Synthesis of Enantiospecific Androgen Receptor Antagonists of Enzalutamide-Resistant Human Prostate Cancer Cells

Abstract

Drugs targeting the androgen receptor (AR) are important therapeutics for castration-resistant prostate cancer (CRPC) but are vulnerable to resistance due to AR mutations. Previously, we reported that the (5R/5S) enantiomer pairs of four novel 3-pyrrolazole carboxamide derivatives 9ab–12ab are antagonists and agonists, respectively, of AR, as demonstrated by transcriptional assays supported by fluorescence microscopy and proliferation studies. Herein, we describe the synthesis and stereochemical characterization of these compounds and document their enantiospecific effect on enzalutamide-resistant AR.

The typically discouraging prognosis for advanced prostate cancer stems from the onset of resistance to the limited treatments available to retard disease progression. , Patients with lethal metastatic castration-resistant prostate cancer (mCRPC) have multiple U.S. Food and Drug Administration (FDA)-approved treatment options, including androgen receptor (AR) pathway inhibitors, taxane chemotherapy, and radiopharmaceuticals. However, durable responses remain limited and the disease remains largely AR-driven, highlighting the urgent need for next-generation inhibitors that specifically target dysregulated mechanisms in mCRPC. Enzalutamide functions as an antagonist to the AR. The AR is predominantly located in the cytoplasm of prostate cells, where agonist binding at the AR ligand-binding domain (AR-LBD) promotes conformational closure of helix-12. This activates a signaling cascade that induces translocation of AR to the nucleus, where interaction with androgen response elements (AREs) regulates gene expression essential for male sexual development. −

Some years ago, Salvati et al. described a series of compounds targeting the AR-LBD, among which the most promising compound was BMS-641988, an ethylsulfonamide derivative of an oxabicyclic scaffold (13a; Figure ). 13a advanced to clinical trials but unfortunately exhibited neurotoxicity and inadequate efficacy, precluding further development.

1.

Structures of 9a–13a (3aR,4R,5R,7R,7aS) and 9b–13b (3aS,4S,5S,7S,7aR). The (R) and (S) labels refer to assignment at C-5.

With the goal of achieving an AR antagonist with an improved safety profile and better efficacy, we explored replacement of the C-5 sulfonamide in 13a with a pyrazole 3-carboxamide moiety, illustrated by 9a–12b (Figure ), taking into account the potential importance of the stereocenter at position C-5 of the oxabicyclic scaffold. Intriguingly, all of the (5R) stereoisomers (9a–12a) were AR antagonists, whereas the (5S) counterparts were potent AR agonists, as measured by AR transactivation assays, including nuclear hyperspeckling, ARE luciferase reporter activity, and AR-dependent gene transcription, as reported previously.

In this paper, we present the synthesis of the 9a/9b, 10a/10b, 11a/11b, and 12a/12b stereoisomers. We also report the preparation and crystal structure of 13b. Using site of metabolism (SOM) computer model predictions (9a−12a) confirmed by a liver microsome assay (9a, 12a), we demonstrate that compounds 9a–12a are not metabolized to the C-5 oxidized metabolite to which the observed toxicity of 13a was attributed. The stereoisomer pairs 9a/9b–10a/10b and 11a/11b–12a/12b consistently exhibit antagonism/agonism duality in ARE luciferase assays. The C-5 (R) stereoisomers 9a and 12a showed robust antagonism of mutation-induced enzalutamide-resistant AR, suggesting further development of this platform.

Compounds were synthesized using the general route shown in Scheme , with the early steps being adapted from the synthesis of 13a. , Thus, Diels–Alder cycloaddition of maleimide 2.1 or 2.2 to the MEM ester of 2,5-dimethyl-3-furoic acid selectively yields the exoisomers (3.1 and 3.2), which were used in the following steps without separation. It is to be noted that the yield for Diels–Alder cycloaddition increased from 25% to 90% for the 2′-fluoro analogue. Subsequent catalytic hydrogenation led to the formation of esters 4.1/4.2. Treatment of 4 with 3 N hydrochloric acid (HCl) resulted in a racemic mixture of acids 5.1/5.2. Sequential Curtius rearrangement followed by trifluoroacetic acid (TFA)-promoted cleavage of resulting trimethylsilylethoxycarbamate (Teoc) gave racemic amines 7.1/7.2. It was reported that Curtius rearrangement of the separated 6 enantiomers retains stereochemistry at C-5. Coupling of amines 7.1/7.2 with carboxylic acids 8 afforded the respective target amide-containing oxabicyclic pharmacophores. In the analogue containing a 2′-fluoro group, exposure to methanol should be avoided during purification to prevent ring opening of pyrrolidine dione.

