1α,20S-Dihydroxyvitamin D3 Interacts with Vitamin D Receptor: Crystal Structure and Route of Chemical Synthesis

1α,20S-Dihydroxyvitamin D3 [1,20S(OH)2D3], a natural and bioactive vitamin D3 metabolite, was chemically synthesized for the first time. X-ray crystallography analysis of intermediate 15 confirmed its 1α-OH configuration. 1,20S(OH)2D3 interacts with the vitamin D receptor (VDR), with similar potency to its native ligand, 1α,25-dihydroxyvitamin D3 [1,25(OH)2D3] as illustrated by its ability to stimulate translocation of the VDR to the nucleus, stimulate VDRE-reporter activity, regulate VDR downstream genes (VDR, CYP24A1, TRPV6 and CYP27B1), and inhibit the production of inflammatory markers (IFNγ and IL1β). However, their co-crystal structures revealed differential molecular interactions of the 20S-OH moiety and the 25-OH moiety to the VDR, which may explain some differences in their biological activities. Furthermore, this study provides a synthetic route for the synthesis of 1,20S(OH)2D3 using the intermediate 1α,3β-diacetoxypregn-5-en-20-one (3), and provides a molecular and biological basis for the development of 1,20S(OH)2D3 and its analogs as potential therapeutic agents.

Pregnenolone acetate (4) has often been used as the starting material for 20S(OH)D 3 analogs 12-14 , in which 1α-hydroxylation was necessary to display potent stimulation of the VDR 12,13 . Owing to the lack of appropriate 1α-OH intermediates, the production of 1α-OH derivatives of 20S(OH)D 3 analogs was dependent on the purification of recombinant CYP27B1. The limited amount of 1α-OH derivatives that could be made was thus a hurdle for extensive biological testing. The production of 1α,3β-diacetoxypregn-5-en-20-one (3) in this report enables production of various analogs of 1,20S(OH) 2 D 3 for future studies.
We experienced inconsistent yields during the Birch reduction of epoxide 13 in the initial trials. In fact, the addition of NH 4 Cl (a quenching step) is the key to the success of this reaction. Quick addition (<10 min) of NH 4 Cl gave predominantly intermediate 14a, whereas slow addition (>2 h) afforded mainly the desired product 14. To our knowledge, 14a as a semi-reduced intermediate was obtained and characterized for the first time.
HPLC showed matched retention times of chemical and enzymatic 1,20S(OH) 2 D 3 . In addition to the UV spectra and NMR identification (Supplementary Information), the chemically synthesized 1,20S(OH) 2 D 3 was analysed by HPLC under two different solvent systems, either an acetonitrile in water gradient or a methanol in water gradient. We conclude that the chemically synthesized 1,20S(OH) 2 D 3 and the enzymatically produced counterpart are identical on the basis of their UV and NMR spectra, as well as their HPLC retention times (Fig. 4). Co-migration of the chemically and enzymatically synthesized 1,20S(OH) 2 D 3 was further confirmed by chromatography of a mixture of the two (see Supplementary Information).

Identification of 15 as having a 1α-OH by NMR analysis.
To identify the formation of the 1α-hydroxyl, the structure of intermediate 15 was characterized from its NMR spectra (Supplementary Information). The NOESY spectrum of 15 gave a strong NOE integral (0.42, see Supplementary Fig. S1) of 1H β to 19-CH 3 , using the NOE integral of 1H β to 2H β as an internal reference. In contrast, the NOE signal of 1H β to 2H α was not observed, suggesting the presence of 1α-OAc group in 15.

