Design and synthesis of novel bicalutamide and enzalutamide derivatives as antiproliferative agents for the treatment of prostate cancer

Prostate cancer (PC) is one of the major causes of male death worldwide and the development of new and more potent anti-PC compounds is a constant requirement. Among the current treatments, ( R )- bicalutamide and enzalutamide are non-steroidal androgen receptor antagonist drugs approved also in the case of castration-resistant forms. Both these drugs present a moderate antiproliferative activity and their use is limited due to the development of resistant mutants of their biological target. Insertion of ﬂ uorinated and per ﬂ uorinated groups in biologically active compounds is a current trend in medicinal chemistry, applied to improve their ef ﬁ cacy and stability pro ﬁ les. As a means to obtain such effects, different modi ﬁ cations with per ﬂ uoro groups were rationally designed on the bicalutamide and enzalutamide structures, leading to the synthesis of a series of new antiproliferative compounds. Several new analogues displayed improved in vitro activity towards four different prostate cancer cell lines, while maintaining full AR antagonism and therefore representing promising leads for further development. Furthermore, a series of molecular modelling studies were performed on the AR antagonist conformation, providing useful insights on potential protein-ligand interactions. © 2016 The Authors. Published by Elsevier Masson SAS. This is an open access article under the CC BY license (http:/


Introduction
Prostate cancer (PC) is a leading cause of male death worldwide and it is the most frequently diagnosed cancer among men aged 65e74 [1].The prognosis varies greatly, being highly dependent on a number of factors such as stage of diagnosis, race and age.Currently, PC treatment includes androgen deprivation, surgery, radiation, endocrine therapy and radical prostatectomy.
PC cell growth is strongly dependent on androgens, therefore blocking their effect can be beneficial to the patient's health.Such outcomes can be achieved by antagonism of the androgen receptor (AR) using anti-androgen drugs, which have been extensively explored either alone or in combination with castration [2].Flutamide (Eulexin ® )( 1) (in its active form as hydroxyflutamide (2)), bicalutamide (Casodex ® )( 3), nilutamide (Niladron ® )( 4) and enzalutamide (previously called MDV3100) (Xtandi ® )( 5) are all non-steroidal androgen receptor antagonists approved for the treatment of PC (Fig. 1).In many cases, after extended treatment over several years, these anti-androgens become ineffective and the disease may progress to a more aggressive and lethal form, known as castration resistant prostate cancer (CRPC).The major cause of this progressive disease is the emergence of different mutations on the AR, which cause the anti-androgen compounds to function as agonists, making them tumour-stimulating agents [3].
Among the drugs used for the treatment of PC, bicalutamide and enzalutamide selectively block the action of androgens while presenting fewer side effects in comparison with other AR antagonists [4e6].The structure of these molecules is characterised by the presence of a trifluoromethyl substituted anilide, which appears to be critical for biological activity (Fig. 1).As a means to improve the anti-proliferative activity of these compounds, and in order to exploit the well established potential of the fluorine atom in enhancing the pharmacological properties and drug-like physicochemical characteristics of candidate compounds [7e9],a wide array of diverse new structures has been rationally designed and synthesised, through the introduction of fluoro-, trifluoromethyland trifluoromethoxy groups in diverse positions of both aromatic rings of the parent scaffolds.Our modifications resulted in a marked improvement of in vitro anti-proliferative activities on a range of human PC cell lines (VCap, LNCaP, DU-145 and 22RV1).In addition, we probed full versus partial AR antagonism for our new compounds.

Results and discussion
Trifluoromethyl and trifluoromethoxy functions were systematically inserted on both aromatic rings of bicalutamide and enzalutamide, with the aim to explore the effect of perfluoro groups on their biological activity.As a means to further expand structureactivity relationship studies, the insertion of different linkers was also envisaged, along with the replacement of the bicalutamide methyl group with a trifluoromethyl function.
The main proposed modifications are summarised in Fig. 2.

