RO215535 – Ap-iii-a40

Novel calcitriol analogue with an oxolane group: In vitro, in vivo, and in silico studies

Correspondence
Alejandro C. Curino, Instituto de Investigaciones Bioquímicas Bahía Blanca (INIBIBB), Departamento de Biología, Bioquímica y Farmacia, Universidad Nacional
del Sur (UNS)‐CONICET, Camino La
Carrindanga, Km 7‐C.C. 857, 8000 Bahía Blanca, Argentina.
Email: [email protected]

Funding information
Fondo para la Investigación Científica y Tecnológica, Grant/Award Number: PICT 2012‐0966; CITACA Strategic Partnership, Grant/Award Number: ED431E 2018/07; Xunta de Galicia, Grant/Award Number: ED431C2017/70; Secretaria de Ciencia y Técnica de la Universidad Nacional del Sur; Consejo Nacional

Abstract
The active form of vitamin D3, calcitriol, is a potent antiproliferative compound. However, when effective antitumor doses of calcitriol are used, hypercalcemic effects are observed, thus blocking its therapeutic application. To overcome this problem, structural analogues have been designed with the aim of retaining or even increasing the antitumor effects while decreasing its calcemic activity. This report aims at gaining insights into the structure–activity relationships of the novel oxolane‐ containing analogue, AM‐27, recently synthesized. We herein demonstrate that this compound has antiproliferative and antimigratory effects in squamous cell carcinoma, glioblastoma, and breast cancer cell lines. Analyses of the mechanisms underlying the AM‐27 effects on cell viability revealed induction of apoptosis by the analogue.

Importantly, nonmalignant cell lines were little or not affected by the compound. In addition, the analogue did not produce hypercalcemia in mice. Also, in silico studies
involving docking and molecular dynamics techniques showed that AM‐27 is able to
bind to the human vitamin D receptor with a higher affinity than the natural ligand calcitriol, a feature that is mostly derived from an electrostatic interaction pattern. Altogether, the proapoptotic effect observed in cancer cells, the lack of calcemic
activity in mice, and the differential effects in normal cells suggest the potential of AM‐27 as a therapeutic compound for cancer treatment.

KEYWORDS
analogue, calcitriol, cancer, oxolane, vitamin D receptor

INTRODUCTION

D. J. Obiol and A. Martínez have contributed equally to this work.

The active metabolite of vitamin D, 1α,25‐dihydroxy vitamin D3 (calcitriol), is classically known to regulate calcium and phosphate
homeostasis.[1] In addition, calcitriol has also been demonstrated to act as a potent anticancer agent in multiple cell culture and animal models of cancer[2]; however, this antitumor activity occurs at supraphysiolo- gical concentrations leading to the occurrence of toxic effects such as hypercalcemia, hypercalciuria, and increased bone resorption.[3] There- fore, structural analogues of the hormone are being designed and tested to find those that maintain, or even increase, the antitumor effects without the undesirable occurrence of side effects.

Up to now few calcitriol analogues with an oxolane moiety in their side chains have been designed and synthesized.[4] Two stereo diastereoisomers AMCR277A and AMCR277B were synthesized bearing an oxolane ring, with the aim at optimizing the aliphatic side chain conformation with a subsequent entropy benefit. Incorpora- tion of the oxolane ring bridging C20 and C23 drastically altered the biological profile of vitamin D compounds.[5] Only AMCR277A was described as a vitamin D receptor (VDR) superagonist with antiproliferative activity, although it also produced hypercalcemia in vivo.[6]

Subsequently, these structures were C2α‐methylated to increase their stability in the ligand binding pocket to enhance superagonistic activity. The C2α‐methylated derivative of AM- CR277A was shown to exhibit superagonistic and antitumoral properties but displayed calcemic effects, and its epimer C2α‐methyl AMCR277B did not produce hypercalcemia but it behaved like the natural ligand in vitro.[7] Besides these oxolane analogues, the novel AM‐27 has been recently synthesized (Figure 1).[8]