Confirmation of 15 by X-ray crystallographic analysis.
To confirm the structure of 15, crystals were produced in hexane for X-ray crystallographic analysis (Supplementary Information). The X-ray structure of 15 (CCDC code: 1527430, Fig. 3) confirmed its absolute structure as the desired product reported in the Fig. 3.  differential VDRE stimulatory effects, the Danio rerio VDR (zVDR) ligand binding domain (LBD) was crystallized in the presence of 1,20S(OH) 2 D 3 or 1,25(OH) 2 D 3 . The overall structure of VDR-1,20S(OH) 2 D 3 (PDB code: 5MX7) is highly homologous to the VDR-1,25(OH) 2 D 3 structure, adopting the canonical active conformation. When compared to the zVDR LBD-1,25(OH) 2 D 3 structure 19, 20 , the Cα atoms of the zVDR LBD-1,20S(OH) 2 D 3 complex have a root mean square deviation of 0.25 Å over 238 residues. The ligand binds similarly to 1,25(OH) 2 D 3 with the notable difference being that the 20S-OH forms a weak H-bond with His305 (3.42 Å) and does not interact with His397 (note that the residues numbers correspond to hVDR). The H-bond with His305 causes a ligand-induced conformational change in the receptor where His305 (loop6-7) is shifted by 0.63 Å to enable this H-bonding interaction. (Fig. 5). The 1α-OH and 3β-OH form similar hydrogen bonds to the zVDR to those seen with 1,25(OH) 2 D 3 .
An additional difference in the structures is that the 20S-OH forms a Van der Waals interaction with Val300. While most of the Van der Waals interactions are maintained, the side chain and terminal methyl groups that are differently positioned to interact differently with some of the residues (Fig. 5). Weaker interactions are formed with Leu227 (4.1 Å instead of 3.8 Å with C26) and Tyr399 (4.1 Å instead of 3.8 Å with C27), interactions compensated by stronger interactions with Val234 (3.9 Å instead of 4.2 Å with C22), and Leu412 (3.9 Å instead of 4.2 Å with C27). Overall, the hydrogen bonding interaction of 20S-OH with His305 and hydrophobic contacts formed by the ligand explains its agonist activity, however, with less potency than that of 1,25(OH) 2 D 3 .
Anti-inflammatory activity. The anti-inflammatory effect of 1,20S(OH) 2 D 3 was determined in mouse splenocytes stimulated by lipopolysaccharide prior to treatments with the secosteroids. The concentrations of IFNγ and IL1β in the culture media were significantly reduced by 1,20S(OH) 2 D 3 , compared with the control ( Table 1). The effect of 1,20S(OH) 2 D 3 (1.0 nM) was comparable with or slightly weaker than that of 1,25(OH) 2 D 3 but weaker than 22-Oxa for reduction of IFNγ production. In contrast, 1,20S(OH) 2 D 3 (100 nM) showed equal efficacy to 22-Oxa, and higher than that of 1,25(OH) 2 D 3 for reduction of IL1β production. These studies suggested that 1,20S(OH) 2 D 3 , acting similarly to 1,25(OH) 2 D 3 and 22-Oxa, is a potent anti-inflammatory agent.

Conclusions
Similar to 1,25(OH) 2 D 3 , 1,20S(OH) 2 D 3 can interact with the VDR with high potency, as evidenced by its ability to stimulate its translocation to the nucleus, regulate VDR downstream genes (including but not limited to VDR, CYP24A1, TRPV6 and CYP27B1), and exert strong anti-inflammatory activity. The crystal structure of 1,20S(OH) 2 D 3 bound to the VDR reveals differences from the 1,25(OH) 2 D 3-bound form with respect to their interactions, including the important role of the H-bond between the 20S-OH and His305 that shifts the position of this residue compared to the 1,25(OH) 2 D 3 -bound form. This difference may contribute to their differential  Table 1. Stimulation of VDRE-reporter activity and inhibition of cytokine production by 1,20S(OH) 2 D 3 . Note: VDRE stimulation activity = EC 50 ± standard deviation, cytokine level in splenocyte cultures = value ± standard error of the mean (pg/mL).
activities of these secosteroids such as the lower calcemic activity of 1,20S(OH) 2 D 3 compared to 1,25(OH) 2 D 3 8 . This study provides a molecular basis for the rational design and practical synthesis of novel 1,20S(OH) 2 D 3 analogs that interact with VDR for future drug development. 1,20S(OH) 2 D 3 was successfully chemically synthesized for the first time, providing ample material for further characterization of its biological activities, including animal studies in the future. The 1α,3β-diacetoxypregn-5-en-20-one (3) intermediate can serve as a common precursor for production of other 1,20S(OH) 2 D 3 analogs which will facilitate the synthesis of similar secosteroids containing a 1α-OH group.

