaptation of the hVDR's LBP to different ligands can be described with the differential changes in the volumes of LBPs and bound ligands.In addition the parameter representing the % of LBP filling with ligand can provide useful information about the activity of ligand [55].All these parameters are summarized in Table 1.Although the two diastereomer have the same molecular weight and differ only in the position of the C3-OH group, the 1a,25(OH) 2 -3-epi-D 3 takes a slightly more compact conformation in the LBP.The graphical 0.2 A ˚mesh representation of the superimposed LPBs presented in Figure 4A and B show the surface area which is enlarged in case of 1a,25(OH) 2 D 3 (in green) and the one enlarged in case of 1a,25(OH) 2 -3-epi-D 3 bound hVDR LBP (in blue).This suggests that the hydrophobic residues lining the LBP are closer to the 3epimer and may compensate for the canonical hydrogen bonds.We observed a notable adaptation with the displacement of the side chain of the residue Tyr147 by 2.0 A ˚compared to the 1a,25(OH) 2 D 3 bound complex and the reorientation of the Glu277 side chain away from the 1a,25(OH) 2 -3-epi-D 3 due to the a-position of the C3-OH group (Figure 4B).These specific rearrangements lead to a more compact conformation resulting in a 5% decrease in the volume of the LBP compared to 1a,25(OH) 2 D 3 .

Specific interactions of the 1a,25(OH) 2 -3-epi-D 3
The hydrophobic and electrostatic interactions between the receptor and the ligand are similar between the two structures except around the C3-OH group.While the C1-OH and C25-OH display the canonical hydrogen bonds, the 3-epi-OH of 1a,25(OH) 2 -3-epi-D 3 interacts through hydrogen bonding only with Tyr143 instead of interacting with both Tyr143 and Ser278 (Figure 5).A significant feature of the 1a,25(OH) 2 -3-epi-D 3 is the compensation of the loss of interaction with Ser278 by a watermediated hydrogen bond with the water molecule H 2 O1 (W1 in [50]).As such, the position of water H 2 O1 is moved 0.7 A ˚towards 1a,25(OH) 2 -3-epi-D 3 , thereby facilitating the specific watermediated contacts.This water molecule is part of the network connecting another water molecule H 2 O2 to Arg274.All these water molecules are also present in the 1a,25(OH) 2 D 3 -hVDR complex [50].The C3-OH hydrogen bonds have longer distances in the 3-epimer (3.0A ˚instead of 2.8 A ˚with Tyr143 and 3.1 A ẘith the water molecule instead of 2.9 A ˚with Ser278).A study on the mutations of the residues forming the hydrogen bonds with the hydroxyl groups of 1a,25(OH) 2 -3-epi-D 3 revealed that mutated residues contacting the 3-hydroxyl group are the less affected in term of activity.Mutation of Ser278 in Ala may result in a lower binding affinity for 1a,25(OH) 2 D 3 [56] while showing a similar potency to activate the transcription [57,56].Due to the shift of the side chain of Tyr147, a hydrophobic interaction with this residue is lost in the 3-epimer structure.These structural data agree well with the lower binding affinity of this compound for VDR and to its induced biological activity.
In conclusion, we described a more effective synthesis of the highly stable 1a,25(OH) 2 -3-epi-D 3 , a natural metabolite.We have  The absolute values in A ˚3 as well as relative values in reference to those of 1a,25(OH) 2 D 3 (100%) are indicated.From these values the percent filling of the LBP with ligands was also calculated.*and ** Connolly solvent accessible surfaces calculated by GRASP and Voidoo respectively The quality of the cubic grid spacing for the surface for both ligands and LBP = 0.5 A ˚. doi:10.1371/journal.pone.0018124.t001solved the crystal structure of hVDR LBD in complex with 1a,25(OH) 2 -3-epi-D 3 , which provides a mechanistic insight for the specific recognition of the two naturally occurring 3-epimers by hVDR.Indeed, the crystal structure reveals that the 3-epimer metabolite maintains the number of H-bonds by an alternative water-mediated interaction.In MCF-7 cells, the 1a,25(OH) 2 -3epi-D 3 on CYP24 gene promoter retains significant transcriptional activity.In addition, the anti-proliferative action of 1a,25(OH) 2 -3epi-D 3 is cell specific and the IC 50 values of 1a,25(OH) 2 D 3 and 1a,25(OH) 2 -3-epi-D 3 in primary keratinocytes are in the same nanomolar range.Therefore, we conclude that the C3-epimerization pathway produces an active metabolite with similar biochemical and biological properties to those of the 1a,25(OH) 2 D 3 .The physiological role of this pathway as the specific biological action of the 3-epimer remains unclear and needs further investigation.However, its high metabolic stability together with its significant biologic activity makes this natural metabolite an interesting ligand for clinical applications.Further study on its target specificity and selectivity is required to the design of selective analogues.Our new findings contribute to a better understanding at molecular level how natural metabolites of 1a,25(OH) 2 D 3 lead to significant activity in biological systems.