The antineoplastic effects of calcitriol do generally involve the binding to its nuclear receptor VDR, which belongs to the super-
family of nuclear receptors for steroid hormones.[9–15] Therefore, effective binding of the vitamin D ligand to this protein is of crucial importance. There has been significant progress on the knowledge about VDR effects since the crystallographic structure of the
calcitriol‐VDR complex (PDB code: 1DB1)[16] was deposited in the Protein Data Bank.[17] This allowed to obtain information about the functional potential of VDR ligands[18] by using bioinformatics tools. In the present work, the results from biological and computational
studies of AM‐27 are shown. This analogue is a diastereoisomer of (1S,3R)‐dihydroxy‐(20S)‐[(2′′‐hydroxy‐2′′‐propyl)‐tetrahydrofuryl]‐22,23, 24,25,26,27‐hexanor‐1α‐hydroxyvitamin D3, which presents a different stereochemistry at positions C2 and C5 of the oxolane ring branched at
carbon C22 (C2RC5S).[4] The biological assays included testing analogue hypercalcemic activity in mice and cellular viability and migration in various tumor cell lines grown in culture. The interaction between AM‐27 and human VDR at the molecular level was studied by applying computational techniques, such as molecular docking, molecular dynamics (MD), and free energy of binding analyses.

2 | RESULTS

2.1 | Biological assays
We first checked if AM‐27 displayed antitumoral activity in culture
and for this purpose various normal and malignant cells were used: HN12 and HN13 head and neck squamous cell carcinoma cells; GL26, T98G, and U251 glioblastoma multiforme cells; and LM3 and 4T1 breast adenocarcinoma cells. Also, HC11 nonmalignant mammary epithelial cell line and primary cultures of human astrocytes were used as the normal counterpart of breast cancer and glioma, respectively. The cells were plated into 96‐multiwell dishes, treated with 0.01–100 nM of calcitriol, AM‐27, or vehicle for 120 hr, and counted. As shown in Figure 2, the analogue reduced cellular viability of both head and neck cancer cell lines (HN12 and HN13; Figure 2a,b). In addition, AM‐27 displayed antitumoral effects on GL26, T98G, and U251 glioma cell lines (Figure 2c,e,f).

The half maximal inhibitory concentration (IC50) is shown in the right upper corner of each graph. Importantly, no effect on the growth of normal human astrocytes was observed (Figure 2d). Similarly, AM‐27 was effective in reducing cell viability of the breast cancer LM3 and 4T1 cell lines (LM3, IC50 = 155 nM; 4T1, IC50 = 0.038 nM) while displaying lower potency in the nonmalignant HC11 mammary cell line
(IC50 = 278 nM; Figure 2g–i). Calcitriol reduced the viability of the human normal astrocytes and of all cell lines tested except for the glioblastoma T98G cell line, which responded only to AM‐27.

It is essential for cancer progression that the tumor cells acquire invasive skills. In this regard, calcitriol and some analogues have proved to inhibit migration and invasion of various tumor types,[2,19] thus preventing the metastatic process. Therefore, we next studied the effect of AM‐27 on cell migration by means of the wound closure assay. As shown in Figure 3, AM‐27 impaired
the migration process of the U251 and LM3 cell lines while no significant differences were observed for HN12, HN13, GL26, and T98G cell lines. AM‐27 was effective in reducing the cell count of all cancer cell lines tested. Therefore, we continued with the analyses of the mechanisms by which the analogue affects cell viability.

2.2 | Computational procedures
It is well known that the genomic effects of calcitriol are exerted through its binding to VDR which is present in, at least, 30 tissues.[26] Vitamin D is transported in serum by α‐globulin, vitamin D bindingn protein, and after dissociating from this protein, it enters target cells and binds to the VDR. The receptor is formed by 427 amino acids and contains two main functional regions: the DNA binding region and the ligand binding domain (LBD). The binding site of calcitriol is located nwithin the LBD region and exerts the activation of the receptor by eliciting a conformational modulation of the H‐12 helix of the LBD. This allosteric modulation originates the so‐called functional activated domain AF‐2, which in turns mediates the binding of transcriptional comodulators. This complex translocates to the nucleus and hetero- dimerizes with any of three isoforms ofretinoid X receptor. Finally, this heterodimer binds to specific sequences in promoter regions of target genes regulating the expression of >900 genes.[27] In light of the above‐described mechanism of action, computational techniques (molecular docking, MD, and free energy of binding analyses) were applied to comparatively study the binding of calcitriol and AM‐27 to VDR, analyzing their pharmacodynamic behavior and ability to elicit the allosteric modulation required for bioactivity.