1. Synthesis of EMI-17##, 9a/9b–12a/12b .

^a Reagents and conditions: (a) X = H: maleic anhydride, glacial acetic acid, reflux overnight, 59%; X = F: (1) maleic acid monomethyl ester, EDC–HCl, HOBt, TEA, DMF, 0–5 °C and (2) SOCl₂, DCM/DMF, 58% over two steps; (b) 2-MEM 2,5-dimethylfuran-3-carboxylate, 125 °C, 2 h, then rt overnight, 25–90%; (c) H₂, Pd/C, EtOAc, rt overnight, 75–90%; (d) 3 N HCl, THF, rt for 16 h, quantitative yield; (e) 2-trimethylsilylethanol, DPPA, Et₃N, 4 Å MS, 1,4-dioxane, 75 °C, 53–79%; (f) TFA, CH₂Cl₂, rt for 2 h, quantitative yield; (g) carboxylic acid 8, DIPEA, HATU, DMF, rt overnight, 26–66%; and (h) chiral HPLC separation.

In the original synthesis of 13a, the 4.1 enantiomers (4.1a and 4.1b) were separated using semi-preparative chiral HPLC and only the C-5 (R) stereoisomer (4.1a) was carried forward to produce the target compound [thus, corresponding C-5 (S) stereoisomer 13b was not described]. In our syntheses, 4.1 and 4.2 were not resolved, but instead, the racemate was carried forward en route to the final target C-5 (R/S) mixtures; at that point, the racemic final product was chirally resolved to afford both individual C-5 stereoisomers.

Effective resolution is critical, as contamination of one stereoisomer by its counterpart could lead to significant changes in their observed biological activities. For example, 2.5% contamination by 9b (95% CI = [2.1%–2.9%]) halved (R) counterpart 9a antagonistic activity, and 9% contamination (95% CI = [6%–12%]) negates the antagonist effect entirely. A high stereoisomeric purity (>99.8%) is thus required for accurate biological study of this family of compounds, motivating additional rounds of chiral HPLC purification.

The specific rotation of 13a had not been reported by Attar et al. Balog et al. determined its absolute configuration via X-ray crystallography. After confirming the absolute stereochemistry of 13a by X-ray crystallography (CCDC 2452778; Table S1), we determined the specific rotation of 13a to be [α]_D = −28.1 in methanol. The specific rotation of its (S) stereoisomer was measured to be [α]_D = +26.1 in methanol, with its absolute stereochemistry also verified by X-ray crystallography (CCDC 2452779; Figure and Table S2).

2.

Crystal structures of (A) 13a and (B) 13b.

It is seen (Table ) that the final C-5 (R) isomers that exhibit negative specific rotations are AR antagonists. Conversely, the (S) isomers that have a positive specific rotation are AR agonists, consistent with the activity reported previously.

1. Specific Optical Rotations of 9a–13b .

Compound	[α]λ T	Compound	[α]λ T
9a	[α]D  = −28.7	9b	[α]D  = +31.6
10a	[α]D  = −48.7	10b	[α]D  = +49.0
11a	[α]D  = −31.0	11b	[α]D  = +34.0
12a	[α]D  = −46.9	12b	[α]D  = +48.2
13a	[α]D  = −28.1	13b	[α]D  = +26.1

^a c = 0.3 in methanol.

^b c = 0.127 in acetonitrile.

^c c = 0.2 in acetonitrile.

^d c = 0.5 in methanol.

In preclinical studies, 13a was metabolized to BMS-570511, which was further reduced to BMS-501949 (Figure ). BMS-501949 was associated with neurotoxicity, specifically seizure activity in dogs at doses of 25 mg/kg that corresponded to high plasma concentrations. Our compounds 9a and 12a, on the other hand, exhibit a different metabolic profile. First, we predicted their blood–brain barrier (BBB) penetration relative to known androgen antagonists using the Schrödinger QikProp module.

3.

Drug metabolism of 13a.

Setting a logBB of < –1 as a threshold we obtained computational predictions with good correlations with available in vivo data (Figure A). For example, flutamide, enzalutamide, and BMS-501949 were predicted to readily penetrate the BBB, whereas bicalutamide, darolutamide, 13a, as well as our preclinical candidates all had logBB values of < –1, suggestive of low BBB permeability. Additionally, toxic metabolites (BMS-501949 for 9a and plausibly o-fluoro-BMS-501949 for 12a) that accumulated in the 13a sample were not predicted or detected for 9a or 12a even after 8 h of incubation with liver microsomes (Figures and B and Figure S1).

5.

