Tetrahydroisoquinoline Phenols: Selective Estrogen Receptor Downregulator Antagonists with Oral Bioavailability in Rat

Abstract

A series of tetrahydroisoquinoline phenols was modified to give an estrogen receptor downregulator-antagonist profile. Optimization around the core, alkyl side chain, and pendant aryl ring resulted in compounds with subnanomolar levels of potency. The phenol functionality was shown to be required to achieve highly potent compounds, but unusually this was compatible with obtaining high oral bioavailabilities in rat.

The estrogen receptor (ER) is a clinically validated target for the treatment of hormonally driven breast cancer.¹ The discovery of the ER antagonist tamoxifen represented a breakthrough in targeted cancer therapy with activity due to the formation of the 4-hydroxy metabolite, which mimics the phenol of the endogenous ligand estradiol (1).² Considerable efforts have been made to develop alternative selective estrogen receptor modulators (SERMs) with a common issue encountered being that the phenol moiety required for high levels of binding to the estrogen receptor frequently leads to low bioavailability, often attributed to glucuronidation of the phenol. A notable exception is lasofoxifene (2), which exhibits low clearance and high oral bioavailability across species, attributed to its nonplanar topology reducing the rate of gut wall glucuronidation.³

A further step change in treatment for hormonally driven breast cancer occurred with development of selective estrogen receptor downregulators (SERDs) such as fulvestrant.⁴ These molecules are capable of both antagonizing estrogen receptor driven cell proliferation as well as effecting a conformational change resulting in ubiquitinylation and degradation of the receptor. Although effective, the low oral bioavailability of fulvestrant means that it is therapeutically administered as a monthly intramuscular implant. Identification of an orally bioavailable SERD would therefore address an unmet medical need. We recently reported our initial attempts to develop a coumarin based series of SERDs.⁵ Further efforts in this area have resulted in the disclosure of 3 (ARN-810)⁶ and 4 (AZD9496),⁷ both of which are currently in Phase I clinical trials. Herein we report an investigation, carried out in parallel to the discovery of 4, on a phenolic tetrahydroisoquinoline (THIQ) scaffold with the aim of identifying a potent, orally bioavailable SERD.

Phenolic THIQs with basic side chains have been previously shown to act as SERMs by groups at Pfizer⁸ and Novartis^9,10 with compounds such as 5 and 6 acting as antagonists at the receptors (Figure 1). Notably 6 was reported to have good bioavailability in rat (F = 49%).¹⁰ We speculated that replacement of the basic substituent with an acrylic acid could confer the desired antagonist downregulator phenotype and may provide a way to generate orally bioavailable SERDs. We therefore prepared a series of THIQ derivatives with representative examples described below (Scheme 1).

Figure 1.

Chemical structures of estradiol (1), lasofoxifene (2), ARN-810 (3), AZD9496 (4), and examples of THIQ SERMs from Pfizer (5) and Novartis (6).

Scheme 1. Synthesis of Key THIQs in the Unsubstituted, (3R)-Methyl, and 3,3-Dimethyl Series.

Reagents and Conditions: (a) (i) Et₃N, toluene, 90 °C, 2 h; (ii) TFA, 90 °C, 1 h, 84%; (b) NaBH(OAc)₃, DCE/AcOH, r.t., 71%; (c) LiOH, MeOH/H₂O, 40 °C, 4 h, 91%; (d) ⁱPr₂NEt, 4-bromobenzoyl chloride, DCM, r.t., 18 h, 84%; (e) (i) POCl₃, 110 °C, 3 h; (ii) NaBH₄, MeOH, 0 °C, 1 h, 97%; (f) ^tBu acrylate, PdCl₂(dtbpf), N(Me)Cy₂, Bu₄NCl, MeCN, 150 °C, 45 min, 42% trans, 25% cis; (g) BBr₃, DCM, −78 °C to r.t., 3 h, 85%; (h) TFA, DCM, r.t., 1 h, 56%; (i) (i) TBDMSCl, N-methylmorpholine, THF, r.t., 3h; (ii) PhN(SO₂CF₃)₂, Et₃N, DCM, r.t., 18 h, 65%; (j) Pd(OAc)₂, Et₃SiH, (Ph₂P)₂C₃H₆, DMF, 60 °C, 6 h, 36%; (k) (i) K₃PO₄, MeCN/H₂O, 50 °C, 18 h; (ii) TFA, 100 °C, MW, 2 h, 46%; (l) TBDMSCl, DMAP, pyridine, reflux, 7 h, 80%; (m) ⁱBuOTf, iPr₂NEt, 1,4-dioxane, r.t., 24 h, 84%; (n) LiOH, MeOH/THF/H₂O, r.t., 5 h, 95%.