Chemistry
Several reported methods for the synthesis of racemic bicalutamide were explored to find a rapid methodology that would allow the preparation of a wide range of new derivatives [10e12].Modification and optimization of these procedures led to the development of the synthetic pathway shown in Scheme 1. Phenylacrylamides 12e16 were prepared by reacting the corresponding aniline (7e11) with methacryloyl chloride (6)i n dimethylacetamide (DMA), modifying a reported methodology [10].In particular, due to the presence of electron withdrawing groups (nitro, cyano, trifluoromethyl) in different positions, anilines 7e11 were in some cases (9 and 10 in particular) of low reactivity towards nucleophilic displacement.Synthetic efforts were made to achieve good yields (Supplementary data).Phenylacrylamides 12e16 were converted into the corresponding epoxides 17e21 in the presence of a large excess of hydrogen peroxide and trifluoroacetic anhydride in dichloromethane [11].Opening of the epoxide rings of 17e21 with commercially available phenols and thiophenols gave a series of ethers (27e31) and thioether derivatives (22e26), respectively, in good yields after purification by column chromatography [11].Thioethers 22e25 were finally oxidised to the corresponding sulfones 32e35 using mCPBA, maintaining the temperature at 25 C [12].Racemic bicalutamide (3)was prepared as a positive control following this route.
Since R-bicalutamide is known to be the most active enantiomer [13], chiral synthesis of selected R-bicalutamide analogues was carried out as shown in Scheme 2.( R)-N-Methacryloylproline 37, prepared using (R)-proline (36) and methacryloyl chloride (6), was reacted with N-bromosuccinimide in DMF to afford bromolactone 38 as a single enantiomer [14].Acid hydrolysis of 38 resulted in the formation of bromohydrin acid 39, which was then converted into the corresponding chiral anilide (40e41) [14].Amide derivatives 40e41 were reacted with the sodium salt of different commercial thiophenols in tetrahydrofuran to give, after purification by silica gel chromatography, the desired (R)-thioethers (42e44) [15].R eaction of amide 40 with the sodium salt of 4-cyanophenol gave the reference compound (S)-enobosarm (44e).Oxidation of thioethers 42e43 with mCPBA provided sulfones 45e46, with the desired R absolute configuration [12].Reference (R)-bicalutamide (45a)w a s prepared following this synthetic route.
For one of the most active compounds initially found, 23d, replacement of the central methyl group with an extra trifluoromethyl function was planned and carried out (Scheme 3).
Several methods have been reported for the preparation of enzalutamide, all showing as a synthetic challenge the formation of the N-substituted thiohydantoin ring [17e19].The method selected in this study was a three-step synthesis involving the preparation of different isothiocyanates (54e58), obtained in quantitative yield after treating the corresponding aniline (7e10, 53) with thiophosgene (Scheme 4) [17].Reaction conditions were optimised for the different anilines used, with 9 and 10 requiring higher temperature, longer reaction time and the addition of extra equivalents of thiophosgene due to their low reactivity.Strecker reaction of substituted anilines (11,53) with acetone and trimethylsilyl cyanide generated the desired cyanomines 59e60.Reaction of 59e60 with isothiocyanates 54e58 in DMF followed by the addition of HCl and MeOH gave the desired thiohydantoins 61e65.The last reaction step was characterized by poor yields possibly due to low reactivity and the formation of several side products (mainly due to the degradation of the reagents into the original starting anilines).Moreover, attempted preparation of enzalutamide derivatives 71b and 71d, in which the nitro and cyano groups are removed, was unsuccessful following this route.
Commercially available 3-bromobenzotrifluoride (66) was coupled with 2-aminoisobutyric-acid to obtain amino propionic acid derivative 67, which was then converted into its methyl ester 68 after treatment with iodomethane.Reaction of 68 with different isothiocynates (70b, d) yielded the desired thiohydantoins 71b and 71d.Following this procedure, standard enzalutamide (5) was also prepared.