Neurotoxicity potential of 9a. (A) Calculated logBB of 9a vs. reference compounds. (B) Liver microsome incubation data. Extracted ion chromatograms for m/z 379.0911 are shown, corresponding to the calculated [M – H]^– mass of BMS-501949, the known toxic metabolite of 13a. The bottom panel shows the BMS-501949 reference, while the third panel shows a signal corresponding to BMS-501949 produced from 13a degradation. We note that the ion intensity in the top panel is for a different species, potentially unique to degradation of 9a (Table ); it has a later retention time and a measured m/z of 377.0737 (the third isotope of which was within our selected ion extraction window).

4.

Schrödinger P450 site of metabolism (SOM) predictions in (A) 13a and (B) 12a. Size of the green circle in overall SOM is positively correlated with the probability of metabolism, and the value in intrinsic reactivity is positively correlated with the probability of metabolism. Results for 9a, 10a and 11a were similar.

Finally, the rates of intrinsic clearance in liver microsomes were comparable across 9a (15.3 μL min^–1 mg^–1), 12a (10.3 μL min^–1 mg^–1), and 13a (10.1 μL min^–1 mg^–1) (Table ). In this experiment, 9a and 12a were converted to metabolites with m/z = Taken together, these results suggest improved safety profiles of 9a and 12a compared to 13a and enzalutamide.

2. Metabolism of Representative Compounds in Liver Microsomes .

Compound	Formula, Exact Mass	T 1/2 (min)	Clint (μL min–1 mg–1)	Metabolite	Detected in LCMS
9a	C22H18F3N5O4, 473.1311	56.6	15.3	unnamed C22H18F3N5O5	yes
				BMS-501949 C18H15F3N2O4	no
12a	C24H21F4N5O5, 535.1479	47.3	10.3	unnamed C24H19F4N5O5	yes
				o-fluoro-BMS-501949 C18H14F4N2O4	no
13a	C20H20F3N3O5S, 471.1076	86.4	10.1	BMS-570511 C18H13F3N2O4	yes
				BMS-501949 C18H15F3N2O4	yes

^aRates of intrinsic clearance Cl_int were established by measuring levels of intact compound after increasing incubation times using mass spectrometry. Metabolite lists the molecular formula corresponding to the observed LCMS peak.

The clinical AR double-mutation F877L/T878A has shown to elicit resistance to enzalutamide by converting the antagonist to an agonist. , To evaluate the effect of 9a and 12a, we performed ARE luciferase assays using PC3 cell lines stably expressing the AR wild type and double-mutation F877L/T878A. As expected in wild type AR, 9a and 12a effectively inhibited R1881-induced AR activation and did not display notable agonistic activity in androgen-deprived conditions (−R1881), similar to the clinical drug enzalutamide (Figure , top row). However, in the presence of AR double-mutation F877L/T878A, enzalutamide switches from antagonist to agonist when activities were compared to untreated control conditions (Figure , dark gray bars). In contrast, 9a and 12a retained their antagonistic activity and displayed no significant agonism in castrate conditions (−R1881) (Figure , yellow bars, bottom row). In summary, these results demonstrate that our novel compounds remain robust antagonists in the presence of mutation-induced enzalutamide resistance AR, warranting continued laboratory development.

6.

9a [(R)-EMI-1702] and 12a [(R)-EMI-1707] target enzalutamide-resistant AR. ARE luciferase assays in cells treated with 10 μM drug + 1 nM R1881. Data (n = 2) are the mean ± SD. Linear model (one-sided), drugs of <NTC in the +R1881 test and >NTC in the −R1881 test. NTC, no treatment control; ENZ, enzalutamide. (∗) p < 0.05, (∗∗) p < 0.01, (∗∗∗) p < 0.001, and (n.s.) not significant. p values corrected for multiple comparisons at a family-wise error rate of 0.05.

Glossary

Abbreviations Used

CRPC

castration-resistant prostate cancer

mCRPC

metastatic castration-resistant prostate cancer

FDA

Food and Drug Administration

AR

androgen receptor

AR-LBD

AR ligand-binding domain

ARE

androgen response element

SOM

site of metabolism

TFA

trifluoroacetic acid

Teoc

trimethylsilylethoxycarbamate

HPLC

high-performance liquid chromatography

BBB

blood–brain barrier

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmedchemlett.6c00085.

• Synthetic details and characterization data (NMR spectra and MS) for 3.2, 4.2, 5.2, 6.2, 7.2 and 9a/9b–12a/12b, X-ray crystallographic data for 13a/13b and experimental details of biological assays and computed metabolism modeling (PDF).

This research was funded in part by a collaboration agreement between the University of Southern California (USC) and the Ellison Medical Institute (DBA-0031) and by a USC Dornsife Graduate Fellowship to Pyone Myat Thwe.

The authors declare the following competing financial interest(s): Dr. McKenna is a paid consultant for the Ellison Institute. USC has also licensed intellectual property under study in the research project to the Ellison Institute and may receive royalty income.

No unexpected or unusually high safety hazards were encountered.