The unsubstituted THIQ core of 32 was constructed via a Pictect–Spengler reaction between 3-(2-aminoethyl)phenol (30) and (E)-methyl 3-(4-formylphenyl)acrylate (31). Reductive amination allowed selective N-alkylation to 33 and subsequent ester hydrolysis gave 10, which was separated into enantiomers 10a and 10b using chiral chromatography (Scheme 1, steps a–c).

For a series of 3-methyl substituted THIQs, chiral amine 34 was converted to the corresponding 4-bromo aryl amide 35 then subjected to a Bischler–Napieralski reaction to afford the THIQ core 36 as an inseparable mixture of trans/cis isomers (Scheme 1, steps d–j). A Heck reaction was used to convert the bromide to the acrylic ester 37, and at this stage the mixture was separated to give enantiomerically pure trans isomer 37a in 42% yield over the two steps. Treatment with boron tribromide converted the 6-methoxy group to the free phenol and hydrolyzed the ester to the acid 12. Alternatively the ester could be selectively hydrolyzed using trifluoroacetic acid to give methyl ether 18. The des-phenol 17 was obtained by transiently protecting the acid of 12 as the silyl ester, converting the phenol to the triflate 38 and then treating with triethylsilane under palladium catalysis.

The dimethyl THIQ core 41 was constructed via a Pictect–Spengler reaction between O-benzyl protected phenethylamine 39 and aldehyde 40 with debenzylation of the phenol group occurring under the reaction conditions (Scheme 1, steps k–n). Silyl protection of the phenol gave 42 followed by N-alkylation to 43. Treatment with lithium hydroxide hydrolyzed both the ester and silyl group to give 25, which was separated into individual enantiomers 25a and 25b using chiral chromatography.

Compounds were profiled in a suite of assays designed to identify compounds with the requisite ER downregulator-antagonist profile (Table 1).¹¹ Physicochemical properties were also measured in an effort to highlight compounds suitable for oral dosing, including a glutathione (GSH) trapping assay designed to screen for the formation of reactive metabolites.¹² With the addition of the acrylic acid functionality to the THIQ scaffold, N-phenyl (7) and N-benzyl (8) THIQs were shown to bind to the receptor and have a downregulator-antagonist phenotype. Exploration of the SAR around the phenyl ring identified that activity in the binding and downregulation assays could be increased by the addition of substituents such as p-OCF₃ (9), consistent with previously reported SAR,⁵ but the compounds were too lipophilic (logD_7.4 3.7) and highly bound to plasma proteins (Rat PPB 0.06% free). As an alternative approach, we also examined N-alkyl substituents such as ⁱBu (10), which allowed a shift to lower lipophilicity space (logD_7.4 1.3) while maintaining downregulator-antagonist activity. Separation of the individual enantiomers showed that the activity resided predominantly in one isomer (10a), which is consistent with work carried out previously.¹³ Subsequent profiling showed reasonable free drug levels (Rat PPB 3.7% free) and moderate clearance in rat hepatocytes (38 μL/min/10⁶ cells), which translated to a modest clearance in vivo (44 mL/min/kg) and some oral levels (F 26%) when dosed in a pharmacokinetic study to Han Wistar rats (Table 6). Given the high potency in the downregulation assay (IC₅₀ 4.4 nM) and the encouraging oral exposure in rat we felt that 10a represented a promising hit for further optimization.

Table 1. Data for Variation of N-Substitution on the THIQ Core.

Cpd	ER bind IC50 (nM)a	ER DR IC50 (nM)b	LogDc	Rat/Hu % freed	Rat/Hu heps Clinte	GSH ratiof
7	4.7	71	2.4	1.2/1.2	87/67	0.25
8	36	90	2.2	1.3/0.7	40/150	0.20
9	0.96	35	3.7	0.06/0.06	20/11	0.56
10a	26	4.4	1.3	3.7/6.3	38/38	<0.04
10b	1100	42	1.4	8.7/1.9	49/4	0.06

^aER binding based on n ≥ 2 with SEM within 0.2 units except for 10b (n = 1).