Ligands
1a,25(OH) 2 D 3 was purchased from Cayman Chemical (Tallinn, Estonia) and the synthesis of 1a,25(OH) 2 -3-epi-D 3 is described in more details in the Methods S1.Additional ligands and reference compounds are described in Methods S2.IUPAC rules were used for the name of the compounds.In addition to NMR spectra (summarized in Methods S1), HPLC analysis was used to determine the purity (.95%) of the vitamin D analogues.

Luciferase reporter gene construct
The fragment of the proximal promoter region (2414 to 264) of the human CYP24A1 gene was fused with the thymidine kinase promoter driving the firefly luciferase reporter gene [60].
Transient transfections and luciferase reporter gene assays MCF-7 cells were seeded into 24-well plates (100,000 cells/well) and grown overnight in phenol red-free Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% charcoalstripped fetal bovine serum (FCS) and 0.6 mg/ml insulin.Plasmid DNA containing liposomes were formed by incubating 40 ng of an expression vector for hVDR, 100 ng of reporter plasmid and 10 ng pEGF-C2 with Fugene 6 (Roche Diagnostics, Switzerland) transfection reagent according to the recommendation of the manufacturer for 15 min at room temperature.After dilution with 500 ml of phenol red-free DMEM, the liposomes were added to the cells.Phenol red-free DMEM supplemented with 500 ml of 20% charcoal-stripped FCS was added 4 h after transfection, in the presence of ligands or solvent.The cells were lysed 16 h after the onset of stimulation using reporter gene lysis buffer (Roche Diagnostics, Switzerland).The lysates were assayed for luciferase activity as recommended by the supplier (Perkin-Elmer, The Netherlands).The luciferase activities were normalized to GFP expression.Data represent one triplicate for which the mean and the S.D. of the mean was calculated.

Data analysis for dose response curves
A non-linear curve fit was performed for the AlphaScreen and reporter gene assay experimental dose response data and from sigmoidal dose response curve then the EC 50 values for the respective ligands were calculated using GraphPad Prism 4 (GraphPad Software Inc., San Diego, CA).The Student's unpaired t-test and Pearson correlation were performed with the SPSS software (SPSS Inc., version 14.0, Chicago, IL, USA).