Computational studies were performed using as reference the crystallographic structure of the calcitriol‐VDR complex.[16] Marvin software was used for drawing, displaying, and characterizing initial structures of calcitriol and AM‐27.[28] Afterward, structural and energetic analyses were performed using the Gaussian 03[29] software to obtain the minimum energy conformation by applying semiempirical (AM1) and ab initio (HF/6‐311 + G*) methods. Figure 6a shows the intermolecular interactions between calcitriol and AM‐27 to VDR as was obtained from docking studies using FRED software.[30,31] As can be seen, AM‐27 was able to fit within the VDR ligand binding site, establishing most of the intermolecular interactions observed for calcitriol. To assess the conformational space available for calcitriol and AM‐27, conformers enumeration at different energy barrier were studied by means of OMEGA software.[32,33] It was found that both bioactive conformations were present at an energy barrier of 10 kcal/mol, with 71 conformers being found for calcitriol and 49 for AM‐27.

This observation indicates that AM‐27 is able to establish the required intermolecular interactions with VDR in a restricted conformational space. To further study the pharmacodynamic behavior of AM‐27, explicit solvent MD simulations were performed. From these
studies, structural and energetic analyses were performed by applying the MM‐PBSA technique,[34] both for calcitriol and AM‐27.
As can be seen in Figure 6b, AM‐27 exhibited higher total interaction energy compared with that of calcitriol (−82.5 and
−73.5 kcal/mol, respectively), feature that is originated in an enhanced electrostatic interaction of AM‐27 with VDR compared with that of calcitriol. From the per‐residue interaction energy analysis (Figure 6c), it can be seen that this higher electrostatic interaction is originated in intermolecular contacts with Arg274 and His397. It is noteworthy that both calcitriol and AM‐27
established intermolecular interactions with Arg274 and His397 as evidenced by molecular docking studies (Figure 6a), so it is suggested that the enhanced interaction observed for AM‐27 is derived from a more sustained and conformationally favored
interaction.

To analyze this effect, hydrogen bond interactions between both ligands and Arg274 and His397 were monitored. As can be seen in Figure 6d, AM‐27 was able to establish stable hydrogen bond interactions with Arg274 and His397, with calculation lifetimes between 50% and 60% during the whole MD trajectory. When MD simulations obtained for calcitriol were analyzed, the above‐mentioned hydrogen bond interactions were very transient throughout the simulations, with a calculated lifetime of 11% and 10% for the secondary and primary hydroxyl groups, respectively. These analyses demonstrate that the higher affinity for VDR calculated for AM‐27 is mainly originated by
efficient hydrogen bond interactions established with Arg274 and His397, feature that is driven by the conformational restriction of
the five‐membered ring present in AM‐27. The efficiency of the interaction of the VDR‐bound ligand with His397 seems to be determinant for the stable interaction with Arg274.

Finally, to assess the effect of AM‐27 to maintain the structural conformation of H‐12 helix observed for the natural ligand (calcitriol), the root mean square fluctuation (RMSF) versus time for this loop was calculated. As can be seen in Figure 7, AM‐27 maintained the structural conformation on the H‐12, suggesting that the binding of this analogue is compatible with its VDR‐mediated transcriptional modulation.

Atomic positional fluctuations (RMSF) relative to the crystallographic structure calculated for residues comprising H‐12 of vitamin D receptor, obtained during molecular dynamics trajectories for calcitriol (black line) and AM‐27 (red line). RMSF: root mean square fluctuation

3 | DISCUSSION

AM‐27 has been recently described as a weak VDR agonist with reduced transcriptional activity from transactivation and mammalian two‐hybrid assays performed in HEK293 and COS7 cells, respectively,[8] which are nonmalignant cell lines. Due to the fact that the coactivators and corepressors present in the cells could determine whether a calcitriol analogue could act as agonist or antagonist[35] and bearing in mind that the presence and ratio of these comodulators is cell‐type‐specific, it was necessary to employ cancer cell lines for evaluating the potential antitumor activity of AM‐27. On the other hand, the initial analyses of AM‐27[8] showed results obtained from crystal-lographic characterization of AM‐27 coupled to zebrafish‐VDR. Similarly, the affinity of AM‐27 to human VDR would better predict the behavior of the analogue in human cancer.

