Discovery of BMS-641988, a Novel Androgen Receptor Antagonist for the Treatment of Prostate Cancer

Abstract

BMS-641988 (23) is a novel, nonsteroidal androgen receptor antagonist designed for the treatment of prostate cancer. The compound has high binding affinity for the AR and acts as a functional antagonist in vitro. BMS-641988 is efficacious in multiple human prostate cancer xenograft models, including CWR22-BMSLD1 where it displays superior efficacy relative to bicalutamide. Based on its promising preclinical profile, BMS-641988 was selected for clinical development.

Carcinoma of the prostate (CaP) is the most common malignancy among men in the US and the second most common cause of cancer-related death worldwide after lung cancer.¹ The androgen receptor (AR) is a member of the nuclear hormone superfamily of ligand-induced transcription factors and is a key signaling pathway leading to the emergence of CaP. Androgen ablation, by surgical or chemical castration in combination with an antiandrogen such as hydroxyflutamide (1) or bicalutamide (2), has been the standard of care for advanced CaP for many years.² This therapy is initially effective in 80–90% of patients; however, >50% of the patients will ultimately develop castration resistant prostate cancer (CRPC) after ∼18 months.³ The treatment of CRPC is challenging due to the sustained AR signaling, which is the result of AR overexpression/activation and the presence of activating AR mutations.⁴ Furthermore, it has been reported that CRPC tumors express the necessary cytochrome P450 enzymes for intratumoral androgen production, thus bypassing the effects of chemical castration, which targets only gonadal androgen production.⁵ These findings suggest that CRPC remains AR dependent, and effective therapies must target AR signaling directly with improved next-generation AR antagonists.

MDV3100 (enzalutamide, 3) is a potent AR antagonist that was recently approved by the US Food and Drug Administration for the treatment of metastatic CRPC patients that have progressed post-treatment with docetaxel.⁶ Although enzalutamide has shown promise in treating these patients, nearly all patients go on to develop resistance to enzalutamide via AR mutations.⁴ Thus, there is a need for novel antiandrogens with distinct interactions in the AR ligand binding domain that could be dosed together or sequentially in the clinic to combat the potential pathways leading to CRPC progression. Our laboratory has been focused on the rational structural-based design of structurally novel, nonsteroidal small molecule AR antagonists for the potential treatment of CRPC.

Previously, we have reported a series of [2.2.1] carbobicyclic⁷ and oxabicyclic⁸ succinimide based AR antagonists (4 and 5, Figure 1). These compounds demonstrated potent binding affinity (K_i) and functional antagonist activity (IC₅₀) against the wild-type AR as found in the MDA-MB-453 cell-line (Table 1). These compounds compared favorably in terms of potency to the clinically used antiandrogens hydroxyflutamide (1) and bicalutamide (2). We designed a series of oxabicyclic-based AR antagonists, such as 5, that demonstrated a superior pharmacokinetic (PK) profile compared to the carbocycles such as 4. Based on our understanding of the need for sustained AR suppression in an effective AR antagonist, we expected that robust PK would be essential for an efficacious AR antagonist.⁸ Accordingly, compound 5 was shown to be efficacious in the CWR22-BMSLD1 human prostate cancer xenograft model, where bicalutamide shows only limited efficacy.⁸ We looked to incorporate PK properties critical for an effective AR antagonist, such as log T_1/2, with a broader activity profile than is seen with first generation agents such as bicalutamide. Accordingly, we screened out agents against the human CaP model CWR22, which has been shown to be refractory to both bicalutamide and hydroxyflutamide. Based on this result, we wanted to expand the scope of this series by using a structure-based approach to identify even more potent AR antagonists.

Figure 1.

Known androgen receptor antagonists.

Table 1. In Vitro Biological Activity.