Methods
General procedures. Reagents and solvents for the synthesis were anhydrous (purchased or self-dried) to ensure good product yield. Solvents used for separations were ACS chemical grade, purchased from commercial sources and used upon arrival. NH 4 Cl was sublimed in our lab for Birch reduction. Reactions for light sensitive Reactions for non-UV active compounds were visualized on TLC by 5% phosphomolybdic acid in ethanol. All NMR data were collected on a Bruker Avance III 400 MHz NMR or a Varian Inova 500 MHz NMR. Samples were dissolved in 0.5 mL CDCl 3 , methanol-d 4 , DMSO-d 6 or actone-d 6 , and NMR data were collected at r.t. Mass spectra of compounds were acquired using a Bruker LC-IT-MS system with an ESI source. High-resolution MS spectra and extracted ion chromatogram (EIC) were obtained using a Waters ACQUITY UPLC I-Class System equipped with a Xevo G2-S QTof mass spectrometer based on our previously reported conditions 26,27 . Reaction mixtures were extracted with ethyl acetate, DCM or hexanes, washed with aqueous Na 2 CO 3 , brine, and water, and then dried over anhydrous Na 2 SO 4 . The solution was transferred to a round-bottom flask and dried by rotary evaporator. The purities of final products were determined by HPLC as >98% (Fig. 4).