Antiproliferative assay
Antiproliferative activity of the newly synthesized compounds was initially evaluated with an in vitro 2D monolayer assay using four human prostate cancer cell lines (LNCaP, 22Rv1, VCaP, and DU145).LNCaP, VCaP and 22Rv1 exhibit some androgen sensitivity, whereas DU145 is hormone-insensitive.The assay allows determination of the compound's capacity to inhibit survival and/or cell proliferation.The antiproliferative results are reported in Tables 1   and 2 (absolute IC 50 in mM).
The overall activity of the four reference compounds (3, 5, 45a, 44e) is consistent with previous reported data for some of these specific cell lines and it represents a confirmation of the reliability of the test [20e22].70% of our new derivatives performed better than bicalutamide, either as a racemic mixture or pure R-isomer (3, 45a), significantly improving its antiproliferative activity up to 50fold (overall antiproliferative activity in the four cell lines reported as geometric mean).
Most inhibitors showed concentration-dependent activity against the four prostate cancer cell lines, with mean IC 50 values ranging from 1.6 mMt o>100 mM.Active inhibitors showed sigmoidal concentrationeeffect curves with total cell kill at high concentrations (Supplementary data).Although the overall antiproliferative activity seems to be the result of the whole structure of each single derivative, a general SAR can be identified, considering three main structural components: linker X, aromatic ring A and aromatic ring B. Thioether or ether compounds (X ¼ So rO )a r e associated with better antiproliferative activity than the corresponding sulfone derivatives (X ¼ SO 2 ), with a decrease of effect in the sulfones up to eight fold (i.e.25l and 28l vs 35l; 23d & 28d vs 33d; 42d vs 45d; 22d & 27d vs 32d; 23c & 28c vs 33c).These data are consistent with previous work, in which thioether derivatives of SARMs (selective androgen receptor modulators) were found to be more active than the corresponding sulfones [20].However, different newly prepared sulfones showed improved antiproliferative activity in comparison with the standard bicalutamide, with a decrease of IC 50 values up to two-fold (i.e.33c, 33b, 45b, etc.).Replacement of the sulfur atom with oxygen is in general associated with retained activity (i.e.23c vs 28c; 22c vs 27c), with only few cases of decreased activity in comparison with the thioether analogues (i.e.25l vs 28l; 23d vs 28d), probably due to the combined effect of the substituent on aromatic ring B and the linker.Several new ether derivatives performed better than the Scheme 3. Synthetic strategy used in the synthesis of 52.Reagents and conditions: (a) NaH (1 equiv.),THF, 0 C to RT, 3 h; (b) KCN (1.2 equiv.),25% H 2 SO 4 ,0 C to RT, 20 h; c) HCl, AcOH, reflux, 24 h; (d) 8, SOCl 2 (1.3 equiv.),DMA, RT, 72 h.Scheme 4. Synthetic strategy toward enzalutamide derivatives.Adapted from reported procedure [17].Reagents and conditions: (a) CSCl 2 (1.5 equiv.),NaHCO 3 ,H 2 O, DCM, RT, 24 h; (b) TMSCN (5.1 equiv.),acetone, 80 C, 12 h; (c) DMF, RT 48 h, followed by HCl, MeOH, reflux, 6 h.Scheme 5. Synthetic strategy toward enzalutamide derivatives 71b-d.Adapted from a reported procedure [19].Reagents and conditions: (a) 2-aminoisobutyric acid (1.5 equiv.),K 2 CO 3 (6.5  standard (S)-enobosarm (44e), improving its activity up to 4-fold (i.e.28m, 28l, 27c, etc.).Modifications on aromatic ring A do not appear to significantly affect antiproliferative activity.Structural modifications on the B ring profoundly influence the antiproliferative effect, improving or totally abolishing activity.In particular, replacement of the original fluorine in the para position with a more lipophilic and bulkier trifluoromethyl moiety is associated with an improvement of anti-proliferative activity of an average of two fold (i.e. 3 vs 32b; 45a vs 45b; 25a vs 25b; 34a vs 34b; 35a vs 35b).Introduction of the trifluoromethoxy group in the No substantial differences were found between racemic and chiral (R)-bicalutamide derivatives (22d vs 42d; 22c vs 42c), confirming that the observed superior efficacy of the R-bicalutamide is mainly due to metabolic properties [13].
Interestingly, the best-performing new compound was 52,i n which the central methyl group is replaced with a trifluoromethyl function.The IC 50 (geometric mean of the overall antiproliferative activity in the four cell lines) is 3.5-fold better than its methyl analogue (23d), and the improvement is evident if compared with bicalutamide (50-fold).
Considering antiproliferative activity for each individual cell line, most of our compounds showed interesting results in all four cell types, with the best results found for those cells expressing the androgen receptor, such as LNCaP (the most sensitive cell line to the new derivatives).This evidence suggests that the newly synthesised structures retain an antagonistic effect on the androgen receptor.Significant antiproliferative activity was also found in the DU-145 cell line (the least sensitive to our new molecules), which does not express the androgen receptor and is insensitive to androgen activity, suggesting that also a different antiproliferative mechanism could be involved, along with the canonical antiandrogen receptor action.This potential off-target effect seems to be present also in the parent bicalutamide (3, 45a), which shows similar IC 50 values across the four cell lines, and might have been enhanced by the newly introduced modifications.
Newly synthesized enzalutamide analogues showed reduced activity in the antiproliferative assay in comparison with standard enzalutamide, with the only exception being 64c.However, the new modifications clearly influenced the activity on LNCaP cells, improving enzalutamide IC 50 up to five-fold (62b).The total absence of activity in the DU-145 cell line could be an indication of a pure anti-androgenic effect for these new compounds.Replacement of the original enzalutamide substituents in aromatic ring A with a para trifluoromethyl group is the most effective modification, while modifications on the B ring do not appear to significantly affect activity.
As a general consideration on the antiproliferative results, the introduction/change of position of trifluoromethyl and trifluoromethoxy groups emerges as an interesting strategy to improve the biological effects of the parent molecules.

