Synthesis and Biological Evaluation of New Isoxazolyl Steroids as Anti-Prostate Cancer Agents

Steroids with a nitrogen-containing heterocycle in the side chain are known as effective inhibitors of androgen signaling and/or testosterone biosynthesis, thus showing beneficial effects for the treatment of prostate cancer. In this work, a series of 3β-hydroxy-5-ene steroids, containing an isoxazole fragment in their side chain, was synthesized. The key steps included the preparation of Weinreb amide, its conversion to acetylenic ketones, and the 1,2- or 1,4-addition of hydroxylamine, depending on the solvent used. The biological activity of the obtained compounds was studied in a number of tests, including their effects on 17α-hydroxylase and 17,20-lyase activity of human CYP17A1 and the ability of selected compounds to affect the downstream androgen receptor signaling. Three derivatives diminished the transcriptional activity of androgen receptor and displayed reasonable antiproliferative activity. The candidate compound, 24j (17R)-17-((3-(2-hydroxypropan-2-yl)isoxazol-5-yl)methyl)-androst-5-en-3β-ol, suppressed the androgen receptor signaling and decreased its protein level in two prostate cancer cell lines, LNCaP and LAPC-4. Interaction of compounds with CYP17A1 and the androgen receptor was confirmed and described by molecular docking.