We have, therefore, followed up with the biological assays employing cancer cell lines and with the computational studies using human VDR. We demonstrated that the new calcitriol analogue displays significant antitumor activity in cell lines belonging to various tumor types. Specifically, it reduced cellular viability of squamous cell carcinoma, breast carcinoma, and glioblastoma cell lines and decreased the migration process in glioblastoma and a breast carcinoma cell line. Although all of the cell lines responded to AM‐27, the potency of this compound varied according to the cell line used. In relation to this observation, it has been reported that vitamin D compounds display differential effects depending on the cell type.[36] This has been ascribed to the different coactivators and corepressors present in the cells. Calcitriol and its analogues have been shown to display antiproliferative activity that could be explained by either induction of apoptosis or inhibition of cell cycle progression.[2]

The analyses of the mechanisms underlying AM‐27 inhibition of proliferation showed that this compound produces an increase in apoptosis in both tumor types studied. In addition, the analogue demonstrated a differential action between tumor and normal or nonmalignant cells because it did not affect cellular viability of human primary astrocytes and displayed mild effects in the normal mammary cell line. This therapeutic window is important when studying the potential anticancer activity of novel drugs.
The in vivo studies demonstrated that, unlike calcitriol, AM‐27 does not display hypercalcemic activity at the high doses tested. In addition, no alterations of the hematocrit were observed. These effects on plasma calcium levels and hematocrit are preliminary parameters that demonstrate a lack of toxic effects in in vivo assays performed in mice. Therefore, although the potency of AM‐27 was
lower than that of calcitriol (the IC50 were higher for AM‐27), the lack of calcemic activity and the differential effects in normal cells
suggest it could be a potential therapeutic compound for cancer treatment.

In addition, our in silico studies demonstrated that the analogue was able to bind with higher affinity to human VDR than calcitriol, characteristic that results mainly from the hydrogen bonding with Arg274 and His397 residues. These results are in accordance with the conclusions of the initial work by Belorusova et al.[8] In this study, the crystals of the oxolane ligands in complex with the zebrafish‐VDR‐LBD were analyzed demonstrating that the introduction of an oxolane ring in the side chain increases thenumber of interactions with the LBD in comparison to calcitriol. Interestingly, the terminal methyl groups of oxolane ligands formed new connections with Phe448 of the H‐12. In relation to this, our RMSF studies with human VDR indicate that AM‐27 maintains the structural conformation of H‐12, suggesting the possibility of creating the activated domain AF‐2 for the recruitment of comodulators, necessary for transcription and thus for the antitumor effects. Overall, these results further our understanding of oxolane‐based calcitriol analogues and provide useful information for the develop- ment of next‐generation calcitriol analogues.

4 | EXPERIMENTAL

4.1 | Biological procedures
4.1.1 | Chemicals and reagents
1α,25‐Dihydroxyvitamin D3 and AM‐27 were reconstituted in 100% high‐performance liquid chromatography‐grade isopropanol and stored protected from light at −20°C. The amount of 1α,25(OH)2‐D3 and AM‐
27 was determined by UV spectrophotometry (SpectroquantPharo 300; Merck, Darmstadt, Germany) between 200 and 300 nm. Both drugs
were dissolved in isopropanol to the concentration of 10–3 M and
subsequently diluted in the culture medium to reach the required concentrations (range, 0.01–100 nM).