Androgen receptor (AR) agonist/antagonist assay
Twenty-two derivatives were selected for the evaluation of their AR antagonist/agonist effect using the GeneBLAzer ® Betalactamase reporter technology for Nuclear receptors (NRs), to assess whether the antiproliferative activity found is related to any interference with the AR function [23].
All tested compounds were found to be antagonist of the androgen receptor in a single concentration antagonism experiment (10 mM concentration, antagonistic effect > 80%), which measures their ability to reduce the receptor activation induced by the known androgen receptor agonist R1881 [24,25].A1 0concentration antagonism assay (Supplementary data) showed that the new molecules possess an antagonistic IC 50 in the same range of reference (R)-bicalutamide and enzalutamide (Table 3).The (R)-bicalutamide IC 50 is comparable with previously reported values in similar assays [26,27].
In the bicalutamide-derived series of analogues, the antagonistic effect is not influenced by the substituents in aromatic ring A (28d vs 23d vs 25d vs 22d), with a small beneficial effect found for both the O and S linkers (28d vs 23d vs 33d) in the racemic derivatives, whereas the SO 2 linker is the most effective found for the (R) structures (42b vs 45b and 45h).Structural modifications on aromatic ring B have a major influence for antagonistic activity; substitutions in the meta position are associated with improved IC 50 values (i.e.28d, 28l, 23d), substitutions in the ortho position are well tolerated (i.e.23c), while substitutions in the para position result in higher IC 50 values (i.e.42b) if compared with (R)-bicalutamide.Overall, the results obtained fall in a similar range, with no significant changes in the anti-androgenic activity in comparison with control bicalutamide.The presence of a cyano group in the para position of ring B (28m, 28n), as in control (S)-enobosarm, is associated with the best antagonistic activity, while the introduction of a second substituent is tolerated.Compound 52, the best performing in the antiproliferative assay, showed an antagonistic IC 50 comparable to the one of (R)-bicalutamide.Considering the enzalutamide series, no significant change in the antagonistic effect was observed for the new derivatives tested.For 14 compounds, a further analysis was performed in order to verify any potential agonistic effect on the androgen receptor.The compounds were tested at a single concentration (10 mM) under the same experimental conditions used for the antagonistic assay, in the absence of R1881.All new derivatives showed either total absence or an agonistic effect <10% (not significant), confirming their full antagonistic behaviour.The partial agonist nature of (S)enobosarm (44e) was also confirmed in our assay: after an initial antagonistic activity at low concentration (best antagonistic IC 50 ), it acts as an agonist, activating the androgen receptor (agonistic effect >20%).Interestingly, the new ether analogues, which could be considered as enobosarm derivatives, were found to be full AR antagonists (i.e.28m, 28n, 25d, 28l).
The AR antagonist/agonist assay confirmed a pure AR antagonistic nature for the newly prepared compounds, which showed a concentration-dependent antagonistic activity.Moreover, an activity switch from partial agonist to full antagonist can be observed for our enobosarm analogues.
No correlation between AR assay and antiproliferative data can be identified, implying that other modes of action or other parameters could play an important role in the antiproliferative activity found for our new compounds.

Molecular modelling studies
Due to the retained activity of the newly synthesised compounds on the AR, and in an attempt to investigate the switch from partial agonist to full antagonist found for the newly prepared enobosarm derivatives, a series of molecular modelling analyses was designed and carried out.

AR antagonist conformation homology model
The functional switch from the open (antagonist) to the closed (agonist) conformation of the AR is caused by the movement of helix 12 (AR-H12), one of the 12 helices that define the ligandbinding domain (LBD) [28,29].The crystal structure of the complex bicalutamide-AR is only available in the closed conformation due to the W741L mutation (PDB ID: 1Z95), which causes bicalutamide to occupy the AR binding site in an agonist fashion, allowing the closure movement of helix 12 and therefore the activation of the AR [28,29].For the purpose of this part of the study, a homology model of the human WT-AR open conformation was built, using a single template approach in the MOE2014 homology modelling tool following a previously reported methodology [30].The crystal structure of the progesterone receptor in its open antagonistic form (PDB ID: 2OVM) was used as template, considering the good sequence identity (54%) with the human AR [31].The final 3D model was obtained as the Cartesian average of 10 generated intermediate models [32].Validation of the new model was performed using the Rampage Server, the Errat plot and calculating the RMSD with the 2AM9 crystal structure [33e35].Fig. 3 shows the shift of helix 12 between the new model and the mutated ARbicalutamide crystal (PDB ID: 1Z95) [28].

