Investigation of 20S-hydroxyvitamin D3 analogs and their 1α-OH derivatives as potent vitamin D receptor agonists with anti-inflammatory activities

20S-hydroxyvitamin D3 [20S(OH)D3] is anti-inflammatory and not hypercalcemic, suggesting its potential as a lead compound. In this study, side chain modified 20S(OH)D3 analogs (4, 13, 23 and 33) together with their 1α-OH derivatives were synthesized and their metabolism and biological activities tested. 4, 13 and 23 are good substrates for CYP27B1, enabling enzymatic synthesis of their 1α-OH derivatives 5, 14 and 24. However, 33 could not be hydroxylated by CYP27B1 and acts as an inhibitor. All analogs were poorer substrates for CYP24A1 than calcitriol, indicating improved catabolic stability. While the parent analogs showed minimal VDR stimulating activity, their 1α-OH derivatives were potent VDR agonists. 4, 5, 14 and 24 significantly upregulated the expression of CYP24A1 at the mRNA level, consistent with their VDR activation abilities and indicating that 1α-hydroxylation is required to produce analogs with strong activity. These analogs have anti-inflammatory activities that are influenced by side chain composition and by 1α-hydroxylation. To understand their molecular interactions with the VDR, 20S(OH)D3, 4 and 33 were co-crystalized with the VDR ligand binding domain, which revealed subtle differences to the calcitriol-bound receptor. This study demonstrates the potential of the 20S(OH)D3 scaffold for the development of novel anti-inflammatory agents.


Synthesis of 20S(OH)D 3 analogs and their 1α-OH derivatives.
To investigate the effects of sidechain modification of VDR interaction and minimize metabolism by CYP24A1, a series of 20S(OH)D 3 analogs were synthesized. In addition, to determine the influence of having the 1α-OH present, their 1α-OH derivatives were biosynthetically made using purified CYP27B1. The chemical procedures and details are listed in the Supplementary information.
The synthesis of the first analog is shown in Fig. 2. We started from pregnenolone acetate 1 which was transformed into the 7-dehydrocholesterol (7DHC) type intermediate 2 by a well-established procedure with a 36% yield 1,4,23 . Grignard reaction using a self-made Grignard reagent removed the 3-acetyl, generated 20S-OH and added the modified side chain in one step to produce 3 with a satisfactory yield (87%). To open the B-ring to produce the D 3 -like structure, 3 dissolved in ethyl ether was irradiated with UVB light, followed by heat-induced isomerization to give the D 3 -like product 4 bearing a 24-ene with a 12% yield. Preparative HPLC was used to purify 4 from the reaction mixture using acetonitrile (MeCN) and water as mobile phases. The 1α-OH derivative 5 was produced by CYP27B1-mediated enzymatic catalysis which specifically adds the 1α-OH to the D 3 -like structures based on its function 24 , and was purified by HPLC.
The synthetic route for 13 and 14 bearing a 24-oxa group starting from 6 1 is shown in Fig. 3. Coupling of isopropyl bromine and 6 under basic condition gave 7 with a 92% yield. After replacing 3-OTBS with 3-OAc to afford 9, the 7DHC-like intermediate (10) was produced by the above-mentioned procedure (38% yield). EOM deprotection (81%) and ester hydrolysis (93%) were carried out under acidic condition (CSA) and basic condition (KOH), separately, to afford the final 7DHC intermediate 12. Similarly, the D 3 structure 13 and its 1α-OH derivative 14 were obtained from the B-ring opening reaction and enzymatic transformation, respectively. The synthetic route for 23 and 24 bearing a 24-difluoro modification started from 15 23 (Fig. 4). PDC oxidation of 15 gave ketone 16 with 95% yield, which was followed by replacing the 3-OTBS with 3-OAc to afford 18 which underwent DAST fluorination to produce 19 with a 30% yield. The 7DHC-like intermediate (20) was produced by the above-mentioned procedure (50% yield), then underwent EOM deprotection (79%) and ester hydrolysis (96%) under acidic condition (CSA) and basic condition (KOH), separately, to afford the final 7DHC-like   Similarly, the D 3 -like structure 23 and its 1α-OH derivative 24 were obtained from the B-ring opening reaction and enzymatic transformation, respectively.