#	MDA-MB-453 Ki (nM)a	MDA-MB-453 IC50 (nM)b
2	64	173
5	8	10
11	1600	>5000
12	23	22
13	50	34
14	31	60
15	3.0	23
16	50	34
17	31	60
18	7.0	6
19	12	20
20	10	1.0
21	1.0	10
22	2.0	7.0
23	1.7	16

^aBinding (K_i) determined through direct displacement with [³H]-DHT in the MDA-MB-453 cell-line

^bFunctional antagonist activity (IC₅₀) in the MDA-MB-453 cell-line determined through a transiently transfected reporter system utilizing the secreted alkaline phosphatase reporter gene driven by the AR-dependent PSA promoter.

Utilizing available X-ray cocrystal structures of the AR generated at BMS, we developed a molecular model of the WT AR LBD (Figure 2) to aid in the identification of new AR antagonists.^9,10 In this model, interactions of N-705 with both the C-5 hydroxyl and the bridging oxygen of compound 5 are evident. Additional interactions between R752 and Q711 and the aryl nitrile functionality of compound 5 are also present. We postulated that endosubstitution at C-5 or C-6 of the bicycle would result in a direct interaction with helix-12 (H-12), possibly creating a more classical AR antagonist conformation, similar to that predicted for bicalutamide. The key hydrogen bonding interaction with N-705 would be maintained by the bridging oxygen as seen with bicalutamide (2), which is shown in orange. Thus, we set out to investigate the effect of various substitutions on the endoface of the [2.2.1]-oxabicyclic core of compound 5.

Figure 2.

Compound 5 docked into a model of the wild-type AR ligand binding domain.

Our efforts started with the synthesis of compounds 11–13 by the synthetic route shown in Scheme 1. Maleimide 8 was prepared from the aniline 6 and maleic anhydride (7) under standard conditions.¹¹ Diels–Alder cycloaddition between 8 and the MEM ester of 2,5-dimethyl-3-furoic acid occurred at 120 °C to give only the desired exoisomer 9 after precipitation. Catalytic hydrogenation led to formation of the endoester by selective reduction from the beta-face of the olefin. Normal-phase chiral HPLC separation of the racemic endoester gave the desired enantiomer 10 in 45% yield and >99% ee. The optical isomer depicted by compound 10 was determined to be optimal for potent AR antagonist activity, and the absolute and relative stereochemistry was eventually confirmed by X-ray crystallographic analysis of compound 23.¹² Treatment of 10 with 3 N HCl gave the key intermediate acid 11 in good yield. Compounds 12 and 13 were prepared by standard ester and amide formation conditions.

Scheme 1. Synthetic Route to Endo-Substituted Bicycles.

Reagents and conditions: (a)HOAc, 110 °C, 88%; (b) MEM 2,5-dimethyl-3-furoate, 120 °C, 33%, exo-isomer only; (c) H₂, Pd/C, EtOAc, 1 atm, 50%; (d) chiral HPLC separation, 45%, >99% ee; (e) 3 N HCl, THF, 22 °C, 98%; (f) (COCl)₂, DCM then i-PrOH, TEA, 97%; (g) EDC, HOBt, DIEA, N-methylaniline, DMF, 88%; (h) 2-trimethylsilylethanol, DPPA, TEA, 4 Å MS, 1,4-dioxane, 75 °C, 78%; (i) TFA, CH₂Cl₂, 22 °C, 99%.

The acid 11 was found to have poor potency in our cellular in vitro assays (Table 1), but the ester 12 and amide 13 had promising binding and functional antagonist activity in the MDA-MB-453 cell-line. Unfortunately, neither the amide nor the ester had potency that was superior to the exohydroxy analogue 5, so we investigated additional functionalities on the oxabicycle in an effort to find highly potent AR antagonists. The endoamine 14 could be prepared from the acid 11 by sequential Curtius rearrangement¹³ and subsequent TFA-promoted cleavage of the resulting Teoc-carbamate in 77% yield. Compound 14 had only modest affinity to the AR but offered a good handle with which to further functionalize the oxa-bicyclic core.