Docking studies
Docking of (R)-bicalutamide was performed in the LBD of the new model to evaluate its predicted antagonistic binding mode.As shown in Fig. 4, docked bicalutamide occupies the binding site in proximity of Trp741, forming a H-bond with Asn705 through its hydroxyl group, and most importantly interacting with Arg752 through a second H-bond with its cyanide nitrogen [36].A direct comparison with the bicalutamide-resistant AR crystal (Fig. 5) suggests that the sulfone ring orientation of the docking pose is the most plausible considering the presence of Trp741 and the position of helix 12.In fact, in the crystal, due to a W741L mutation, the sulfone aromatic ring is pointing down towards the centre of the protein, lying in the position that would be occupied by the Trp741 indole ring, allowing helix 12 closure [36].The presence of Trp741 in the model does not allow this disposition (steric clash), orienting aromatic ring B toward helix 12, thus blocking its closure.Docking of the more rigid enzalutamide (Fig. 6) confirms the proposed antagonistic binding mode (comparable to the crystal structure of AR-LDB in complex with an enzalutamide-like compound, PDB ID: 3V49) [37].Fig. 7 shows the docking of two newly designed compounds (28m, 33d, 52) in the AR homology model.The three compounds show a consistent binding mode, occupying the LBD in the same manner as bicalutamide, maintaining the same key interactions, consistently with the AR assay results.
Replacement of the fluorine substituent in the B ring with a bigger group such as trifluoromethyl or trifluoromethoxy, increasing the steric hindrance of this part of the molecule, could allow the overcoming of bicalutamide resistance mechanism.This speculation is supported by the docking of 33d in the mutant AR crystal structure 1Z95: as shown in Fig. 8, the bulkier substituent in the B ring does not fit in the AR closed conformation, even in the presence of the adaptive mutation W741L, as indicated by the close proximity of ring B to the protein surface.Biological confirmation of this effect will be the focus of future studies.
Molecular docking of the new enobosarm derivatives (i.e.28d, 28m, 28n, 28l, 25d) was performed on the crystal structure of closed-conformation WT-AR in complex with (S)-enobosarm (PDB ID: 3RLJ) [38], in order to explain the observed activity switch from partial AR agonist to full antagonist.(S)-Enobosarm, after binding the AR and acting as antagonist, activates the receptor by favouring the closure movement of helix 12 even in the absence of the W741L mutation.This movement, as shown in the crystal structure, is made possible by the more flexible and less bulky oxygen linker, which allows a binding conformation similar to the one found for bicalutamide in the W741L AR mutant crystal structure (Fig. 9).Docking results show that the replacement of the cyano substituent in aromatic ring B with a larger group (i.e.2-OCF 3 in 28l or 2-CF 3 in 28d), or the introduction of a second aromatic substituent (i.e.2-CF 3 in 28m), might lead to structural clashes with helix-12, confirming that these new enobosarm derivatives could impede the movement of helix 12 to the AR-closed conformation, hence the absence of any residual AR agonistic activity (Fig. 10).
The results obtained for our new enobosarm analogues might represent a further indication that the newly inserted structural modifications could play a role in overcoming the problem of bicalutamide-induced AR mutation W741L.

Molecular dynamics simulations
To further confirm the reliability of the new homology model, and consequently support the speculations on the binding mode proposed for the newly synthesized compounds, 5 ns molecular dynamics (MD) simulations for selected derivatives (45a, 23d, 25d, 28d, 33d, 62b), either free in solution or in complex with the AR, were performed.These structures were selected in order to cover the range of activities found in the AR antagonist assay.The MD output results were used to compute DG binding energy calculations of the ligand-protein complexes (Table 4).
The stability showed by most of the ligand-protein systems during the simulations could be a consequence of the active site composition, mainly formed by hydrophobic amino acids (Met745, Met787, Leu707, Leu704, Leu712, Val746, Trp741, Phe764, etc.), which provides an ideal environment for the lipophilic nature of these fluorinated molecules.Moreover, the formation of two hydrogen bonds in the bicalutamide/enobosarm derivatives, both present during the entire MD duration, with Arg752 and Asn705, further contribute to the system stability, allowing the molecules to maintain a stable conformation in the active site.The lowest calculated DG binding was obtained for 62b, whereas 33d, a derivative with one of the highest antagonistic IC 50 values, showed the highest energy result.Interestingly, a similar correlation between the calculated DG binding energy and the AR antagonistic IC 50 was found for all the compounds studied.Enzalutamide derivative 62b showed the best result in terms of DG binding , confirming the best antagonistic IC 50 value obtained from the assay and highlighting the high affinity of the rigid enzalutamide-like structure for the androgen receptor.The DG binding /AR antagonistic IC 50 correlation might represent a further confirmation of the reliability of the new AR homology model and of the proposed binding mode for the new fluorinated compounds.