Introduction
Prostate cancer (PCa) is the most common cancer in men in developed countries. More than 80 years have passed since Charles Huggins showed that a decrease in androgen levels in patients with PCa causes tumor regression [1]. The androgenic pathway remains the main target of prostate cancer therapies-it plays a major role in the formation and progression of this type of cancer [2]. Therapy has aimed at reducing the content of testosterone in the blood, which can significantly slow down the process of tumor development and alleviate the patient's condition. Therefore, a number of drugs are used to block the synthesis of androgens in the testes or adrenal cortex as an accepted alternative to surgical intervention (orchiectomy).
The most important step in the biosynthesis of androgens is the conversion of pregnenolone to 17α-OH-pregnenolone, and then to dehydroepiandrosterone (DHEA), secreted by the testes and adrenal cortex [3]. Both reactions proceed with the participation of cytochrome P450 CYP17A1, which combines the functions of 17α-hydroxylase and 17,20-lyase. In 2011, a new CYP17A1 inhibitor, abiraterone, was approved, effective for the treatment of prostate cancer, insensitive to hormone therapy, and reducing the level of testosterone in the blood [4]. Thus, abiraterone (Figure 1), which is a pyridine derivative of DHEA, inhibits two key reactions in the androgen synthesis pathway. The optimal CYP17A1 inhibitor should have significant effect on 17,20-lyase activity, with moderate or no effect towards 17α-hydroxylase activity of the enzyme, to modulate sex steroid biosynthesis with minimal effect on glucocorticoid hormones biosynthesis [5].
Galeterone, the most advanced among them and having a multiple mechanism of action, has reached phase III clinical trials [19].
To date, previous studies have shown that steroids with 5-membered rings containing one nitrogen and one oxygen (oxazole or isoxazole) are of great interest in the development of drugs for the treatment of prostate cancer [6,11,12,14,[20][21][22][23][24][25]. Thus, isoxazole 2 showed potent and non-competitive inhibition of human microsomal 17β-hydroxylase/C17,20-lyase, with an IC50 value of 59 nM, and demonstrated potent and competitive inhibition of 5α-reductase in human prostate microsomes with an IC50 value of 33 nM [21]. It was also shown that 1, at a concentration of 5 μM, exhibits antiandrogenic activity in human prostate cancer cell lines (e.g., LNCaP), preventing the binding of labeled synthetic androgen R1881 (5 nM) to the androgen receptor (AR). Compound 2 had a significant effect on the growth of LNCaP and PC-3 cells, commensurate with that of galeterone [11]. It should be noted that 2 showed no inhibitory potency towards CYP17A1, thus confirming that inhibition of this enzyme is not the only mechanism of anticancer action of such steroids. Obviously, further studies of new nitrogen-containing steroids, in particular the investigation of their effect on various signaling pathways involved in the pathological processes of tumor development, are relevant and of great interest. In this regard, the present paper aims (i) to develop synthesis of a series of novel steroidal isoxazoles 3a,b and (ii) to During the development of new inhibitors of CYP17A1, a large number of androstane and pregnane derivatives have been introduced containing pyridyl-, picolidine-, pyrazolyl-, imidazolyl-, triazolyl-, isoxazolinyl-, dihydrooxazolinyl-, tetrahydrooxazolinyl-, benzimidazolyl-, and carbamoyl-substituents, mainly in positions 16,17, and 22 [6][7][8][9][10][11][12][13][14][15][16][17][18]. Galeterone, the most advanced among them and having a multiple mechanism of action, has reached phase III clinical trials [19].
To date, previous studies have shown that steroids with 5-membered rings containing one nitrogen and one oxygen (oxazole or isoxazole) are of great interest in the development of drugs for the treatment of prostate cancer [6,11,12,14,[20][21][22][23][24][25]. Thus, isoxazole 2 showed potent and non-competitive inhibition of human microsomal 17β-hydroxylase/C17,20-lyase, with an IC 50 value of 59 nM, and demonstrated potent and competitive inhibition of 5α-reductase in human prostate microsomes with an IC 50 value of 33 nM [21]. It was also shown that 1, at a concentration of 5 µM, exhibits antiandrogenic activity in human prostate cancer cell lines (e.g., LNCaP), preventing the binding of labeled synthetic androgen R1881 (5 nM) to the androgen receptor (AR). Compound 2 had a significant effect on the growth of LNCaP and PC-3 cells, commensurate with that of galeterone [11]. It should be noted that 2 showed no inhibitory potency towards CYP17A1, thus confirming that inhibition of this enzyme is not the only mechanism of anticancer action of such steroids.
Obviously, further studies of new nitrogen-containing steroids, in particular the investigation of their effect on various signaling pathways involved in the pathological processes of tumor development, are relevant and of great interest. In this regard, the present paper aims (i) to develop synthesis of a series of novel steroidal isoxazoles 3a,b and (ii) to carry out studies of their effects on the 17α-hydroxylase and 17,20-lyase activity of human CYP17A1 and the ability of selected compounds to affect the downstream androgen receptor signaling. The synthesis of target compounds was initiated with ester 11, obtained in two steps from androstenolone [29]. Initially, the possibility of a one-step conversion of 12 into 18, described for fluoroketones [26] and consisting in the addition of lithium acetylenides to esters in the presence of boron trifluoride etherate, was studied (Scheme 2). However, the reaction of 12 with a lithium salt of 13 resulted in the formation of a complex mixture of products.
Next, an attempt was made to obtain ynone 18 using an approach based on the conversion of ester 12 to aldehyde 16. Its reaction with the lithium salt of 13 gave a mixture of isomeric alcohols in 17, which was oxidized in the last stage to give the target ynone 18. The obvious disadvantage of this method was the necessity to accomplish a multistep reaction procedure. In this connection, the possibility of using Weinreb amides was studied. This approach showed good results for the preparation of ynone 18 via amide 14 and was further used for the synthesis of all other α,β-acetylenic ketones. The synthesis of target compounds was initiated with ester 11, obtained in two steps from androstenolone [29]. Initially, the possibility of a one-step conversion of 12 into 18, described for fluoroketones [26] and consisting in the addition of lithium acetylenides to esters in the presence of boron trifluoride etherate, was studied (Scheme 2). However, the reaction of 12 with a lithium salt of 13 resulted in the formation of a complex mixture of products.
The regioselectivity of hydroxylamine addition to α,β-acetylenic ketones is highly dependent on the reaction conditions. The optimal conditions for the conjugated 1,4addition were found using a mixture of organic solvents with water [28]. The enamine, formed as a result of conjugated addition, undergoes intramolecular cyclization to form 5-hydroxy-4,5-dihydroisoxazoles. We used the reaction conditions (water-THF, NaHCO 3 as a base) proposed in [26]; in this case, the hydroxyisoxazoline 22a-i was obtained in a 56-88% yield.
The next stage involved the dehydration of 22a-i to form the corresponding isoxazole 23a-i. The reaction proceeded relatively smoothly in the case of the derivative 22d-i; however, isoxazoles 23b,c were obtained only in 9 and 45% yields, respectively. Simultaneously, compounds 26b,c (44-50%) were isolated from the reaction mixture (Scheme 4). The possible mechanism of their formation can be explained as follows. 5-Hydroxy-4,5dihydroisoxazoles 22b,c are expected to exist in equilibrium with enehydroxylamines 21b,c. The latter, as a result of reaction with CDI, can give derivatives 25b,c. It is known that such derivatives can undergo 3,3-sigmatropic rearrangement [30,31]. In the case of 25b,c, such a rearrangement resulted in the formation of substituted enaminoketones (26b,c). Their structures were determined by spectral methods, including two-dimensional NMR experiments. Signals at δ 197.1, 134.2, and 157.2 in the 13 C NMR spectrum of 26c were assigned to C-22, C-23, and C-24, respectively. The connectivity of the side chain was established by the key HMBC correlations: H-20 and H-17 correlated to C-22; H-25, H-26, and H-27 correlated to C-24; and H-20 correlated to C-23.
Another direction of the reaction of 5-hydroxy-4,5-dihydroisoxazoles with CDI was found in the case of compound 22a. In addition to the target isoxazole 23a (51%), β-oxonitrile 31 was also isolated in 40% yield. A possible mechanism of its formation is shown in Scheme 5. It is assumed that the enehydroxylamine 21a is converted to β-ketoxime 29, which then reacts with CDI to form the imidazole derivative 30. The latter loses imidazole carboxylic acid in a six-membered transition state [32] to form β-oxonitrile 31.
such derivatives can undergo 3,3-sigmatropic rearrangement [30,31]. In the case of 25b,c, such a rearrangement resulted in the formation of substituted enaminoketones (26b,c). Their structures were determined by spectral methods, including two-dimensional NMR experiments. Signals at δ 197.1, 134.2, and 157.2 in the 13 C NMR spectrum of 26c were assigned to C-22, C-23, and C-24, respectively. The connectivity of the side chain was established by the key HMBC correlations: H-20 and H-17 correlated to C-22; H-25, H-26, and H-27 correlated to C-24; and H-20 correlated to C-23. Another direction of the reaction of 5-hydroxy-4,5-dihydroisoxazoles with CDI was found in the case of compound 22a. In addition to the target isoxazole 23a (51%), β-oxonitrile 31 was also isolated in 40% yield. A possible mechanism of its formation is shown in Scheme 5. It is assumed that the enehydroxylamine 21a is converted to β-ketoxime 29, which then reacts with CDI to form the imidazole derivative 30. The latter loses imidazole carboxylic acid in a six-membered transition state [32] to form β-oxonitrile 31. Scheme 5. Possible mechanism for the formation of ketonitrile 31.
The removal of protective groups completed the synthesis of target isoxazoles 24 containing a steroid at C-5 of the isoxazole heterocycle. The reaction was carried out by treating the esters 23a-i with TBAF or HF. The latter option is preferred for compound 24h, additionally containing tetrahydropyranyl protection.
Attempts were made to carry out some transformations of the sigmatropic rearrangement product 26c in order to obtain compounds suitable for biological testing (Scheme 4). Removal of the silyl protective group proceeded smoothly, without affecting the functional groups in the side chain, to form alcohol 27. Removal of the imidazole-carboxylic fragment was expected to be achieved under alkaline hydrolysis conditions. However, the reaction led to compound 28, containing an oxazolone heterocycle. An attempt to remove the silyl group in 28 (Bu4NF/THF) gave a complex mixture of products.  Another direction of the reaction of 5-hydroxy-4,5-dihydroisoxazoles with CDI was found in the case of compound 22a. In addition to the target isoxazole 23a (51%), β-oxonitrile 31 was also isolated in 40% yield. A possible mechanism of its formation is shown in Scheme 5. It is assumed that the enehydroxylamine 21a is converted to β-ketoxime 29, which then reacts with CDI to form the imidazole derivative 30. The latter loses imidazole carboxylic acid in a six-membered transition state [32] to form β-oxonitrile 31. The removal of protective groups completed the synthesis of target isoxazoles 24 containing a steroid at C-5 of the isoxazole heterocycle. The reaction was carried out by treating the esters 23a-i with TBAF or HF. The latter option is preferred for compound 24h, additionally containing tetrahydropyranyl protection.
Attempts were made to carry out some transformations of the sigmatropic rearrangement product 26c in order to obtain compounds suitable for biological testing (Scheme 4). Removal of the silyl protective group proceeded smoothly, without affecting the functional groups in the side chain, to form alcohol 27. Removal of the imidazole-carboxylic fragment was expected to be achieved under alkaline hydrolysis conditions. However, the reaction led to compound 28, containing an oxazolone heterocycle. An attempt to remove the silyl group in 28 (Bu4NF/THF) gave a complex mixture of products. The removal of protective groups completed the synthesis of target isoxazoles 24 containing a steroid at C-5 of the isoxazole heterocycle. The reaction was carried out by treating the esters 23a-i with TBAF or HF. The latter option is preferred for compound 24h, additionally containing tetrahydropyranyl protection.
Attempts were made to carry out some transformations of the sigmatropic rearrangement product 26c in order to obtain compounds suitable for biological testing (Scheme 4). Removal of the silyl protective group proceeded smoothly, without affecting the functional groups in the side chain, to form alcohol 27. Removal of the imidazole-carboxylic fragment was expected to be achieved under alkaline hydrolysis conditions. However, the reaction led to compound 28, containing an oxazolone heterocycle. An attempt to remove the silyl group in 28 (Bu 4 NF/THF) gave a complex mixture of products.
Simultaneous removal of both protecting groups in 24i gave the diol 32 (Scheme 6). Compound 24i contains a functional group in the isoxazole core that can be used for the synthesis of other derivatives, which was demonstrated in the synthesis of azide 36. Selective removal of the tetrahydropyranyl protecting group was achieved by reaction of 24i with magnesium bromide diethyl etherate [33]. The tosylation reaction of 33 gave chloride 37 instead of the expected tosylate. The desired product 35 was obtained via S N 2 displacement of primary mesylate 34 with the azide group.
Ynone 20d was used as a model compound to study suitable conditions for the preparation of isoxazoles 40. Its reaction with hydroxylamine in aqueous methanol in the presence of NaHCO 3 [27,28] gave oxime 39 as a mixture (1:1) of E/Z-isomers (Scheme 7). The next step in the synthesis of isoxazole 40 was the gold-catalyzed cycloisomerization of acetylenic oximes [34]. The desired product 40d was obtained from 39, but in a moderate 46% yield, as only the Z-isomer underwent the cyclization under these conditions. At the same time, it was found that prolonged heating of the reaction mixture at the stage of hydroxylamine addition led directly to the formation of isoxazoles without any catalysis.
Simultaneous removal of both protecting groups in 24i gave the diol 32 (Scheme 6). Compound 24i contains a functional group in the isoxazole core that can be used for the synthesis of other derivatives, which was demonstrated in the synthesis of azide 36. Selective removal of the tetrahydropyranyl protecting group was achieved by reaction of 24i with magnesium bromide diethyl etherate [33]. The tosylation reaction of 33 gave chloride 37 instead of the expected tosylate. The desired product 35 was obtained via SN2 displacement of primary mesylate 34 with the azide group. Scheme 6. Synthesis of the isoxazoles 32, 36, and 38.
Ynone 20d was used as a model compound to study suitable conditions for the preparation of isoxazoles 40. Its reaction with hydroxylamine in aqueous methanol in the presence of NaHCO3 [27,28] gave oxime 39 as a mixture (1:1) of E/Z-isomers (Scheme 7). The next step in the synthesis of isoxazole 40 was the gold-catalyzed cycloisomerization of acetylenic oximes [34]. The desired product 40d was obtained from 39, but in a moderate 46% yield, as only the Z-isomer underwent the cyclization under these conditions. At the same time, it was found that prolonged heating of the reaction mixture at the stage of hydroxylamine addition led directly to the formation of isoxazoles without any catalysis. For this reason, the transformation of the remaining ynones 20a,e-i was carried out in one step without the isolation of the intermediate acetylenic oximes (Scheme 8). The yield of isoxazoles 40a,e-i was 40-82%, depending on the substituent R. Removal of the silyl protective group was performed out by treatment with Bu4NF or, in the case of compounds 40h,i, containing additionally tetrahydropyranyl ether, with HF in a mixture of THF-MeCN. Ynone 20d was used as a model compound to study suitable conditions for the preparation of isoxazoles 40. Its reaction with hydroxylamine in aqueous methanol in the presence of NaHCO3 [27,28] gave oxime 39 as a mixture (1:1) of E/Z-isomers (Scheme 7). The next step in the synthesis of isoxazole 40 was the gold-catalyzed cycloisomerization of acetylenic oximes [34]. The desired product 40d was obtained from 39, but in a moderate 46% yield, as only the Z-isomer underwent the cyclization under these conditions. At the same time, it was found that prolonged heating of the reaction mixture at the stage of hydroxylamine addition led directly to the formation of isoxazoles without any catalysis. For this reason, the transformation of the remaining ynones 20a,e-i was carried out in one step without the isolation of the intermediate acetylenic oximes (Scheme 8). The yield of isoxazoles 40a,e-i was 40-82%, depending on the substituent R. Removal of the silyl protective group was performed out by treatment with Bu4NF or, in the case of compounds 40h,i, containing additionally tetrahydropyranyl ether, with HF in a mixture of THF-MeCN. For this reason, the transformation of the remaining ynones 20a,e-i was carried out in one step without the isolation of the intermediate acetylenic oximes (Scheme 8). The yield of isoxazoles 40a,e-i was 40-82%, depending on the substituent R. Removal of the silyl protective group was performed out by treatment with Bu 4 NF or, in the case of compounds 40h,i, containing additionally tetrahydropyranyl ether, with HF in a mixture of THF-MeCN.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1.