The synthetic route for 33 using intermediate 25 1 is shown in Fig. 5. According to a previous procedure 25 , 25 was oxidized into acid 26 under mild condition without affecting the acid-sensitive TBS and EOM protection. Amide coupling of 26 and isopropylamine to construct the side chain were catalyzed by PyBOP with a 93% yield. After replacing 3-OTBS with 3-OAc to afford 29, the final 7DHC-like intermediate (32) was produced by the above-mentioned dibromatin/AIBN/TBAB/TBAF reaction, then EOM deprotection and ester hydrolysis, with a 22% yield for the three steps. Similarly, the D 3 -like structure 33 having a 23-amide modification was obtained from the the photochemical opening of the B-ring. Its 1α-OH derivative, however, could not be generated by CYP27B1 (discussed later).
The synthesis of analog 4 is relatively simple due to the commercially available reagent 5-bromo-2-methyl-2-pentene, making the side chain completion possible via a one-step Grignard reaction. In contrast, we have to construct the modified side chains for analogs 13, 23 and 33 by multi-step schemes, and the previously made intermediates (6, 15 and 25) have greatly sped up the synthesis of these analogs. The chemical addition of 1α-OH is not straightforward, thus the well-established biosynthetic strategy using CYP27B1 was applied here to produce sufficient of the 1α-hydroxy-derivatives for biological testing. Recently, we have worked out a synthetic route for producing 1α,20S(OH) 2 D 3 chemically 26 , the same strategy can be also applied to the chemical synthesis of the 1α-OH derivatives in the future.

Metabolism profiles of 20S(OH)D 3 analogs by CYP24A1.
Since CYP24A1 is the major enzyme determining the metabolic stability (half-life) of 20S(OH)D 3 and 1,25(OH) 2 D 3 , we tested its ability to metabolize the 20S(OH)D 3 analogs. Fig. S1 shows HPLC analyses of the products resulting from a 10 min incubation of 20S (OH)D 3 , 4, 13, 23 and 33 with rat CYP24A1 and reveals that this enzyme is capable of metabolizing all these compounds, despite the structural differences. Three major products were observed with 20S(OH)D 3 as substrate ( Fig. S1B) that were not present in the control (Fig. S1A), similar to what was reported before 27,28 . These were identified from authentic standards as 20S,25(OH) 2 D 3 , and the two C24 diastereomers of 20S,24(OH) 2 D 3 23,27 . One of the C24 diastereomers, 20S,24 R(OH) 2 D 3 is the major product and accounts for almost 70% of the total products. The action of CYP24A1 on 4 resulted in at least nine products (Fig. S1D), suggesting a complex pathway of metabolism. There were two major products formed from this analog, RT27 and RT30 (indicated by arrows), which accounted for approximately 20% and 30% of the total products, respectively. Compound 23 (Fig. S1E) was converted to at least 6 different products by the incubation with CYP24A1. In contrast, 13 tested under identical conditions generated only one major product with a retention time of 20 min (RT20) and three minor products (Fig. S1D). By 10 min of incubation, almost all of 13 had been metabolized by CYP24A1. Three major products (indicated by arrows) and at least three minor ones were observed for the metabolism of 33 by CYP24A1 (Fig. S1F).