^bER downregulation based on n ≥ 3 with SEM within 0.2 units.

^clogD_7.4 determined by shake flask method.

^dDetermined from DMSO stock solution by equilibrium dialysis in 10% plasma from Alderley Park Han Wistar rats or 10% human plasma supplied by Quintiles.

^eRate of metabolism (μL/min/10⁶ cells) determined from DMSO stock solution in isolated rat or human cryopreserved hepatocytes diluted to 1 × 10⁶ cells/mL.

^fAmount of GSH adduct formed relative to clozapine as described in ref (12).

Table 6. Rat Pharmacokinetic Parameters for THIQs.

Cpd	Clpa (mL/min/kg)	Vdss (L/kg)	PO t1/2 (h)	IV t1/2 (h)	bioavailability F (%)
10a	44	0.5	3.0	0.2	26
12	49	0.6	3.9	0.2	86
13	19	1.6	5.6	2.4	33
14a	21	5.9	7.4	5.8	66
19	63	0.6	1.9	0.2	3
24	4.2	1.1	6.7	4.2	79

^aCompounds were dosed intravenously at 0.5 mg/kg in 5% DMSO/95% hydroxylpropyl β-cyclodextrin or 10% DMA/90% Captisol and orally as 1 mg/kg using a 0.1% HMPC/tween suspension or 10% DMSO/90% cyclodextrin solution. The data shown is the mean of the data generated in two male rats.

Placement of a methyl group at varying positions around the core was investigated as a way of mapping out the SAR (Table 2). 1-Methyl substitution resulted in an enantiomeric set of compounds, which again showed the SERD activity predominantly resided in one enantiomer (11a).¹⁴ Despite a large increase in lipophilicity associated with this change (ΔlogD_7.4 + 0.9), the potency dropped slightly (ΔpIC₅₀ – 0.4). In contrast, 3-(R)-methyl substitution (12) added less lipophilicity (ΔlogD_7.4 + 0.4) but showed a proportionate increase in activity in the downregulation assay (ΔpIC₅₀ + 0.5).

Table 2. Data for Methyl Substitution of the THIQ Core.

Cpd	ER bind IC50 (nM)	ER DR IC50 (nM)	LogD	Rat/Hu % free	Rat/Hu heps Clint	GSH ratio
10a	26	4.4	1.3	3.7/6.3	38/38	<0.04
11a	44	11	2.2	1.9/0.9	33/40	0.16
12	19	1.3	1.7	7.5/7.0	22/33	0.11
13	12	5.5a	2.7	1.3/1.5	15/18	0.66
14a	39	7.0	1.5	11/3.5	7/6	0.17
15	610	170	1.6	6.3/2.4	32/9	0.06
16	160	43	1.8	8.8/9.1	17/17	0.09

^aResult based on n = 2, SEM 0.14.

We re-examined 1-methyl substitution in combination with the 3-(R)-methyl (13) on the basis the methyl groups may have a conformational influence on each other, but found a similar rise in lipophilicity (ΔlogD_7.4 + 1.0) and reduction in potency (ΔpIC₅₀ – 0.6) similar to that seen with 10a and 11a. The GSH trapping increased with 13, and this was attributed to the increased lipophilicity. Addition of a second methyl group at the 3-position resulted in a compound with slightly lower logD (−0.2 relative to 12) with one of the enantiomeric pairs (14a) significantly more active in the binding assay.¹⁵ Notably the effect of moving to the gem-dimethyl substitution led to a marked reduction in turnover in both human and rat hepatocytes compared to the unsubstituted or monomethyl analogues.