Biology
Biological studies included analysis of the interaction of the prepared compounds with the CYP17A1 active site, testing their effect towards the AR-transcriptional activity and evaluation of their ability to influence the downstream AR signaling. Compounds used in one or more biological tests are shown in Table 1. As the first step in the analysis of the interaction of compounds with the CYP17A1 active site, we performed spectrophotometric titration of selected compounds. Progesterone with Kd app < 1 µM was used as a positive control for the substrate-like ligand (type I). As a positive control for an inhibitor-like ligand (type II), abiraterone with a Kd app < 1 µM was used. DMSO or ethanol was used as a negative control. For the negative controls, no type I or type II spectral responses were observed. Analysis of binding of series of compounds (24a,d,g,j, 32, 36, 38) with human CYP17A1 showed that only four compounds were able to bind to the active site of human CYP17A1 (32 with Kd app -13.41 ± 2.38 µM, 24j with Kd app -1.90 ± 0.23 µM, 24d with Kd app -1.50 ± 0.21 µM, and 24a with Kd app -0.13 ± 0.02 µM). However, all these compounds show type I (substratelike) spectral response, which indicates their potentially low inhibitory ability against this enzyme.
To evaluate the effect of isoxazoles on a potential molecular target, CYP17A1, we performed an inhibitory assay using an in vitro reconstituted system containing recombinant human CYP17A1. We analyzed the inhibitory effect of the compounds (50 µM-final concentration) on two types of reactions catalyzed by CYP17A1: 17α-hydroxylase activity and 17,20-lyase activity. There was almost no inhibition of 17,20-lyase activity of human CYP17A1 with the compound 24j (11% of enzyme inhibition at 50 µM of compound 24j). Most isoxazoles were shown to have a moderate inhibitory effect on human CYP17A1 activity ( Table 2). The maximum inhibition of 17,20-lyase activity was found for isoxazole 41a containing no substituent at C-5 of the heterocyclic ring. A similar inhibitory effect was observed for pyridine derivative 41f. It should be noted that compound 41a showed a minimal inhibitory capacity for 17α-hydroxylase activity, being the most 17,20 lyase selective, which is important for the development of the next generation CYP17 targeted drugs [35].