A series of amides, sulfamides, carbamates, ureas, and sulfonamides was prepared from the amine 14 in library format. Standard coupling techniques were utilized to prepare analogues 15–23 (see Supporting Information); the corresponding in vitro biological data is shown in Table 1. In general, these analogues were potent binders to, and functional antagonists of, the wild-type AR present in the MDA-MB-453 cell line. The amide 15 and the carbamate 18 had excellent potency in vitro, while the sulfamide 16 and urea 17 were weakly active. The sulfonamides 19–23 had the best overall in vitro profile, with robust affinity and potent antagonist activity, superior to that for bicalutamide (2).

We next wanted to investigate the pharmacodynamic effects of these novel AR antagonists in vivo. Compounds 13–23 were progressed into the immature rat prostate weight (IRPW) PK/PD model, where the compound effect on AR-dependent growth of the prostate and seminal vesicles was measured (Table 2).¹⁴ In this model, compounds were dosed orally once a day at 1 or 10 mg/kg for 4 days with plasma concentrations of drug measured 2 h postdose on day 4. Agents that effectively block the proliferative effect of the AR in these tissues would result in a decrease in the total weight of organs relative to a control group. As expected, the amide 13 had poor pharmacodynamic effect in this model, most likely due to modest functional antagonist potency. The amide 15 and the urea 17 gave only modest PD effects even with very high exposure after 1 and 10 mg/kg doses. This result correlated with the very high serum protein binding measured for these two compounds (>99% in mouse serum). The sulfonamide 16 and the carbamate 18 also had only modest PD effects, but this was most likely due to poor exposure relative to compound 2. The sulfonamide series stood out in the IRPW model by having excellent PD with modest exposure, suggesting superior in vivo potency compared to the ureas, amides, carbamates, and sulfamides. Compounds 19–23 all demonstrated robust PD at a 10 mg/kg dose with exposures significantly less than observed for bicalutamide. Of these promising analogues, the ethyl sulfonamide 23 was chosen for further studies due to robust potency in vivo and a promising PK profile in rats.

Table 2. Immature Rat Prostate Weight Assay Results.

#	IRPW 1 mg/kg SV/FBa	IRPW 10 mg/kg SV/FBa	exposure 1 mg/kg (μM)b	exposure 10 mg/kg (μM)b
2	69 ± 24	41 ± 4	2.3 ± 0.4	9.5 ± 1.4
13	88 ± 19	43 ± 8.0	1.8 ± 0.68	12.0 ± 0.82
14	89 ± 22	89 ± 8.1	0.08 ± 0.02	0.48 ± 0.32
15	87 ± 39	34 ± 6.0	24 ± 1.7	190 ± 73
16	89 ± 5.9	36 ± 4.2	0.01 ± 0.001	0.23 ± 0.08
17	115 ± 3.9	56 ± 8.9	8.3 ± 0.4	60 ± 6.1
18	105 ± 24	87 ± 16	0.012 ± 0.004	0.076 ± 0.017
19	64 ± 13	26 ± 4.2	0.29 ± 0.07	4.0 ± 1.2
20	63 ± 14	31 ± 5.2	0.23 ± 0.03	2.4 ± 0.52
21	56 ± 6.5	24 ± 1.8	0.16 ± 0.02	1.9 ± 0.06
22	44 ± 8.6	23 ± 11	0.05 ± 0.01	0.52 ± 0.09
23	58 ± 13	26 ± 3.0	0.79 ± 0.17	4.0 ± 0.42

^aSV/FB is the percentage of weight of the seminal vesicles over the full body weight of the rat (n = 3) where testosterone treated control = 100% and sham = 10%.

^bPlasma exposure measured 2 h postdose on day 4.

Compound 23 was further profiled to determine in vitro safety and ADME properties (Table 3). Inhibition of human cytochrome P450 (CYP) isoforms is very weak (>40 μM), and there is very low potential for CYP induction based on the human PXR transactivation assay. Plasma protein-binding of 23 was measured by equilibrium dialysis, and very low levels of plasma protein binding were observed in all species (>10% free). Compound 23 demonstrated excellent metabolic stability in hepatocyte incubations, and the predicted clearance in all species is low, especially for human. Finally, the PK properties of compound 23 were assessed in mouse, rat, and dog following both oral and IV doses. Consistent with the predicted hepatic clearance, compound 23 demonstrated moderate to long half-lives and very low clearance across species.