An X-ray crystal structure of 10a in complex with a construct of the ERα ligand binding domain (Figure 2a) showed the phenol group interacting with a conserved water and both elements of a salt bridge formed between residues Glu-351:Arg-394 with distances of 3.1/2.6/3.1 Å consistent with formation of a strong hydrogen bond network. Structures of 11a and 12 (Figure 2b) showed minimal perturbation of the tetrahydroisoquinoline phenol with the 3-(R)-methyl group filling the “Phe-404:Phe-425 lipophilic hole” as described previously.⁷ Methyl substitution ortho to the phenol led to a reduction in potency, more marked in the 5-methyl (15) than the 7-methyl (16) compound. This was consistent with a significant shift in the position of the core to accommodate the 5-methyl substitution as shown in Figure 2c and a corresponding lengthening of the interactions of the phenol with the water and salt bridge (3.6/2.7/3.6 Å). Compounds 12, 13, and 14a were profiled in rats and showed acceptable pharmacokinetic profiles with improved bioavailability relative to 10a (Table 6).

Figure 2.

X-ray crystallography of THIQs bound to the ERα ligand binding domain construct; (a) 10a (pdb code 5fqr) overlaid with AZD9496 (pdb code 5acc); (b) 10a overlaid with 11a (pdb code 5fqs) and 12 (pdb code 5fqp); (c) 10a overlaid with 15 (pdb code 5fqv) and 16 (pdb code 5fqt).

In order to evaluate the contribution of the phenol to the potency of 12, we made the corresponding compounds where the phenol was deleted (17) or capped with a methyl group (18) (Table 3). As expected, both compounds lost significant activity in both the binding and downregulation assays consistent with loss of the phenol interaction seen in the X-ray crystal structures and our previously reported findings.⁵ Notably, the compounds without the phenol showed no activity in a GSH trapping assay, whereas phenol 12 showed a small but reproducible signal potentially indicating a reactive metabolite risk.

Table 3. Data for Phenol Replacement in the THIQ Core.

Cpd	ER bind IC50 (nM)	ER DR IC50 (nM)	LogD	Rat/Hu % free	Rat/Hu heps Clint	GSH ratio
12	19	1.3	1.7	7.5/7.0	22/33	0.11
17	230	82	2.5	1.6/2.1	24/45	<0.04
18	390	160	2.3	3.4/2.0	43/56	<0.04

In the 3-(R)-methyl series, exemplified by 12, we elected to examine the SAR around the N-alkyl substituent and investigate di-ortho fluorination of the pendant aryl ring as a potential way to increase potency based on previous work on an alternative scaffold.⁷ We initially looked to reduce lipophilicity by introduction of an ether to the 3-position of the N-alkyl chain (19). This resulted in a lowering of logD (ΔlogD_7.4 – 0.7 relative to 12) and reduced the amount of GSH trapping observed but was lower in potency. Introduction of di-ortho fluorine substituents (20) did have the desired effect of increasing potency but unfortunately led to increases in both rat and human hepatocyte clearances. Addition of a 3-F to the alkyl chain (21) and the di-ortho fluorinated aryl analogue (22) were both subnanomolar downregulators but suffered from high hepatocyte clearances. Moving the fluorine to the 2-position of the alkyl chain (23) further increased the potency but did not ameliorate the high metabolism or GSH signal seen in vitro. As an alternative approach, we elected to revisit the p-OCF₃ aryl substituent in the methyl series (24) in the hope that the increased lipophilicity would result in reduced free drug levels and consequently the compound would be protected from metabolism. This was effective in that, despite the increased lipophilicity, the compound showed low turnover in both rat and human hepatocytes and had excellent pharmacokinetics in rat with low clearance (Cl 4.2 mL/min/kg) and high bioavailability (F 79%) (Table 6). Despite the high lipophilicity (logD_7.4 4.2), the potency was lower than seen with the alkyls and the GSH signal remained.

Table 4. Data for Compounds with the 3-(R)-Methyl THIQ Core.

Cpd	ER bind IC50 (nM)	ER DR IC50 (nM)	LogD	Rat/Hu % free	Rat/Hu heps Clint	GSH ratio
12	19	1.3	1.7	7.5/7.0	22/33	0.11
19	83a	5.8	1.0	12/16	27/7	0.07
20	2.8	1.2	1.6	5.3/3.8	110/45	0.33
21	6.2	0.40	1.5	2.4/2.6	47/100	1.4
22	1.3	0.37	2.5	1.0/0.9	94/170	0.23
23	0.97	0.25	2.1	1.1/2.3	74/78	0.43
24	1.7b	6.1c	4.2	0.02/0.9	9/9	0.59

^aResult based on n = 4, SEM 0.22.