The Effect of Derivatives on AR Transcriptional Activity and Viability of PCa Cells
Based on the structural similarity of novel derivatives with galeterone and other published compounds (abiraterone, 3, and [6]), we tested the effect of our compounds towards the AR-transcriptional activity. Compounds were evaluated using an AR-dependent reporter cell line (ARE14) with a firefly-luciferase gene under the control of an androgen response element [36].
As shown in Table 3, within the studied isoxazoles, three AR antagonists with moderate activity were found (reduced R1881-stimulated AR transcriptional activity to ≤ 50% at 50 µM concentration), namely 24j, 32, and 41a. Despite that, these derivatives acted as AR-antagonists in dose-dependent manner, and none of them were able to overcome the activity of the standard steroidal antagonist galeterone (≈35% of activity at 10 µM concentration). Based on the obtained results, it is evident that derivatives bearing only unsubstituted isoxazole (41a), or isoxazoles substituted with small polar substituent (-CH 3 -OH in 32, t-butyl in 24j), were able to decrease the AR-transcriptional activity, while isoxazoles bearing longer unsaturated (24d), or bulky aromatic substituents (24e, 24g), were inactive. Importantly, none of novel derivatives displayed AR-agonist activity in the chosen concentrations (Table 3).
Antiproliferative properties of all novel steroids were tested in two AR-positive prostate cancer cell lines (LNCaP and LAPC-4) and one AR-negative cell line (DU145) using the Alamar-blue assay after 72 h treatment. Antiproliferative activities of the most potent derivatives 24j and 32 displayed mid-micromolar values (in agreement to ARantagonist assay) in both AR-positive PCa cell lines, while no targeting of the AR-negative DU145 cells was observed. Compound 36 displayed reasonable antiproliferative activity only in LAPC-4 cell line.
We further evaluated the potency of the most active derivative 24j. We analyzed its effect on AR transcriptional activity in a broad concentration range and found the IC 50 value = 21.11 ± 1.07 µM (Figure 2A) while IC 50 = 7.59 µM for galeterone. On the other hand, galeterone displayed worse antiproliferative activities and effects related to AR signaling. Importantly, no clear agonist activity was observed for 24j in tested concentrations ( Figure 2A).