Crystallization of intermediate 15.
To a clean test tube (13 × 100 mm), 18 mg of compound 15 powder and 3 mL anhydrous n-hexane were added. The tube was shaken until the solid was completely dissolved, then sealed with 5 layers of sealing film (Para film) membrane. The resulting solution was allowed to stand in a quiet environment for 10 days, by which time the hexane had evaporated, leaving crystals of 15 which were collected for crystallographic analysis (Supplementary Information).
Purification was carried out as previously described, including metal affinity chromatography and gel filtration 28 . The protein was concentrated using Amicon ultra-30 (Millipore) to 3-7 mg/mL and incubated with a two-fold excess of ligand and a three-fold excess of the coactivator SRC-1 peptide (686-RHKILHRLLQEGSPS-698). Crystals were obtained in 50 mM Bis-Tris pH 6.5, 1.6 M lithium sulfate and 50 mM magnesium sulfate. Protein crystals were mounted in a fiber loop and flash-cooled under a nitrogen flux after cryo-protection with 20% glycerol. Data collection from a single frozen crystal was performed at 100 K on the ID23-1 beamline at ESRF (France). The raw data were processed and scaled with the HKL2000 program suite 29 . The crystals belong to the space group P6 5 22, with one LBD complex per asymmetric unit. The structure was solved and refined using BUSTER 30 , Phenix 31 and iterative model building using COOT 32 . Crystallographic refinement statistics are presented in Supplementary Table S9. All structural figures were prepared using PyMOL (www.pymol.org/). Biosynthesis of 1,20S(OH) 2 D 3 . Enzymatic synthesis of 1,20S(OH) 2 D 3 involved the 20S-hydroxylation of 1α(OH)D 3 by recombinant bovine CYP11A1 and was carried out as described in detail before 7 . HPLC comparison was determined by using an Agilent HPLC 1100 system and a Phenomenex Luna-PFP C 18 column (5 µm, 250 mm × 4.6 mm, Torrance, CA) at 25 °C and a flow rate of 1.0 mL/min. MeCN: H 2 O and MeOH: H 2 O were used as mobile phases with a gradient comprising 50-100% organic solvent for 30 min. 263 nm was used to display chromatograms.
VDR translocation assay. The effects of 1,20 S(OH) 2 D 3 on VDR translocation from the cytoplasm to the nucleus were tested on the previously described SKMEL-188 cell model 21,33 , using cells stably transduced with pLenti-CMV-VDREGFP-pgk-puro (VDR and EGFP expressed as fusion protein) 34 . Cells were treated with secosteroids (up to 100 nM) for 90 min followed by analysis with Cytation 5 (BioTek, Winooski, VT, US). Translocation to the nucleus was determined by counting cells with a fluorescent nucleus and the results are presented as the percentage of the total cells that displayed nuclear staining, as described previously 21 . The data were obtained from at least two separate experiments, with images taken in the central area from at least three different wells and counted as described 21, 34 . Real-time PCR assay. HaCaT cells were seeded in 60 mm dishes (1 million/dish) in 10 mL DMEM supplemented with 5% FBS and 1% Ab. After overnight incubation they were cultured in FBS-free medium for another 12 h to synchronize the cells. The media were then removed and secosteroids in DMEM (5% FBS and 1% Ab) with a concentration of 100 nM were added to the dishes. After 24 h incubation, media were removed, and 10 mL PBS was used to wash the dish. Cells were then detached by trypsin, centrifuged in Eppendorf tubes, washed with PBS (5 mL), and stored at −80 °C. Absolutely RNA Miniprep Kit (Stratagene, La Jolla, CA, USA) was used to isolate the RNA, and Transcriptor First Strand cDNA Synthesis Kit (Roche Inc., Mannheim, Germany) was used for reverse transcription (100 ng RNA/reaction). Real-time PCR was carried out using cDNA which was diluted 10-fold in sterile water and a SYBR Green PCR Master Mix. The forward reverse primers for VDR, CYP24A1, TRPV6 and CYP27B1 genes were designed based on the rat and mouse sequences using Primer Quest software (Integrated Device Technology, San Jose, CA, USA). Reactions (n = 3) were performed at 50 °C for 2 min, 95 °C for 10 min and then 40 cycles of 95 °C for 15 s, 60 °C for 30 s and 72 °C for 30 s. Data were collected and analyzed on a Roche Light Cycler 480. Using a comparative Ct method 25 , the amount of the final amplified product was normalized to the amount of β-actin as a housekeeping gene.
IFNγ production assay. All animal experiments in this study were performed in accordance with the NIH animal use guidelines and protocol (protocol No.: 15-043.0) approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Tennessee Health Science Center (UTHSC, Memphis, TN). Splenocytes were isolated from 7-week old C57BL/6 female mice and cultured at 2 × 10 6 /mL 500 μL/well for 72 h at 37 °C in a humidified atmosphere. Harvested supernatants were analyze for levels of murine IFNγ by ELISA (R & D Systems Minneapolis, MN) according to the manufacturer's instructions. Results are expressed as mean [IFNγ] ± SEM of triplicate determinations (pg/mL) of culture supernatant. The difference in [IFNγ] between EtOH vehicle + anti-CD3e MOAB (0.08 μg/mL) and [IFNγ] in each of the cultures containing D 3 analog was analyzed by one way ANOVA Sigma. The amount of EtOH in the EtOH + PBS and EtOH + anti-CD3e MOAB culture was equal to that in the cultures of the D 3 analogs at a level of 10 −9 M. Culture medium was RPMI 1640 supplemented with 9% charcoal stripped fetal bovine serum, non-essential amino acids, HEPES buffer Glutamax, penicillin 100 μg/mL, streptomycin 100 μg/mL, fungizone 1 μg/mL (GIBCO, Grand Island, NY) and 50 μM β-mercaptoethanol (Sigma, St. Louis, MO) 35 .
IL1β production assay. Splenocytes were isolated from 7-week old C57BL/6 female mice and cultured at 2 × 10 6 /mL, 500 μL/well, for 24 h at 37 °C in a humidified atmosphere. The vitamin D analogs or EtOH vehicle were added to the splenocyte cultures 2 h prior to addition of Lipopolysaccharide W E. coli 055:B5 (LPS) (Difco Lab. Defrost MI) 100 ng/mL or PBS vehicle. Harvested supernatants were analyzed for levels of murine IL1β by ELISA (R & D Systems Minneapolis, MN) according to the manufacturer's instructions. Results are expressed as mean IL1β concentration ± SEM of triplicate determinations (pg/mL) of culture supernatant. The amount of EtOH in the EtOH + PBS and EtOH + LPS culture was equal to that in the cultures of the vitamin D analogs at a level of 10 -7 M. Culture medium was RPMI 1640 supplemented with 9% charcoal stripped fetal bovine serum, non-essential amino acids, HEPES buffer Glutamax, penicillin 100 μg/mL, streptomycin 100 μg/mL, fungizone 1 μg/mL (GIBCO, Grand Island, NY) and 50 μM β-mercaptoethanol (Sigma, St. Louis, MO). The difference in [IL1β] between control and D 3 analog treatment was analyzed by one way ANOVA (Sigma Plot 13.0).