^bResult based on n = 3, SEM 0.29.

^cResult based on n = 3, SEM 0.25.

Based on the observation that the 3,3-gem-dimethyl substituted core was less potent but more metabolically stable then the 3-(R)-methyl core (14a vs 12), we profiled some combinations of optimal N-alkyl substituents both with and without di-ortho fluorination in the aryl ring. In each case the data for the most potent enantiomer is shown.¹⁵ These compounds were generally less potent than the corresponding matched pairs in the 3-(R)-methyl series¹⁶ but in most cases were more metabolically stable in hepatocytes. Two examples (25a and 27a) were profiled in a rat pharmacokinetic study and both showed high bioavailability (F > 70%). Unfortunately, the in vitro GSH signal was seen with all examples profiled consistent with the presence of the phenol.

As a representative example of this series, compound 12 was profiled further (Table 7). The pK_a of the basic nitrogen was found to be 8.6 and the presence of this cation together with the anion of the acrylic acid (pK_a 4) indicates that 12 will be predominantly zwitterionic at neutral pH.¹⁷ The pK_a of the phenol was measured to be 10.4. The compound showed good aqueous solubility and high membrane flux in a Caco-2 permeability assay and no significant efflux. No activity was observed against the hERG ion channel (IC₅₀ > 30 μM) or five isoforms of the cytochrome P450 enzymes (IC₅₀ > 30 μM) and the compound had a clean profile (IC₅₀ > 30 μM) against a panel of 157 other targets at Cerep with only four weak hits.¹⁸

Table 7. Physicochemical Parameters for THIQ 12.

acid pKaa	phenol pKaa	base pKaa	solubility (μM)b	Caco-2 Pappc	hERG IC50 (μM)d	CYPs IC50 (μM)e
4.0	10.4	8.6	>1500	15(AB) 11(BA)	>30	all > 30

^apK_a measured using UV absorption (Sirius) in 0.1 M potassium chloride at 20 °C.

^bSolubility of compounds in aqueous phosphate buffer at pH 7.4 after 24 h at 25 °C.

^cCompounds were incubated at 10 μM in cultured Caco-2 cells. Permeability was measured in both the apical to basolateral (A to B) and basolateral to apical (B to A) directions.

^dInhibition of the hERG tail current was measured using a plate-based planar patch clamp system (IonWorks).¹⁹

^eInhibition of cytochrome P450 enzymes IC₅₀.

In conclusion, we have demonstrated that by inclusion of an acrylic acid motif, an ER downregulator-antagonist profile could be achieved in a series of tetrahydroisoquinoline phenols. Optimization around the core, alkyl side chain, and pendant aryl ring resulted in compounds with subnanomolar levels of potency. The phenol functionality was shown to be required to achieve highly potent compounds, but unusually this was compatible with obtaining high oral bioavailabilities in rat across a range of lipophilicity. These compounds were observed to trap glutathione in an in vitro trapping assay, potentially indicative of a reactive metabolite risk and, as a consequence, were not progressed. Further studies on efforts to reduce this risk will be reported in due course.

Glossary

ABBREVIATIONS

DR

downregulation

ER

estrogen receptor

GSH

glutathione

PPB

plama protein binding

SAR

structure–activity relationship

SERD

selective estrogen receptor downregulator

SERM

selective estrogen receptor modulator

THIQ

tetrahydroisoquinoline

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.5b00413.

• Experimental procedures for the synthesis of representative compounds together with biological data and associated errors (PDF)

Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.

Table 5. Data for Compounds with the 3,3-gem-Dimethyl-THIQ Core.

Cpd	ER bind IC50 (nM)	ER DR IC50 (nM)	LogD	Rat/Hu % free	Rat/Hu heps Clint	GSH ratio
14a	39	7.0	1.5	11/3.5	7/6	0.17
25a	1.8	1.6	2.7	0.4/0.5	34/33	0.21
26a	7.5	9.9	2.0	4.2/1/0	180/88	0.86
27a	7.4	1.3	1.8	2.8/1.0	24/14	0.58
28a	0.69	1.6	3.1	0.5/0.2	68/39	0.46
29a	2.2	6.5	2.3	2.5/0.7	59/19	0.68