Targeting the AR Signaling Pathway
Further, we evaluated the ability of compounds 24j, 32, and 41a to influence the downstream AR signaling (levels of known transcriptional targets PSA and Nkx3.1) in LAPC-4 and LNCaP cell lines after R1881 stimulation. Western blot analysis ( Figure 3) showed that 24j and 32 were able to markedly suppress R1881 stimulated S81-phosphorylation in both LAPC-4 and LNCaP cell lines after 24 h. Observed effects were accompanied by a profound decrease in Nkx3.1 and PSA protein levels in LAPC-4, while were only limited in LNCaP cells (Figure 3).
Candidate compounds were further tested in the same PCa cell lines (without R1881 activation) for a longer period to monitor the effects on AR stability. Compounds 24j and 32 (12.5 μM) induced a significant decrease in Nkx3.1 and PSA levels in both LAPC-4 and LNCaP after 48 h. Moreover, they both induced a significant decrease in AR protein level that was comparable to galeterone's effect. We also performed colony formation assay (CFA) within 10 days to evaluate the prolonged antiproliferative potency of 24j in the LAPC-4 cell line. CFA is frequently used for the validation of PCa cell lines growth because of their high doubling time in culture. Our compound, 24j, significantly inhibited the formation of cell colonies in a dose-dependent manner after 10 days in LAPC-4, already from a 1.56 µM concentration ( Figure 2B).

Targeting the AR Signaling Pathway
Further, we evaluated the ability of compounds 24j, 32, and 41a to influence the downstream AR signaling (levels of known transcriptional targets PSA and Nkx3.1) in LAPC-4 and LNCaP cell lines after R1881 stimulation. Western blot analysis ( Figure 3) showed that 24j and 32 were able to markedly suppress R1881 stimulated S81-phosphorylation in both LAPC-4 and LNCaP cell lines after 24 h. Observed effects were accompanied by a profound decrease in Nkx3.1 and PSA protein levels in LAPC-4, while were only limited in LNCaP cells ( Figure 3).