In vitro metabolic stability, cell permeability and cytotoxicity studies
The significant improvement obtained with the newly synthesized analogues in the antiproliferative assay, which is associated with retained AR antagonistic activity in comparison with the reference compounds, could be a consequence of an off-target effect, already present in the parent structures and enhanced by the new modifications, or a consequence of altered cellular metabolism, stability and/or cell permeability of the new molecules, due to their high fluorine content.

Metabolic stability in liver microsomes
The selected compounds together with the controls were tested for their stability (oxidative metabolism being the predominant type of biotransformation) in human liver microsomes (phase I drug metabolism).The tested compounds were incubated for 45 min with pooled liver microsomes (Supplementary data), and the intrinsic clearance (CL int ) and half-life (t 1/2 ) values were calculated based on 5 time points (Table 5).Bicalutamide, as reported in the literature, mainly undergoes oxidation in humans by   cytochrome P450 enzymes [26].
As expected, intrinsic clearance (Cl int ) is high and, as consequence, t 1/2 is very low for 22c, 22d, 23c, 23d, 42b, 42c.These compounds are thioethers and therefore they might undergo rapid oxidation of their sulfide linker to the corresponding sulfoxide and eventually to the sulfone.Ether derivatives 28m, 27b and (S)enobosarm and sulfones 33d and (R)-bicalutamide show lower Cl int and higher t 1/2 , with some of them showing complete metabolic stability (no loss of the parent detected during the assay).Oxidative metabolism seems to affect mainly the thioether linker with a small effect on the aromatic systems.
Interestingly, introduction of a trifluoromethyl group in the aromatic ring (33d) completely abolishes any oxidative metabolism on the ring itself, doubling the t 1/2 in comparison with (R)-bicalutamide.The significant antiproliferative activity found for thioether and ether derivatives does not seem to be caused by any oxidative metabolite, as indicated by the high stability of ether analogues and by the decrease in antiproliferative activity found for the respective sulfone derivatives.Moreover, these results suggest that ether derivatives have the best antiproliferative and stability (low intrinsic clearance) profile, making them good potential drug candidates.

Caco-2 cell permeability
An in vitro Caco-2 bi-directional permeability assay was performed to gain information on membrane penetration, efflux from cells and possible interaction with the P-glycoprotein (P-gp) for the new compounds.The data obtained are reported in Table 6 as apparent permeability (P app ) and efflux ratio.Generally, compounds with a P app A / B between 2 and 20 Â 10 À6 cm s À1 are considered as medium permeability drugs, whereas an efflux ratio (BA/AB) higher than 2 is an indication that the compound could be a substrate for a drug transporter (i.e.P-gp, BCRP) and undergo active efflux to the extracellular compartment.These transporters are all over-expressed in cancer cells and are commonly responsible for drug resistance [21,39].
(R)-Bicalutamide (45a) shows a P app A / B value of 32.0 Â 10 À6 cm s À1 , which is comparable to the one reported in literature (22.1 Â 10 À6 cm s À1 ), as both of these values would fall into the rank order category of good permeability [21].Under the test conditions the P app B / A value was found to be 38.1 Â 10 À6 cm s À1 , indicating membrane permeability in both directions for (R)-bicalutamide.The calculated efflux ratio (BA/AB) of 1.19 indicates that (R)-bicalutamide is not an efficient substrate for transporters, in discordance with published data, in which bicalutamide shows an efflux ratio (BA/AB) of 5.4 [21].This discrepancy could be a consequence of using (R)-bicalutamide (45a), since bicalutamide enantiomers have different metabolism, permeability and absorption properties in vivo, which influence their biological activity [13] and possibly the interaction with the Pglycoprotein.Moreover, several experimental variables can influence the transport of compounds in both membrane directions.Due to this variability, the Caco-2 cell data have been used only to verify if any relevant difference in cell permeability is present among the tested compounds.The reliability of our assay and the functional presence of the protein transporters were evaluated using two standard reference compounds (atenolol for low permeability and propranolol for high permeability) and one known P-glycoprotein substrate (talinolol) (Supplementary data).The new molecules appear to be moderately permeable (2<P app A / B>20 Â 10 À6 cm s À1 ), with a low tendency to efflux from the cell.Differences between racemic and (R) derivatives are evident comparing compounds 22c and 42c, in which the pure (R) enantiomer has an increased efflux ratio, still lower than 2. Overall, the  results obtained suggest that these compounds do not show any significant difference in cell permeability in comparison with the standards, therefore no correlation can be identified between antiproliferative data and cell permeability properties.