Table 3. Summary of Androgen Receptor Biological Data and ADME Properties for Compound 23^a.

assay	results
WT AR binding (Ki)	1.7 ± 0.56 nM
MDA 453 (IC50)	16 ± 3 nM
LNCaP (IC50)	153 ± 77 nM
human CYP (1A2, 2B6, 2C8, 2C9, 2D6, 3A4) IC50	>40 μM
PXR-TA EC50	>50 μM
hERG inhibition @ 30 μM	0%
protein binding (% free): mouse, rat, dog, human	14.8, 10.6, 16.5, 10.4
rate of metabolism in hepatocytes: mouse, rat, dog, monkey, human (pmol/min/106 cells)	0, 26, 3.0, 1.0, 2.0
predicted clearance: mouse, rat, dog, monkey, human (mL/min/kg)	0, 16, 1.8, 0.6, 0.8
IV/PO PK: T1/2 (h), CL (mL/min/kg), AUC0-∞ (μM·h), %F
mouse (5/10 mg/kg)	2.4, 4.5, 39.1/74.1, 94.7
rat (5/10 mg/kg)	3.7, 6.5, 31.7/40.5, 63.9
dog (1/2 mg/kg)	21.8, 0.35, 102.1/160.2,79

^aIC₅₀ values are an average of three experiments.

Compound 23 was then tested in the human prostate cancer xenograft model CWR22-BMSLD1 (Figure 3).¹⁵ Treatment with bicalutamide (2) (150 mg/kg, po, qd × 35 days) resulted in good tumor growth inhibition for the initial 10 days, followed by regrowth of the tumor at a rate that was similar to control. When compound 23 was dosed (90 mg/kg, po, qd × 45 days) excellent tumor growth inhibition was observed over the entire dosing period, demonstrating superior efficacy to bicalutamide (2). Additionally, we investigated the possibility of treating bicalutamide resistant tumors in this model by allowing the bicalutamide-treated tumors to triple in size followed by a switch to treatment with compound 23. As shown in Figure 3, tumor growth continued for ∼10 days after switching to compound 23, followed by nearly complete tumor growth inhibition for the remainder of dosing period. We were encouraged by this result as it gave strong evidence that compound 23 has the potential to treat forms of prostate cancer resistant to bicalutamide (2).

Figure 3.

In vivo efficacy of compound 23 in the CWR22-BMSLD1 prostate cancer xenograft model.

Docking compound 23 into the wild-type AR ligand binding domain (Figure 4)¹⁰ revealed the H-bond from N-705 to the bridging oxygen is likely intact as observed for compound 5. To accommodate the endosulfonamide at C-5, F876 must reorient resulting in a positional shift of Helix-11. This significant shift of Helix-11 results in a change in the overall architecture of the LBD, potentially giving compound 23 the promising antagonist profile presented here.

Figure 4.

Compound 23 docked into a model of the WT AR ligand binding domain.

Further profiling of compound 23 demonstrated an acceptable preclinical safety profile both in vitro and in vivo. This compound was selected for clinical development and advanced into phase I clinical trials.^15,16

In summary, we have utilized structure-based design to identify a new series of amino [2.2.1]-oxabicyclosuccinimide AR antagonists. Lead molecules demonstrated potent antagonist activity in cellular binding and transactivation assays in vitro and had robust PK/PD profiles in the IRPW model. Compound 23 was shown to be superior to bicalutamide in the CWR22-BMSLD1 human CaP tumor xenograft model and has the potential to address acquired bicalutamide resistance based on the results from these studies.

Supporting Information Available

Synthetic procedures and characterization data for compounds 9–23 and biological methods. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.5b00173.

The authors declare no competing financial interest.

Dedication

We would like to dedicate this paper to the memory our colleague John D. DiMarco.