Molecular Docking into the Active Site of CYP17A1 and into the AR-LBD
The binding of candidate compounds into their cellular targets was evaluated by rigid molecular docking into the crystal structure of human CYP17A1 co-crystalized with abiraterone and heme (PDB:3RUK) and by flexible docking into the AR ligand-binding domain (LBD) from the crystal structure with DHT (PDB:2PIV).
The best binding pose of 41a in the active site of CYP17A1 was oriented in nearly the same pose as abiraterone and showed similar binding energy (ΔGVina = −12.6 kcal/mol and −13.0 kcal/mol, respectively) ( Figure 4A,B). The isoxazole ring is oriented towards the Fe 2+ central ion of heme, similar to the pyridine ring in abiraterone. The most promising compound, 24j, was modelled into the CYP17A1 as well, with a similar pose as abiraterone and 41a, but with lower binding energy (ΔGVina = −10.1 kcal/mol, picture not shown).
In the case of the AR-LBD structure, two key amino acid residues in both extremities of the cavity (Arg752 and Thr877) were set as flexible. The docking of 24j revealed a pose with extensive binding in AR-LBD, with binding energy comparable to galeterone (ΔGVina = −10.5 kcal/mol and −10.8 kcal/mol, respectively ( Figure 4C,D). The position of the steroid core is conserved. C3-OH on the A-ring forms a typical hydrogen bond with Arg752, while the oxygen atom of the isoxazole ring forms a hydrogen bond with Thr877 at the other extremity of the LBD-cavity. The steroid core is further positioned by hydrophobic interactions with Gln711, Met745, Met746, and Leu 704. The tert-butyl substituent could be hydrophobically binded with Leu701, Phe647, and Leu880, which could be a key for the activity and selectivity of 24j ( Figure 4C). Candidate compounds were further tested in the same PCa cell lines (without R1881 activation) for a longer period to monitor the effects on AR stability. Compounds 24j and 32 (12.5 µM) induced a significant decrease in Nkx3.1 and PSA levels in both LAPC-4 and LNCaP after 48 h. Moreover, they both induced a significant decrease in AR protein level that was comparable to galeterone's effect.

Molecular Docking into the Active Site of CYP17A1 and into the AR-LBD
The binding of candidate compounds into their cellular targets was evaluated by rigid molecular docking into the crystal structure of human CYP17A1 co-crystalized with abiraterone and heme (PDB:3RUK) and by flexible docking into the AR ligand-binding domain (LBD) from the crystal structure with DHT (PDB:2PIV).
The best binding pose of 41a in the active site of CYP17A1 was oriented in nearly the same pose as abiraterone and showed similar binding energy (∆G Vina = −12.6 kcal/mol and −13.0 kcal/mol, respectively) ( Figure 4A,B). The isoxazole ring is oriented towards the Fe 2+ central ion of heme, similar to the pyridine ring in abiraterone. The most promising compound, 24j, was modelled into the CYP17A1 as well, with a similar pose as abiraterone and 41a, but with lower binding energy (∆G Vina = −10.1 kcal/mol, picture not shown).

Chemistry
Commercially available reagents were used without further purification. If necessary, solvents were distilled and dried before use by standard methods. Column chromatography was performed through silica gel (200-300 mesh). Thin layer chromatography (TLC) was performed using Silica gel 60 F254 plates and visualized using UV light or phosphomolybdic acid. 1 H and 13 C NMR spectra were recorded in CDCl3, on a Bruker AVANCE 500 spectrometer. Chemical shifts in 1 H NMR spectra are reported in parts per million (ppm) on the δ scale from an internal standard of residual non-deuterated solvent in CDCl3 (7.26 ppm). Data for 1 H NMR are reported as follows: chemical shift, multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, br = broad), coupling constant in Hertz (Hz), and integration. Data for 13 C NMR spectra are reported in terms of chemical shift in ppm from the central peak of CDCl3 (77.16 ppm). High resolution mass spectrometry (HRMS) analysis was performed using a Q Exactive HFX (Thermo Scientific) mass spectrometer in ESI ionization mode.
3.1.1. Methyl 2-(3β-((tert-butyldimethylsilyl)oxy)-androst-5-en-17-yl)acetate (12) A solution of alcohol 11 (prepared from androstenolone in two steps according to [29]) (1.92 g, 5.55 mmol), TBSCl (1.25 g, 8.29 mmol), and imidazole (838 mg, 12.3 mmol) in dry DMF (8 mL) was stirred at 90 °C for 12 h. After the reaction was completed, the mixture was diluted with water, the organic layer was separated, and the reaction product was extracted from the aqueous layer with PE. The combined organic extracts were dried with sodium sulfate. Then, the solvent was removed under reduced pressure, and the resulting residue was purified by column chromatography on SiO2 (PE/EtOAc, 20:1) to In the case of the AR-LBD structure, two key amino acid residues in both extremities of the cavity (Arg752 and Thr877) were set as flexible. The docking of 24j revealed a pose with extensive binding in AR-LBD, with binding energy comparable to galeterone (∆G Vina = −10.5 kcal/mol and −10.8 kcal/mol, respectively ( Figure 4C,D). The position of the steroid core is conserved. C3-OH on the A-ring forms a typical hydrogen bond with Arg752, while the oxygen atom of the isoxazole ring forms a hydrogen bond with Thr877 at the other extremity of the LBD-cavity. The steroid core is further positioned by hydrophobic interactions with Gln711, Met745, Met746, and Leu 704. The tert-butyl substituent could be hydrophobically binded with Leu701, Phe647, and Leu880, which could be a key for the activity and selectivity of 24j ( Figure 4C).