Cell viability assay (MTT)
A MTT cell viability assay was performed in a human hepatocarcinoma (HepG2) cell line, in order to evaluate the cytotoxicity profile and to further explore the possible presence of an off-target effect for the new compounds.The MTT assay measures the effect of molecules on cell proliferation and other events that eventually lead to cell death [40].
Data are reported in Table 7 as minimum effective concentration (MEC) related to a vehicle control (DMSO) and AC 50 .All tested compounds showed a concentration-dependent decrease in formazan production across the range studied (example plots in the Supplementary data), indicating a decrease in cell viability associated with some growth inhibitory and toxic effects.Compared with control (R)-bicalutamide (45a), the new compounds are in general characterized by a higher cytotoxicity, with the lowest MEC and AC 50 values obtained for 23d, which is also one of the most active compounds in the antiproliferative assay (antiproliferative data Table 7).
These results might represent an important indication that the newly prepared analogues, besides the canonical AR antagonist activity, possess an extra effect, which positively contributes to their antiproliferative action.Such effect seems to be particularly evident for 23d, while other derivatives (e.g.28m, 42b) do not show a significant increase in cytotoxicity, maintaining instead a remarkable improvement in the antiproliferative assay in comparison with bicalutamide.The mechanism and effect (necrosis, apoptosis, reduction of cellular proliferation or effect on mitochondria) causing this cytotoxicity, which seems to be a consequence of the new structural modifications, will be the main focus of future studies.

Conclusions
The structures of bicalutamide and enzalutamide, two nonsteroidal androgen receptor antagonists in use to treat prostate cancer, have been chemically modified leading to the synthesis of new and more potent antiproliferative compounds.Different chemical modifications have been planned in order to benefit from the properties of fluorine in improving the activity and pharmacological profile of drugs.The novel chemical entities were found to be full androgen receptor (AR) antagonists, with no residual agonist effect found also for structural analogues of enobosarm (ether derivatives), a well-known AR partial agonist.Introduction of extra perfluoro groups (CF 3 and OCF 3 ) confers metabolic stability to phase I drug metabolizing enzymes, maintaining the same good membrane permeability of the parent compounds.A very notable improvement in the in vitro anti-proliferative effect on four different PC cell lines (VCap, LNCaP, DU-145 and 22RV1) has been achieved, improving the activity of the parent structures up to 50 folds, thus making some of the new molecules potential drug candidates for the treatment of prostate cancer.This marked activity improvement did not parallel an enhanced AR antagonistic effect.Moreover, increased cytotoxicity found for some of the new analogues in the MTT assay could be an indication that the compounds developed in this work, while certainly binding to AR LBD and exhibiting full antagonist behaviour, might possess another biological effect, which positively contributes to their antiproliferative action in the prostatic cancer cell lines.The reliable AR homology model prepared and all the docking information provided can offer useful insights for the development of pure AR antagonists.The new model appears well validated by all observed data.
The results achieved in this work will be the starting point for further modification of the parent compounds, in order to improve both antiproliferative and anti-androgen properties.All successful structural modifications will be further investigated, giving priority to the replacement of the central methyl group of bicalutamide with a CF 3 (52).
Future biological studies will aim to understand the mechanism of action and the biological consequences of potential off target effects, along with the evaluation of our novel AR antagonists in a W741L mutant AR.Finally, the most promising compound for each structural family (23d, 27b, 33d) has been selected for in vivo preclinical studies in mouse models.

Experimental
All chemistry, biology and molecular modelling experimental procedures, along with compound characterisation, are fully described in the Supplementary Data.All final compounds were purified by column chromatography or recrystallization, fully characterised by NMR ( 1 H, 13 C, 19 F) and LRMS.HRMS is reported for representative final compounds (one example for each synthetic method).All final compounds were found to be >95% pure by HPLC.

Fig. 1 .Fig. 2 .
Fig. 1.Structure of anti-androgen small molecules approved by FDA or in clinical development for the treatment of PC.

Fig. 3 .Fig. 4 .
Fig. 3. Superposition between the crystal structure of the AR in the closed form (PDB ID: 1Z95) (pink) and the new open AR model (light blue).Mutated Leu741 (pink) reduces the steric hindrance in comparison with the non-mutated Trp741 (light blue).Co-crystallised bicalutamide (carbon atoms in salmon) is shown as stick model.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5 .Fig. 6 .
Fig. 5. Comparison between our AR model (light blue) and bicalutamide-resistant AR crystal 1Z95 (pink).Crystallised (carbon atoms in salmon) and docked bicalutamide (carbon atoms in pink) highlighted.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 7 .
Fig. 7. Predicted binding mode of new derivatives 28m, 33d and 52 in the AR homology model.