Time courses for the metabolism of the 20S(OH)D 3 analogs by CYP24A1. The activity of CYP24A1 with the different analogs was determined as a function of time to determine the initial linear region that could be used for kinetic studies, and to determine any differences in the pattern between analogs. These time courses revealed that by 20 min there was 6-fold greater metabolism of 13 than for 20S(OH)D 3 (Fig. 6). The major product of 13 (Fig. S1C, RT20), accounted for 71% of the total secosteroids (products and substrate) at one min and by 20 min it was 95% of all secosteroids. The time course for 20S(OH)D 3 in comparison to 4 and 33, reached maximum product formation early in the incubation, with the amount of product at the end of the incubation being only 2.2-fold higher than at one min. By 20 min, less than 15% of the 20S(OH)D 3 had been metabolized. These data suggests that products of the metabolism of 20(OH)D 3 by CYP24A1 may inhibit the enzyme. While initially being metabolized at the lowest rate of any of the analogs, 33 maintained its initial rate longer with more product being present at the end of the 20 min incubation than for any of the other analogs except 13 (Fig. 6). Kinetics of the metabolism of 20S(OH)D 3 analogs by CYP24A1. To more fully document the differences in catalytic activities between the analogs revealed by Fig. 6, the kinetic parameters of CYP24A1 for the metabolism of the 20-hydroxy analogues were determined. An incubation time of 1 min was chosen to obtain the initial rate based on the time courses (Fig. 6) and previous studies 28,29 . 1,25(OH) 2 D 3 was included for comparison and all analogs displayed higher K m values than this major substrate for CYP24A1, reflecting their lower affinity for the active site of the enzyme. Only 13 displayed a higher maximum catalytic activity than 1,25(OH) 2 D 3 under saturating substrate conditions (k cat ), with 4 and 33 displaying the lowest k cat values which were approximately 4-fold lower than for 1,25(OH) 2 D 3 . K m values varied greatly between the 20S(OH)D 3 analogs (Table 1). 4 had the lowest K m for CYP24A1, whereas 20S(OH)D 3 displayed the highest which was over 17-fold higher than for 4. 13 and 23 displayed similar K m values, 3-4 fold lower than that for 20S(OH)D 3 , while 33 gave a K m half that of 20S(OH)D 3 . The resulting k cat /K m values which give a measure of catalytic efficiency under conditions of low substrate concentration, show that 13 is most efficiently metabolized by rat CYP24A1 with a value 72% higher than for the next best analog, 4. 20S(OH)D 3 and 33 displayed similar catalytic efficiencies which were the lowest of the analogs tested and 13-fold lower than for 13. Thus, the modifications at C24 that were introduced generally increased the efficiency of their metabolism. This can be explained in part by the presence of additional functional groups on the side chain increasing the interaction with CYP24A1 active site, thus lowering the K m . The known ability of CYP24A1 to hydroxylate the vitamin D side chain from C23 to C27 depending on what prior functional groups are present 27,29 provides an explanation as to why blocking of one or two carbons on the side chain from hydroxylation does not prevent hydroxylation at neighboring carbons. It is important to note that all the new 20S(OH)D 3 analogs were metabolized by CYP24A1 with a catalytic efficiency at least 5-fold lower than 1,25(OH) 2 D 3 ( Table 1), suggesting that they are likely to display increased metabolic stability in vivo.

Metabolism of 20S(OH)D 3 analogs by CYP27B1.
In order to make the 1α-OH derivatives of the 20S(OH)D 3 analogs enzymatically, and to assess their likely capacity to be 1α-hydroxylated in vivo, the ability of mouse CYP27B1 to hydroxylate these metabolites was examined. 25(OH)D 3 , the major natural substrate for CYP27B1 2,30 , was included for comparison. The 20S(OH)D 3 analogs were incorporated into phospholipid vesicles and incubated with CYP27B1, then the extent of their metabolism was determined by HPLC ( Table 2). All analogs tested except 33 were metabolized by CYP27B1. In each case only a single metabolite was produced, assumed   Table 1. Kinetic data for the metabolism of the 20S(OH)D 3 analogs by rat CYP24A1. Analogs were incorporated into phospholipid vesicles and incubated for one min in a reconstituted system with rat CYP24A1 (0.14 µM) and a range of substrate concentrations. Kinetic parameters were determined by fitting hyperbolic curves to the data using KaleidaGraph version 4.1. Data for K m and k cat are shown ± SE from the curve fit. PL: phospholipid.