Chemistry
Commercially available reagents were used without further purification. If necessary, solvents were distilled and dried before use by standard methods. Column chromatography was performed through silica gel (200-300 mesh). Thin layer chromatography (TLC) was performed using Silica gel 60 F 254 plates and visualized using UV light or phosphomolybdic acid. 1 H and 13 C NMR spectra were recorded in CDCl 3 , on a Bruker AVANCE 500 spectrometer. Chemical shifts in 1 H NMR spectra are reported in parts per million (ppm) on the δ scale from an internal standard of residual non-deuterated solvent in CDCl 3 (7.26 ppm). Data for 1 H NMR are reported as follows: chemical shift, multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, br = broad), coupling constant in Hertz (Hz), and integration. Data for 13 C NMR spectra are reported in terms of chemical shift in ppm from the central peak of CDCl 3 (77.16 ppm). High resolution mass spectrometry (HRMS) analysis was performed using a Q Exactive HFX (Thermo Scientific) mass spectrometer in ESI ionization mode.
3.1.1. Methyl 2-(3β-((tert-butyldimethylsilyl)oxy)-androst-5-en-17-yl)acetate (12) A solution of alcohol 11 (prepared from androstenolone in two steps according to [29]) (1.92 g, 5.55 mmol), TBSCl (1.25 g, 8.29 mmol), and imidazole (838 mg, 12.3 mmol) in dry DMF (8 mL) was stirred at 90 • C for 12 h. After the reaction was completed, the mixture was diluted with water, the organic layer was separated, and the reaction product was extracted from the aqueous layer with PE. The combined organic extracts were dried with sodium sulfate. Then, the solvent was removed under reduced pressure, and the resulting residue was purified by column chromatography on SiO 2 (PE/EtOAc, 20:1) to yield ether 12 (2.35 g, 92%) as an oil. 1 (18) A 2M solution of BuLi in hexanes (0.7 mL, 1.4 mmol) was added to a cooled −70 • C solution of tert-butyldimethyl(prop-2-yn-1-yloxy)silane (13) (prepared according to [37], 211 mg, 1.24 mmol) in THF (4 mL). After 15 min, a solution of aldehyde 16 (380 mg, 0.88 mmol) in THF (2.5 mL) was added to the reaction mixture. It was stirred for 20 min at −70 • C, then the cooling bath was removed, and the mixture was allowed to get ambient temperature. NH 4 Cl (150 mg) was added, then the mixture was diluted with water and extracted with EtOAc. The organic layers were dried over Na 2 SO 4 , then evaporated, and the residue was chromatographed on SiO 2 to give a mixture of isomeric at C-22 alcohol 17 (467 mg), which was used in the next step without further purification.
A mixture of alcohol 17 (467 mg, 0.78 mmol), Dess-Martin reagent (2.30 g, 5.42 mmol), and DCM (15 mL) was stirred under argon at 0 • C for 3 h. Then it was diluted with water and extracted with DCM. The residue after evaporation of the extracts was chromatographed on SiO 2 (PE/EtOAc, 98:2) to give ketone 18 (388 mg, 73% from 16) as an oil. 1 (14) To a stirred suspension of Weinreb salt (555 mg, 5.72 mmol) in dry toluene (10 mL), a 1M solution of Me 3 Al in heptane (5.7 mL, 5.7 mmol) was added dropwise at 0 • C. After stirring for 40 min at this temperature, a solution of ester 12 (1.00 g, 2.17 mmol) in toluene (10 mL) was added dropwise. The cooling bath was removed, and the reaction mixture was allowed to stir overnight at room temperature. A 2N solution of HCl was added on cooling until pH 2 was reached. the mixture was diluted with water and extracted with EtOAc. The combined organic extracts were washed with saturated aqueous NaHCO 3 , saturated NaCl, dried over anhydrous Na 2 SO 4 , and evaporated to dryness. The residue was purified by silica gel column chromatography (PE/EtOAc, 100:0→70:30) to give the Weinreb amide 14 (650 mg, 61%) as an off-white solid. 1

General Procedure for the Synthesis of Alcohols (24a-i)
A 1M solution of silyl ethers 23a-i (1 eq.) and TBAF (1.2 eq.) in THF was kept at room temperature for 24 h. On completion of the reaction, the mixture was diluted with saturated NH 4 Cl and extracted with EtOAc. The combined organic layers were washed with water, brine, dried over Na 2 SO 4 and concentrated under reduced pressure. The residue was purified by silica gel chromatography (PE:EtOAc) to give alcohols 24a-i.