Fig. 8 .
Fig. 8. Docking of 33d (carbon atoms in light pink) in the 1Z95 crystal structure.The red circle highlights how the ring B of the new compound is in close proximity to the protein surface, potentially causing steric clashes that might impede the helix 12 movement to the closed conformation of the W741L AR mutant.The occupational molecular surface of the binding site is shown in grey.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 9 .
Fig. 9. Superimposition between the crystal structure of WT-AR co-crystallised with (S)-enobosarm (green) and the W741L-AR co-crystallized with bicalutamide (salmon).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 10 .
Fig. 10.A)(S)-Enobosarm co-crystallized in the WT-AR closed conformation (PDB ID: 3RLJ).B) Docking of 28d (carbon atoms in turquoise) and 28l (carbon atoms in white) in the 3RLJ crystal structure.Replacement of the cyano substituent with a bulkier group might cause structural clashes (red circles) with helix-12.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1
Antiproliferative activity for bicalutamide/enobosarm derivatives.All data are mean values from triplicate experiments.

Table 2
Antiproliferative activity for enzalutamide derivatives.All data are mean values from triplicate experiments.paraposition does not have a significant effect or slightly decreases the activity if compared with eCF 3 analogues.The 4-CN group present in (S)-enobosarm (44e) slightly improves or leaves unaltered the antiproliferative activity, depending on the X linker and the substituent in the A ring (44e, 28e, 29e).Repositioning of the trifluoromethyl group from the para to the meta and the ortho position greatly improves antiproliferative activity up to ten-fold.The best results were obtained with a substituent in position ortho, either a CF 3 (23d) or an OCF 3 (25l) group.The IC 50 values for each new derivative appear to be a combination of the substituent in the meta or ortho position of aromatic ring B, the linker (thioether analogues are the most active) and the substituent in the A ring (nitro derivatives perform slightly better than cyano analogues).Generally, all meta and ortho substitutions on the B ring performed better than standard bicalutamide, leading to a significant activity improvement.Replacement of the phenyl ring B with either a 4-trifluoromethyl pyridine or a 5-trifluoromethyl pyridine abolishes activity with only few exceptions (25o, 25p), which are probably a consequence of the presence of a 4-NO 2 -2-CF 3 aromatic substituent in the A ring.Introduction of a second fluoro substituent in the B ring, such as a 2,4-fluoro, 2,3-fluoro, 4-CN,3-F or 4-CN,2-CF 3 , is in general associated with a better activity profile if compared with standard bicalutamide and enobosarm.In particular, the introduction of an extra trifluoromethyl group in the ortho position of enobosarm aromatic ring B improved its antiproliferative activity up to 4-fold (28m and 29m vs 44e).

Table 3
AR antagonist/agonist assay results.* Compounds were considered full antagonist if at a single concentration of 10 mM the reduction of receptor activation by R1881 was greater than 80%; ** Compound at 10 mM concentration in absence of R1881 was used; *** Compounds tested at 10 different concentrations; **** Due to poor solubility, the highest concentration used was 1 mM; N.E, no antagonistic effect.

Table 4
Calculated ligand-interaction energies for the compounds analysed by Molecular Dynamics and antagonistic IC 50 values.

Table 5
Metabolic stability in human liver microsomes; * CL int : theoretical unrestricted maximum clearance of unbound drug by an eliminating organ, in absence of blood or plasma protein binding limitations; ** SE: standard error; ***Compounds metabolically stable with no loss of parent detected for the duration of the assay.

Table 6 Caco
-2 cell permeability test; * (A / B): apical-basolateral flux; ** (B / A): basolateral-apical flux; this information was used for the apparent permeability (Papp) evaluation; ***Efflux Ratio: Mean Papp B / A/Mean Papp A / B; y Compounds tested under standard conditions using HBSS buffer at pH ¼ 7.4.All other compounds were tested under the same conditions but the cell media contained BSA (1% w/v).All compounds were tested an initial concentration of 10 mM.

Table 7
MTT assay; all the compounds have been tested at 8 different concentrations.* MEC: minimum effective concentration that significantly crosses vehicle control threshold; ** AC50: concentration at which 50% of maximum effect is observed; ***Geometric mean.Bassetto et al. / European Journal of Medicinal Chemistry 118 (2016) 230e243