4.1.2 | Cell lines
Biological evaluation of AM‐27 was performed on three different
tumors, each of them represented by various cell lines. The glioblastoma multiforme (T98G and U251) and head and neck squamous cell carcinoma (HN12 and HN13) cell lines were kindly donated by Silvio Gutkind. The murine breast adenocarcinoma cell line LM3 and the murine breast carcinoma 4T1 were a generous gift from E. Bal de Kier Joffé (Instituto de Oncología Ángel Roffo, Buenos Aires, Argentina). All cell lines along with the primary cultures of human astrocytes were maintained at 37°C, 5% CO2 in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 100 U/ml penicillin, 100 U/ml streptomycin, 4 mM glutamine, and 10% fetal bovine serum (FBS) except for the LM3 that was 5% FBS. GL26 (murine glioblastoma multiforme) and 4T1 (murine breast adenocarcinoma) were maintained in RPMI (Sigma‐Aldrich, St. Louis, MO) supplemented with 10% (v/v) FBS (Gibco, Carlsbad, CA), L‐glutamine (5 mM, Gibco), penicillin (100 U/ml, Gibco), and streptomycin (100 μg/ml, Gibco). HC11 (nonmalignant murine mammary epithelial cell line) was maintained under the same conditions as GL26, but it was supplemented with insulin (5 μg/ml; Gibco). The cells were passed every 3–4 days.

4.1.3 | Effects of AM‐27 on cellular viability
We performed dose‐response experiments. The cell lines were plated at a density of 500–2000 cells/well into 96‐multiwell dishes in complete medium. They were treated with 0.01–100 nM of calcitriol, AM‐27, or vehicle (isopropanol), resulting in a maximum amount of 0.1% (v/v) alcohol in the assay dishes. We did not find that this amount or lower quantities of alcohol had any significant effect on these cell lines (medium was changed every 2 days). Cells were incubated for 120 hr.
They were washed with phosphate‐buffered saline (PBS) 1×, trypsinized
and resuspended in 100 μl complete medium. They were counted manually using a hemocytometer. Data was analyzed using two‐way analysis of variance (ANOVA) followed by Bonferroni’s post‐test analysis to determine the effects of increasing concentrations of
calcitriol, AM‐27, and vehicle on cellular survival. Graphs were plotted using the Prism 5.0 (GraphPad Prism Software Inc., San Diego, CA).

4.1.4 | Effects of AM‐27 on cellular migration
The different cell lines tested were seeded in 35 mm Petri dishes and when they reached appropriate confluence, cell culture was injured
with a tip of 200 µl, immediately after, the lesion area was quantified, and treatments with vehicle (isopropanol) or analogue AM‐27 were performed. The wound closure was monitored by measuring and quantifying the wound area not covered at different time laps with ImageJ 1.37v (Wayne Rasband, National Institute of Health; http:// rsb.info.nih.gov/ij/). The results were analyzed by two‐way ANOVA
with Bonferroni’s post‐test. The line graph shows the percentage of the wound area.

4.1.5 | Effects of AM‐27 on cell cycle
For fluorescence‐activated cell sorting (FACS) analysis, HN12 cells were plated at a density of 100,000 cells/100 mm dish in DMEM
supplemented with 10% FBS. Twenty‐four hours after plating, cells were treated with AM‐27, calcitriol or vehicle (100 nM) during 120 hr. Then, cells were harvested by trypsinization, washed twice with ice‐cold 1× PBS, fixed by 70% ethanol and incubated at −20°C during 24 hr. Before flow cytometry analysis, the cellular double‐ stranded nucleic acids were stained with PI (50 μg/ml; Roche, Mannheim, Germany) and RNase A (100 U/ml) to degrade double‐ stranded RNA in PBS for 60 min at room temperature. Cell cycle analysis was performed with FACScan flow cytometry (Becton Dickinson). Data were analyzed using CellQuest software (Becton Dickinson). One thousand forward scatter gated events were collected per sample. Data were analyzed using two‐way ANOVA
with Bonferroni’s post‐test analysis.