Cell Viability Assay
Cells were seeded into the 96-well tissue culture plates and, on the other day, compounds were added in different concentrations in duplicate for 72 h. Upon treatment, the resazurin solution (Sigma Aldrich, St. Louis, MI, USA) was added for 4 h, and then the fluorescence of resorufin was measured at 544 nm/590 nm (excitation/emission) using a Fluoroskan Ascent microplate reader (Labsystems, Budapest, Hungary). The GI 50 value was calculated from the dose-response curves that resulted from the measurements using GraphPad Prism 5.

Colony Formation Assay
LAPC-4 (10,000 cells per well) were seeded into 6 well plates and cultivated for 2 days. Next, the medium was removed and replaced with fresh medium containing different concentrations of the compound. Cells were cultivated with the compounds for 10 days. After that, the medium was discarded, colonies were fixed with 70% ethanol, washed with PBS, and stained with crystal violet (1% solution in 96% ethanol). Finally, wells were washed with PBS until the bottom was clear and colonies were visible and the photograph was captured.

Immunoblotting
After the treatment, cells were washed twice with PBS, pelleted, and kept frozen in −80 • C. Cells were lysed, as usual, in ice-cold RIPA lysis buffer supplemented with protease and phosphatase inhibitors. Cells were disrupted by ultrasound sonication on ice and clarified by centrifugation at 14,000× g for 30 min. Protein concentration was measured and balanced within samples. Protein solutions were denatured in SDS-loading buffer, and proteins were separated by SDS-PAGE and electroblotted onto nitrocellulose membranes. Membranes were blocked in 4% BSA and incubated overnight with primary antibodies. On the next day, membranes were washed and incubated with secondary antibodies conjugated with peroxidase. Peroxidase activity was detected by SuperSignal West Pico reagents (Thermo Fisher Scientific, Waltham, MA, USA) using a CCD camera LAS-4000 (Fujifilm, Minato, Japan). Primary antibodies purchased from Merck (Darmstadt, Germany): (anti-β-actin, clone C4; anti-phosphorylated AR (S81)) and from Cell Signaling Technology (Danvers, MA, USA) (anti-AR, clone D6F11; anti-PSA/KLK3, clone D6B1; anti-Nkx3.1, clone D2Y1A; anti-rabbit secondary antibody (porcine anti-rabit immunoglobulin serum)). All antibodies were diluted in 4% BSA and 0.1% Tween 20 in TBS.

Molecular Docking
Molecular docking of compounds 3-7j and 6-4z was performed into the crystal structure of CYP17A1 co-crystalised with heme and abiraterone (PDB:3RUK). The abiraterone molecule was extracted from the protein target before docking, for which the protein was set rigid. For molecular docking into the AR-LBD structure, its crystal structure with DHT was used (PDB:2PIV), and two key amino-acid residues in both extremities of the cavity (Arg752 and Thr877) were set flexible. Accuracy of the docking was assured by re-docking of abiraterone and galeterone into the protein targets and comparison with crystal structure or previously published docking poses. The 3D structures of all compounds were prepared, and their energy was minimized by molecular mechanics with Avogadro 1.90.0. Polar hydrogens were added to molecules with the AutoDock Tools program [42], and docking was performed using AutoDock Vina 1.05 [43]. Figures were generated in Pymol ver. 2.0.4 (Schrödinger, LLC, Cambridge, UK).

Conclusions
In summary, in this paper, we present the synthesis and biological studies of steroids containing an isoxazole fragment on their side chain. The presented synthetic approach allowed the preparation of regioisomeric isoxazole derivatives bearing a steroid moiety at both C-3 and C-5 of the heterocycle using common intermediates. Biological studies of the obtained compounds included an examination of their effects on 17α-hydroxylase and 17,20-lyase activity of human CYP17A1 and the ability of selected compounds to influence the downstream AR signaling.
Most of the compounds have a moderate inhibitory effect on the activity of human CYP17A1. The most promising results (predominant inhibitory effect on 17/20-lyase reaction over effect on 17α-hydroxylase activity of CYP17A1) were obtained for the compounds 41a and 41k. These molecules are the most perspective for further optimization. Compounds 41f,g,j also had a predominant effect on the 17,20-lyase reaction of CYP17A1. Moreover, binding and interactions of 41a in CYP17A1 was described using molecular docking and was found nearly identical, compared to abiraterone. Several compounds were further evaluated for their ability to affect the AR transactivation and the viability of several PCa cell lines. Within prepared compounds, three AR antagonists were found to abolish the AR transcriptional activity and the viability of AR-positive PCa cell lines in mid-micromolar concentrations. Candidate compound 24j decreased the AR protein level and blocked its downstream signaling and significantly inhibited colony formation of LAPC-4 cells. Binding of 24j in AR-LBD was described to be similar to galeterone. Overall, the results support the development of novel steroidal derivatives targeting CYP17A1 and AR as anticancer agents in PCa therapy.