to be the 1α-OH derivative based on the known high specificity of the enzyme for the 1α-position 24,30,31 . Analog 13 showed the highest conversion to product in the 20 min incubation, with greater than 90% conversion which was slightly higher than for 25(OH)D 3 and 4.3 fold higher than that observed with 20S(OH)D 3 . Metabolism of 4 was almost twice that of 20S(OH)D 3 and metabolism of 23 almost three fold. Thus, the modifications to C24 and/or C25 seen in 13, 4 and 23 enhanced their ability to be metabolized by CYP27B1. The replacement of carbon 24 with an oxygen atom had a similar effect to adding a hydroxyl group at C24 which we previously showed increased the catalytic efficiency of 1α-hydroxylation by both increasing the k cat and decreasing the K m relative to the values seen for 20S(OH)D 3 31 . In contrast, the introduction of an amide linkage into the side chain, as in 33, prevented its hydroxylation by CYP27B1. In a separate experiment we found that when present with an equal concentration 25(OH)D 3 , 33 was able to reduce the metabolism of 25(OH)D 3 from 82% to 56% in a 20 min incubation. This suggests that it is a competitive inhibitor and can compete for binding to the active site of CYP27B1 with 25(OH)D 3 , but binds in an unfavorable position for hydroxylation. VDRE stimulation activity. Using our previously established VDRE-Luciferase reporter models 1,4,23,32 , the VDR-induced transcriptional activity of the analogs and their 1α-OH derivatives were investigated in three different cell lines (Caco-2, HaCaT and Jurkat). As shown in Table 3, both positive controls, 1,25(OH) 2 D 3 and 22-Oxa, showed strong activity for VDR activation in all three cell lines, with 22-Oxa being the most potent one among all compounds tested. 1,25(OH) 2 D 3 and 22-Oxa showed better activities in Jurkat cells than in Caco-2 or HaCaT cells, suggesting that they are more sensitive to immune cells. 20S(OH)D 3 analogs were unable to significantly activate VDR except 4 which showed moderate activity. In Caco-2 and HaCaT cells, all 1α-OH derivatives displayed potent activities which were better than or comparable with that of 1,25(OH) 2 D 3 , the native ligand of VDR, suggesting that these analogs were also strong VDR agonists. The activities of 1α-OH derivatives in Jurkat cells were as potent as in Caco-2 and HaCaT cells. The better activities of 1α-OH derivatives over their parent analogs suggested the importance of 1α-OH for VDR activation with the synthetic VDRE used in these assays, which is consistent with our previous studies on 20S,23 S/R(OH) 2 D 3 1 and 20S,24 S/R(OH) 2 D 3 23 . One thing noteworthy is that 4 and 5 had relatively better activities than the other analogs and stood out amongst the 20S(OH) D 3 analogs and their 1α-OH derivatives, respectively. It would therefore appear that the presence of the 24-ene group inside the VDR binding pocket enhances VDR activation, particularly when the 1α-OH is also present.

RT-PCR-based analysis of CYP24A1 expression.