4.1.6 | Effects of AM‐27 on induction of apoptosis
Cells were seeded onto coverslips at a density of 50,000 cells/35 mm dish and after 24 hr they were treated with AM‐27 or vehicle (100 nM) during 120 hr. Then the cells were fixed with 4% formaldehyde in PBS for 60 min at room temperature, washed with PBS twice for 5 min and incubated with blocking solution (3% H2O2 in methanol) during 10 min. After that, coverslips were washed with PBS twice for 5 min and coated with permeabilization solution (0.1% Triton X‐100 in 0.1% sodium citrate, freshly prepared) and the In Situ Cell Death Detection Kit, POD (Roche) was used to detect apoptotic cells. Photographs were taken using an inverted phase microscope Nikon Eclipse TE 2000S (Nikon Instruments Inc., Tokyo, Japan). Ten fields per condition were evaluated at ×200 magnification. In addition, the mitochondrial membrane potential was measured as follows: cells were seeded onto coverslips at a density of 50,000 cells/35‐mm dish and after 24 hr they were treated with AM‐27 (IC50) or vehicle during 120 hr. Then the coverslips were incubated with the fluorescent probe at 200 nM (0.1 g/ml; MitoTracker; Molecular Probes, Eugene, OR) and afterward were washed, fixed in paraformaldehyde 4% for 20 min and mounted with mounting media (VectaShield®, Vector Labs) onto slides. The cells were visualized and photographed by using a Nikon Eclipse TE 2000S at ×200. Images were quantified with ImageJ (NIH).