Keratinocyte cell cultures
Normal human keratinocytes were isolated from fresh adult skin obtained from surgery and immediately transported to the laboratory under sterile conditions.Isolation and culture under serum-free conditions and without a feeder layer followed a modified protocol as used by Bikle et al [11].The isolated epidermis was incubated in a 0.25% trypsin solution for 45 min at 37uC.Thereafter, the cells were scraped off and put in 50 ml Hank's balanced salt solution (HBSS) containing 10% FCS to block further trypsin digestion and centrifuged at 2000 rpm/2 min.The resulting cell pellet was suspended in Keratinocyte Growth Medium (KGM, Clonetics Corp., San Diego), a defined serum-free medium at low (0.06 mM) calcium containing 0.1 ng/ml epidermal growth factor, 5 mg/ml insulin, 0.5 mg/ml hydrocortisone, bovine pituitary extract, antibiotics (gentamycin, amphothericin) gave the primary culture.After 24 h, the cells were incubated at 37uC in 95% air/5% CO 2 and the attached cells were washed and provided with fresh KGM medium.The culture medium was changed every other day and the cells were passaged when they reached 80-90% confluency (usually 6-10 days after plating).

Incubations of primary keratinocytes with 3 H-25(OH)D 3
Confluent human keratinocytes in 1 ml KGM and in 6-well plates were incubated in duplicates at 37uC with 20.6 nM 3 H-25(OH)D 3 (around 600 000 dpm/ml) for 1-23 h.Incubations were stopped with 1 ml methanol/well, the cells were scraped off, transferred to a test tube together with the supernatant and two washings (with 1 ml methanol and 0.8 ml water).Unmodified 3 H-25(OH)D 3 and most of the products were totally extracted from the combined solutions plus cell pellet according to the method of Bligh and Dyer [61] by three subsequent extractions with 2, 1 and 1 ml volumes of CHCl 3 at room temperature. 3H-activity in the CHCl 3 -phase, in the water and total 3 H-yield were determined.The combined CHCl 3 extracts were then evaporated under argon at 35uC, the residues dissolved in 0.4 ml ethanol and an aliquot (containing around 250 000 dpm 3 H-activity) subjected to HPLCanalysis (see Methods S2).

H-Thymidine incorporation (anti-proliferation assay in primary keratinocytes)
Keratinocytes (second passage) in 200 ml KGM (low calcium) were plated in 96-well plates at an initial density of 10 4 cells/well, kept 24 h at 37uC in an incubator with 95% air/5% CO 2 .Thereafter, the test compounds 1,25(OH) 2 D 3 or its 3-epimer were added in 1 ml ethanol to give final concentrations ranging from 0 to 100 nM, each condition in triplicates.After further 24 h, 50 ml 3 H-thymidine (1 mCi) were added and incubation continued for additional 7 h.Then, incubations were stopped by cell harvesting (Filtermate 196 Harvester, Packard-Canberra) and lysis: After removing the supernatant (see below), the adherent cells were released by 5 min treatment with 100 ml 0.125% trypsin in PBS at 37uC, harvested on a filterplate and washed 3 x with redistilled water.After drying the plates, their bottoms were sealed with a film and 50 ml scintillation cocktail (MicroScint O, Packard) were added.The whole plates were sealed with Packard Cover Film and 3 H-activity counted on a Microplate Scintillation Counter (TopCount, Packard Canberra).To check whether proliferative ( 3 H-thymidine incorporating) cells could have been shed off, the supernatants were soaked through a 96-well filterplate (Unifilter Plate GF/C) and 3 x washed with redistilled water: in all conditions, 3 H-activity was undetectable on these filterplates (in order to roughly assess cell numbers and check for substance related morphological changes/toxic effects, photographs were taken prior to compound addition and immediately before harvesting.)Data -used as means 6 SD -were normalized (incorporated 3 H-activity sample vs. blank) and analyzed using the GRAFIT Erithacus 4.0.19IC 50 software.