We have previously shown that 20S(OH)D 3 causes poor stimulation of CYP24A1 expression compared to 1,25(OH)2D 3 32,33 , despite having comparable potency for inhibiting keratinocyte proliferation 11,33 . We therefore compared the activity of the five synthetic analogs on the   To test whether the 20S(OH)D 3 analogs also show anti-inflammatory properties, the inhibition of IFNγ production was performed using our established assay 1,4,23 . As shown in Table 3, positive controls 1,25(OH)2D 3 , 22-Oxa and 20S(OH)D 3 showed similar effects on IFNγ production, decreasing its level by 60%, 51% and 48% at 100 nM, respectively. All chemically synthesized 20S(OH)D 3 analogs significantly reduced IFNγ concentrations, however, to a lesser degree than 1,25(OH)2D 3 , 22-Oxa and 20S(OH)D 3 . After 1α-hydroxylation, the inhibitory activities of all 1α-OH derivatives (37% to 62% decreases) were significantly improved as compared with their parent analogs (16.7% to 30% decreases). Among these compounds, 5 was the most active one because it showed comparable reduction of IFNγ production to 1,25(OH) 2 D 3 and larger reduction than 22-Oxa (p < 0.001). These results suggest that 1α-hydroxylation of these 20S(OH)D 3 analogs may improve their anti-inflammatory activity, which may also be affected by specific side chain modifications.  (Fig. 8). 20S(OH)D 3 interacts similarly to 1,20S(OH) 2 D 3 but lacks the H-bonds with Ser237 and Arg302 (Fig. 8A). The interaction of the 20S-OH with H305 in 4 is similar to that for 1,20S(OH) 2 D 3 forming a weak H-bond. Due to the 24-ene group in 4, the terminal methyl groups are differentially positioned forming stronger contacts with C-terminal residues, Val418 and Phe422, compared to 20S(OH)D 3 , that stabilize the agonist conformation of VDR in agreement with the strongest transcriptional activity of analog 4 amongst the 20S(OH)D3 analogs (Fig. 8B). Analogs 4 and 33 cause conformational changes in the position of His305 and His397 (Fig. S2) with quite a substantial change in the case of 33 which formed a hydrogen bond between His397 and the amide group which is in contrast to 20S(OH)D 3 which does not interact with this residue. While the exact relationship between these conformational changes induced by 33 and VDR function remain to be elucidated, the change indicates the possibility of producing analogs that display a unique subset of the actions seen for 1,25(OH) 2 D 3 (i.e. are biased agonists), as already seen to some degree for the parental 20S(OH)D 3 . They also substantiate previous molecular modeling studies that showed high docking scores for 20S(OH)D 3 and functional assays showing its ability to efficiently translocate VDR from the cytoplasm to the nucleus 9,15,17,18 .

Conclusion
In this study, four novel 20S(OH)D 3 analogs with side chain modifications were chemically synthesized, and their 1α-OH derivatives were produced enzymatically using recombinant CYP27B1. Enzymatic studies showed that CYP27B1 can hydroxylate the novel 20S(OH)D 3 analogs at C1α (except 33). All analogs served as substrates for CYP24A1 despite the modifications made at C24, but were much more metabolically stable than 1,25(OH) 2 D 3 . The 20S(OH)D 3 analogs showed VDR-stimulatory and VDR-downstream-gene regulatory activities that were significantly improved by 1α-hydroxylation. In addition, 1α-hydroxylation improved the inhibition of IFNγ production by activated lymphocytes. Co-crystal structures of VDR in complex with 20S(OH)D 3 , 4, and 33 revealed notable differences in their molecular interactions in the binding pocket of VDR, which will be insightful for developing novel VDR agonists as anti-inflammatory agents.

Methods
General methods. All reagents and solvents in synthetic and separation procedures were purchased from commercial sources and were used as received unless otherwise noted. Reactions of 5,7-diene structures were all protected by wrapping the flasks with aluminum foil. Moisture-or oxygen-sensitive reactions were performed under an argon atmosphere. All reactions were routinely monitored by TLC on silica gel, and visualized by 5% phosphomolybdic acid in ethanol for non-UV active compounds or UV lights for compounds with absorption at 254 nm. Ethyl acetate was used for extraction of reaction mixtures and then dried over anhydrous Na 2 SO 4 , filtered and removed using a rotary evaporator under reduced pressure. Mass spectra of all compounds were obtained using a Bruker ESQUIRE-LC/MS system equipped with an ESI source (Billerica, MA, USA). The purities of final D 3 compounds, as analyzed by an Agilent 1100 HPLC system (Santa Clara, CA, USA), were above 98%. Highresolution MS spectra were obtained from a Waters UPLC-Q/Tof-MS system with the function of molecular formula prediction (Milford, MA, USA). NMR data were collected at 25 °C. Chemical shifts were referenced to residual solvent peaks of methanol-d 4 or CDCl 3 . NMR measurements were performed on either a Bruker Avance III 400 MHz (Bruker BioSpin, Billerica, MA, USA), or a Varian Unity Inova 500 MHz spectrometer (Agilent Technologies Inc., Santa Clara, CA, USA).