4.1.7 | Effects of AM‐27 on blood calcium levels
Inbred normal CF1 mice aged 8–10 weeks and weighing 40 g were obtained from the animal facility of the Biology, Biochemistry and
Pharmacy Department of the Universidad Nacional del Sur (Bahía Blanca, Argentina). The mice were treated in accordance with the institutional animal care and use committee guidelines. Calcium level studies were performed following daily intraperitoneal injection of 5 μg/ kg body weight of either AM‐27 or vehicle (isopropanol), during four consecutive days. Blood samples were collected from mice (basal levels as well as at 24, 48, 72, and 96 hr). Animals were anesthetized with Acedan (Holliday Scott, Beccar, Argentina) 0.22 mg/kg body weight and heparinized capillary tubes were used to collect blood from the retro‐ orbital sinus. Samples were held on ice, protected from light and processed at 4°C. Plasma was separated by centrifugation at 10g and stored at −20°C until assayed. Approximately 10–15 µl of plasma/ mouse was obtained each time. Calcium concentration was determined using Ca‐Color Arsenazo III AA Kit (Wiener Lab, Rosario, Argentina), measuring the absorbance at 650 nM using a spectrophotometer. The calcium concentration was calculated from calcium standards provided by the manufacturer. To adjust for differences in hemolysis among samples, blanks were prepared and the absorbance reading was subtracted from the test reading. We also measured the hematocrit for each mouse before and following daily treatments to determine if
they were healthy. Data was analyzed by two‐way ANOVA with Bonferroni’s post‐test. A p < 0.05 was considered significant. 4.2 | Computational procedures Computational studies were performed using as reference the crystallographic structure of the calcitriol‐VDR complex (PDB code: 1DB1).[18] Initial structures of calcitriol and AM‐27 were built using the MarvinSketch software,[28] after which structural and energetic analyses were performed using the Gaussian 03 software[29] to obtain the minimum energy conformation by applying semiempirical (AM1) and ab initio (HF/6‐311 + G*) methods. Molecular docking assays were performed using software packages developed by OpenEye Scientific Software (Santa Fe, NM, http://www. eyesopen.com). Docking procedures consisted of three sequential stages: (a) a ligand conformer library generation, which was conducted at an energy threshold of 10 kcal/mol using the OMEGA software,[32,33] (b) the docking runs, which were performed by applying a fast rigid exhaustive docking approach as implemented in the FRED3 soft- ware,[30,31] with the ChemGauss3 scoring function being used to evaluate and rank resulting docked poses. The lowest energy docked pose was considered for further analyses. Stage (c) involved three‐ dimensional visualization and intermolecular interactions predictions, which were performed using the VIDA[37] and LigPlot+[38] software packages, respectively. To obtain and analyze MD trajectories, the AMBER14 software package was used.[39] Atomic charges and molecular parameters corresponding to calcitriol and AM‐27 were assigned from the GAFF force field,[40] while those corresponding to the macromolecule were assigned from the AMBER force field 99SB.[41] Complexes predicted by molecular docking were used as initial structures, solvated with a pre‐equilibrated TIP3P octahedral box of explicit water molecules, and subjected to energy minimization. The minimized systems were heated to 298 K for 100 ps, using a time step of 2 fs under constant pressure and temperature conditions. The SHAKE algorithm was applied to constrain bonds involving hydrogen atoms. After conclud- ing the heating phase, an equilibration stage (1 ns) was performed and followed by the corresponding production stages (80 ns). Analyses of the MD trajectories were carried out using the Cpptraj module of AMBER14, with energetic and per‐residue decomposition analyses being performed by applying the molecular mechanics Poisson–Boltzmann surface area (MM‐PBSA) approach as implemen- ted in the MM‐PBSA.py tool.[34] The resulting trajectories were visualized using VMD v.1.9 software.