Protein purification and Crystallization
Purification and crystallization of the hVDR LBD complexed with 1a,25(OH) 2 -3-epi-D 3 were performed as previously described [49].The LBD of the hVDR (residues 118-427 D166-216) was cloned in pET-28b expression vector (Novagen) to obtain an Nterminal 6xHis fusion protein and was overproduced in E. coli BL21 (DE3) strain.Cells were grown in Luria Bertani medium and subsequently incubated for 6 h at 20uC with 1 mM isopropyl thiob-D-galactoside.The protein purification included a metal affinity chromatography step on a Co 2+ -chelating resin (Clontech).The 6xHis tag was removed by thrombin digestion overnight at 4uC, and the protein was further purified by gel filtration on a Superdex S200 16/60.The sample buffer prior to protein concentration contained 10 mM Tris, pH = 7.5, 100 mM NaCl, and 10 mM dithiothreitol.The protein was concentrated to 3.5 mg/ml and incubated in the presence of a 1.5-fold molar excess of ligand.The purity and homogeneity of the protein were assessed by SDS-PAGE.The protein crystals were obtained at 4uC by vapor diffusion method using crystals of hVDR LBD-1a,25(OH) 2 D 3 as microseeds.The reservoir solution contained 0.1 M MES and 1.4 M ammonium sulfate pH = 6.0.

X-Ray data collection and structure determination
The crystal was mounted in fiber loop and flash cooled in liquid nitrogen after cryoprotection with a solution containing the reservoir plus 30% glycerol and 2% polyethylene glycol 400.Data collection from a single frozen crystal was performed at 100 K on the beamline ID29 of the European Synchrotron Radiation Facility (Grenoble, France).The crystal belongs to the orthorhombic space group P2 1 2 1 2 1 with one monomer per asymmetric unit.Data were integrated and scaled using MOSFLM [62] (see statistics in Table S1).A rigid body refinement was used with the structure of the hVDR LBD complexed to 1a,25(OH) 2 D 3 as a starting model.Refinement involved iterative cycles of manual building and refinement calculations.The programs Refmac [63] and COOT [64] were used throughout structure determination and refinement.The omit map from the refined atomic model of hVDR LBD was used to fit the ligand to its electron density, shown in Figure 2A.Individual Batomic factors were refined isotropically.Solvent molecules were then placed according to unassigned peaks in the difference Fourier map.In the hVDR/1a,25(OH) 2 -3-epi-D 3 complex, refined at 1.9 A ˚with no s cutoff, the final model consists of residues 118-423 (D166-216), the ligand, two sulphate ions and 372 water molecules.According to PROCHECK [65] 92.6% of peptide lies in most favored regions and 7.4% in additional allowed regions.Data are summarized in Table S1.The volumes of the ligand-binding pockets and ligands were calculated as previously reported [49].

Supporting Information
H[26,27]-25(OH)D 3 at physiological concentration(20.6 nM).During the first two hours, we observed a rapid appearance of 1a,25(OH) 2 D 3 , from which at a slower rate the 3epimer was irreversibly formed.In total, some 60 independent incubation experiments were performed on the kinetics of 3 H[26,27]-25(OH)D 3 using primary keratinocytes from various donors and skin sites.In all experiments, highly comparable time course of 1a,25(OH) 2 D 3 and 1a,25(OH) 2 -3-epi-D 3 were recorded with 3-epimer exceeding 1a,25(OH) 2 D 3 after longer incubation as shown in Figure2Cand in the detailed HPLC analysis in Figure

Figure 4 .
Figure 4. Adaptability of the hVDR LBP upon 1a,25(OH) 2 -3-epi-D 3 binding.(A) The adaptation of the LBP is depicted by mesh representation of the superimposed LBP volumes calculated with Voidoo software.The green surface represent the LBP area where the 1a,25(OH) 2 D 3 bound pocket is larger.The blue area represents similar increase but for 1a,25(OH) 2 -3-epi-D 3 and the two main expanded regions are highlighted with red circles.(B) Adaptation of the residues Tyr147 and Glu277 in the LBP of the 1a,25(OH) 2 -3-epi-D 3 hVDR complex.The distances between the ligand-specific positions of the residues are displayed in A ˚. doi:10.1371/journal.pone.0018124.g004

Table 1 .
Volume of VDR ligands and their resulting LBPs.