Metabolism of analogs by CYP24A1 and CYP27B1.
Rat CYP24A1, mouse CYP27B1 and adrenodoxin and human adrenodoxin reductase were expressed in E. coli and purified as described before 28,30,35 . To test metabolism of each analog, they were incorporated into phospholipid vesicles made from dioleoyl phosphatidylcholine and cardiolipin by sonication, as before 24,28 . The substrates in vesicles (510 µM phospholipid) were incubated at 37 °C with either CYP24A1 (0.14 µM) or CYP27B1 (0.8 µM) in a reconstituted system containing adrenodoxin (15 µM) and adrenodoxin reductase (0.4 µM). Samples from incubations with CYP24A1 were extracted with dichloromethane and analysed by HPLC using a 25 cm Grace Alltima C18 column, as before 27,28 . Products (except 33) were separated using an acetonitrile on water gradient (45% to 100% for 20 min then 100% acetonitrile for 40 min at a flow rate of 0.5 mL/min). For the more polar 33, the acetonitrile gradient was 30% to 100% acetonitrile for 30 min then 100% acetonitrile for 20 min, at 0.5 mL/min. Products from incubations with CYP27B1 were similarly extracted with dichloromethane and analyzed by reverse phase HPLC using a 15 cm Grace Smart C18 column and an acetonitrile in water gradient (10 min 45% to 100% acetonitrile then 20 min at 100% acetonitrile, at 0.5 mL/min).
VDRE reporter assays. Caco-2, HaCaT and Jurkat cells were cultured as described previously 1 IFNγ inhibition assay. Secosteroids were solubilized in absolute EtOH at 10 −4 M and diluted to 10 −6 M by adding Eagles Minimal Essential Medium (EMEM) containing 9% charcoal-stripped fetal calf serum, 100 U/ mL penicillin and 100 µg/mL streptomycin, non-essential amino acids, 2.5 mM 2-mercaptoethanol, 2.5 mM L-glutamine 16 . Splenocytes from mice were isolated, erythrocytes lysed by hypotonic shock, washed twice with EMEM, and suspended at a concentration for 2 × 10 6 cells/mL in EMEM as described above. To each well in a 48-well tissue culture plate, 450 µL of the splenocytes were added. Secosteroids (50 µL of the 10 −6 M stock) or EtOH diluted 1:100 with the above culture medium were added to triplicate wells and then incubated at 37 °C in 5% CO 2 in a humidified tissue culture incubator for 2 h, after which 1 µg/well of rat anti-mouse CD3 MOAB was added. After 72 h culture, supernatants from each well were harvested and analyzed by ELISA for levels of D-murine IFNγ (RAD Systems, Minneapolis, MN, USA), according to the manufacturer's instructions. The concentration of IFNγ is supernatants from cultures containing secosteroids were compared to the concentration of IFNγ in the supernatants of EtOH-treated control cultures, by ANOVA.
Co-crystallization of VDR in complex with secosteroids. zVDR LBD (156-453 AA) was produced and purified as previously described 36 . The protein was concentrated using an 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 ID30 beamline at ESRF (France). The raw data were processed and scaled with the HKL2000 program suite 37 . The crystals belong to the space group P6522, with one LBD complex per asymmetric unit. The structure was solved and refined using BUSTER 38 , Phenix 39 and iterative model building using COOT 40 . Crystallographic refinement statistics are presented in Table S1. All structural figures were prepared using PyMOL (www.pymol.org/). Accession codes. PDB ID codes 5OW9, 5OWD, 5OW7, 5MX7 and 2HC4 for co-crystals of 20S(OH)D 3 , 4, 33 1,20S(OH) 2 D 3 and 1,25(OH) 2 D 3 in complex with VDR ligand binding domain.