[42] In all cases, MD trajectories were obtained using CUDA designed code (pmemd.cuda), with computational facilities provided by the GPGPU Computing group at the Facultad de Matemática, Astronomía y Física (FAMAF), Universidad Nacional de Coŕdoba, Argentina. ACKNOWLEDGMENTS This study was supported by grants and fellowships awarded by the National Council of Scientific and Technical Research (CONICET), the National Agency for Scientific and Technological Promotion (ANPCyT, PICT 2012‐0966 and PICT 2012‐1595), and the Uni- versidad Nacional del Sur (PGI 24/B221, SGCyT‐UNS), Argentina and the Xunta de Galicia, Spain: (ED431C2017/70) and CITACA Strategic Partnership (ED431E 2018/07). M. A. Q. acknowledges the GPGPU Computing Group from the Facultad de Matemática, Astronomía y Física (FAMAF), Universi- dad Nacional de Córdoba, Argentina, for providing access to computing resources. A. M. thanks the University of Vigo for a fellowship. CONFLICT OF INTEREST The authors declare no conflict of interest. ORCID María M. Facchinetti http://orcid.org/0000-0001-7596-4769 REFERENCES [1] M. F. Holick, N. Engl. J. Med. 2007, 357, 266‐281. [2] D. G. Salomon, E. Mascaro, S. M. Grioli, M. J. Ferronato, C. A. Vitale, G. E. Radivoy, et al., Curr. Top. Med. Chem. 2014, 14, 2408. [3] K. C. Chiang, C. C. Sun, M. H. Chen, C. Y. Huang, J. T. Hsu, T. S. Yeh, et al., J. Steroid Biochem. Mol. Biol. 2016, 155, 26. [4] A. Martínez, Z. Gándara, M. González, G. Gómez, Y. Fall, Tetrahedron Lett. 2013, 54, 3514. [5] W. Sicinska, M. Kurcinski, J. Steroid Biochem. Mol. Biol. 2010, 121, 34. [6] S. Hourai, L. C. Rodrigues, P. Antony, B. Reina‐San‐Martin, F. Ciesielski, B. C. Magnier, et al., Chem. Biol. 2008, 15, 383. [7] P. Antony, R. Sigüeiro, T. Huet, Y. Sato, N. Ramalanjaona, L. C. Rodrigues, et al., J. Med. Chem. 2010, 53, 1159. [8] A. Y. Belorusova, A. Martínez, Z. Gándara, G. Gómez, Y. Fall, N. Rochel, Eur. J. Med. Chem. 2017, 134, 86. [9] T. M. Beer, A. Myrthue, Mol. Cancer Ther. 2004, 3, 373. [10] R. Bouillon, A. Verstuyf, C. Mathieu, S. Van cromphaut, R. Masuyama, P. Dehaes, et al., Best. Pract. Res. Clin. Endocrinol. Metab. 2006, 20, 627. [11] A. J. Brown, A. Dusso, E. Slatopolsky, Am. J. Physiol. 1999, 277, F157. [12] F. C. Campbell, H. Xu, M. El‐Tanani, P. Crowe, V. Bingham, Biochem. Pharmacol. 2010, 79, 1. [13] L. L. Issa, G. M. Leong, J. A. Eisman, Inflamm. Res. 1998, 47, 451. [14] N. Lopes, B. Sousa, D. Martins, M. Gomes, D. Vieira, L. A. Veronese, et al., BMC Cancer 2010, 10, 483. [15] S. Segaert, R. Bouillon, Curr. Opin. Clin. Nutr. Metab. Care 1998, 1, 347. [16] N. Rochel, J. M. Wurtz, A. Mitschler, B. Klaholz, D. Moras, Mol. Cell 2000, 5, 173. [17] P. W. Rose, A. Prlić, C. Bi, W. F. Bluhm, C. H. Christie, S. Dutta, et al., Nucleic Acids Res. 2015, 43, 345. [18] C. Carlberg, F. Molnár, A. Mouriño, Expert Opin. Ther. Pat. 2012, 22, 417. [19] C. Leyssens, L. Verlinden, A. Verstuyf, Endocr. Relat. Cancer 2013, 20, 31. [20] J. R. Muindi, R. A. Modzelewski, Y. Peng, D. L. Trump, C. S. Johnson, Oncology 2004, 66, 62. [21] D. G. Salomón, S. M. Grioli, M. Buschiazzo, E. Mascaró, C. Vitale, G. Radivoy, et al., ACS Med. Chem. Lett. 2011, 2, 503. [22] Z. Gándara, M. Pérez, D. G. Salomón, M. J. Ferronato, M. E. Fermento, A. C. Curino, et al., Bioorg. Med. Chem. Lett. 2012, 22, 6276. [23] M. J. Ferronato, D. G. Salomón, M. E. Fermento, N. A. Gandini, A. López Romero, M. L. Rivadulla, et al., Arch. Pharm. 2015, 348, 315. [24] U. Windberger, A. Bartholovitsch, R. Plasenzotti, K. J. Korak, G. Heinze, Exp. Physiol. 2003, 88, 431. [25] OECD 1995. OECD Test Guideline for Testing of Chemicals, Section 4, Health Effects. OECD Guideline No. 407, Updated July 27, 1995. [26] I. Bendik, A. Friedel, F. F. Roos, P. Weber, M. Eggersdorfer, Front. Physiol. 2014, 248, 1. [27] I. Guessous, BioMed. Res. Int. 2015, 2015, 563403. [28] ChemAxon Ltd. MarvinSketch v.6.31. http://www.chemaxon.com [29] M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Rob, J. R. Cheeseman, et al. 2003, Gaussian 03. [30] M. McGann, J. Chem. Inf. Model. 2011, 51, 578. [31] Fred 3.0.0: OpenEye Scientific Software. http://www.eyesopen.com [32] P. C. D. Hawkins, A. Nicholls, J. Chem. Inf. Model. 2012, 52, 2919. [33] Omega.2.4.3: OpenEye Scientific Software. http://www. eyesopen.com [34] B. R. Miller 3rd, T. D. McGee, J. M. Swails, N. Homeyer, H. Gohlke, A. E. Roitberg, J. Chem. Theor. Comput. 2012, 8, 3314. [35] B. Bandera Merchan, S. Morcillo, G. Martin‐Nuñez, F. J. Tinahones, M. Macías‐González, J. Steroid Biochem. Mol. Biol. 2017, 167, 203. [36] J. M. Wierzbicka, A. Binek, T. Ahrends, J. D. N. Ka, A. Szydłowska, Ł. Turczyk, et al., Int. J. Oncol. 2015, 47, 1084. [37] VIDA 4.2.1: OpenEye Scientific Software. http://www.eyesopen.com [38] R. A. Laskowski, M. B. Swindells, J. Chem. Inf. Model. 2011, 51, 2778. [39] R. Salomon‐Ferrer, D. A. Case, R. C. Walker, Wiley Interdiscip. Rev. Comput. Mol. Sci. 2013, 3, 198. [40] J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, D. A. D.A., J. Comput. Chem. 2004, 25, 1157. [41] V. Hornak, R. Abel, A. Okur, B. Strockbine, A. Roitberg, C. Simmerling, RO215535 ,Proteins 2006, 65, 712.
[42] W. Humphrey, A. Dalke, K. Schulten, J. Mol. Graph. 1996, 14, 33.