Discovery and Characterization of 2-Aminooxazolines as Highly Potent, Selective, and Orally Active TAAR1 Agonists

Abstract

2-Aminooxazolines were discovered as a novel structural class of TAAR1 ligands. Starting from a known adrenergic compound 1, structural modifications were made to obtain highly potent and selective TAAR1 ligands such as 12 (RO5166017), 18 (RO5256390), 36 (RO5203648), and 48 (RO5263397). These compounds exhibit drug-like physicochemical properties, have good oral bioavailability, and display in vivo activity in a variety of animal models relevant for psychiatric diseases and addiction.

Trace amines (TAs) are metabolites of amino acids with structural similarity to biogenic amines and represent the endogenous ligands of the trace amine associated receptor 1 (TAAR1).^1,2 Dysregulation of TAs in the brain has been linked to a variety of psychiatric diseases and selective TAAR1 ligands have gained much interest as potential therapeutics for depression, schizophrenia, bipolar disorder, ADHD, and psychostimulant addiction.³

The modulatory role of this G protein-coupled receptor on monoaminergic neurotransmission has recently been investigated and confirmed by characterization of a transgenic mouse line overexpressing TAAR1 in central nervous system neurons.⁴ Further efforts were made to identify potent and selective TAAR1 agonists with favorable pharmacokinetic properties in order to prove the modulatory effect on dopaminergic signaling in vivo.⁵

In an endeavor to discover a novel and selective TAAR1 chemotype, we considered application of the SOSA approach (Selective Optimization of Side Activities) to adrenergic ligands as a viable lead identification strategy.⁶ Structural similarity of TAAR1 agonists with adrenergic agonists was already known from our previous work⁵ pointing toward a similarity of the binding regions of both receptors.^7,8 Therefore, we searched the literature and in-house databases for adrenergic ligands reported as drugs or development candidates, whereupon the alpha 2 adrenergic receptor partial agonist S18616 from Servier (1) caught our attention.⁹ Pleasingly, testing this candidate revealed (besides its expected activity at the human α_2A receptor) a high functional activity at human TAAR1 (EC₅₀ = 15 nM).¹⁰

A medicinal chemistry optimization program was then started aiming for compounds selective for TAAR1, where the known TAAR1 pharmacophore motif (aromatic moiety linked to a basic headgroup) of S18616 was kept, but the linker region was modified by opening the central six-membered ring (Figure 1). For all such derived compounds selectivity data was obtained by measuring functional activity at the human TAAR1 receptor (EC₅₀ hTAAR1) and in addition at the human adrenergic α_2A receptor (hα_2A) (Table 1).

Figure 1.

S18616 (1) and derived 2-aminooxazolines.

Table 1. Yields, Functional Activity Data at hTAAR1, and Selectivity vs hα_2a for Compounds 1–14.

compd	R1	A	R2	configuration	synthetic yield	hTAAR1 EC50 (nM)	hTAAR1 Emaxa	functional selectivity ratio hTAAR1 vs hα2A
1						15	60%	0.047
2	H	CH2	o-Cl	S	41%	154	100%	2.1
3	H	CH2	H	S	66%	330	100%	1.3
4	H	CH2	H	R	54%	2900	47%	n.d.
5	H	CH2–CH2	m-Cl	S	56%	18	87%	57
6	H	CH2–CH2	H	S	42%	27	72%	31
7	Me	CH2–CH2	m-Cl	S	21%	330	90%	2.0
8	H	CH2–O	m-Cl	S	76%	270	90%	6.4
9	H	CH2–NH–	m-Cl	S	60%	580	58%	4.8
10	H	CH2–NMe–	m-Cl	S	47%	27	84%	3.8
11	H	CH2–NEt–	m-Cl	S	41%	29	85%	38
12	H	CH2–NEt–	H	S	53%	59	87%	36
13	H	CH2–NiPr–	H	S	47%	140	58%	48
14	H	CH2–NEt–	H	R	51%	230	94%	8

^aThe E_max value describes the degree of functional activity compared to 100% for the natural ligand and full agonist phenethylamine.

All 2-aminooxazolines were synthesized from the corresponding amino alcohols 15 and cyanogen bromide in the presence of a base as depicted in Scheme 1. The enantiomerically pure amino alcohols were obtained from the chiral pool (e.g., via reduction of amino acids or their derivatives). The procedures are described in the Supporting Information.

Scheme 1. Synthesis of 2-Aminooxazolines from 15.

Reagents and conditions: (a) BrCN, K₂CO₃, THF, RT, 18 h, 21–76%.

We observed that the (S)-benzyl derivatives such as 2 or 3 showed functional activity at hTAAR1 but were not selective vs α_2A (Table 1). In contrast, the (S)-phenethyl derivatives such as 5 or 6 were much more potent hTAAR1 ligands and, surprisingly, showed promising selectivity vs the adrenergic receptor. An additional methyl substituent at the chiral center as in 7, however, was less tolerated.

To establish further SAR and to avoid a metabolically labile benzylic position we investigated derivatives 8 to 14 where the benzylic carbon was replaced by oxygen or nitrogen. The N-alkyl derivatives 10–12 showed excellent functional TAAR1 activity, and we decided to characterize compound 12 further. Compound 12 (= RO5166017) exhibited high potency at mouse TAAR1 (EC₅₀ = 3 nM) and rat TAAR1 (EC₅₀ = 14 nM) as well, was highly selective (as evaluated from radioligand binding assays for a panel of other targets, performed at Cerep¹¹), and was successfully used as the first tool compound from this new aminooxazoline class to be tested in mouse experiments.^12,13

However, we observed high metabolic clearance of 12 in our second model species, the rat, which would limit the use of this compound in behavioral studies. We attributed this to the presence of the central C–N linker motif and could confirm N-dealkylation as the major metabolic pathway by metabolite ID.¹⁰ In addition, testing 12 and analogues for GSH adduct formation, a screen for reactive metabolite formation,¹⁴ revealed the pronounced formation of GSH adducts in vitro upon metabolic activation in both rat and human liver microsomes. These findings were attributed to the presence of the aniline structural motif, which is linked to reactive metabolite formation via oxidative generation of quinone imine intermediates.¹⁵ Owing to foreseeable difficulties for clinical development of compounds, which might pose a risk for idiosyncratic toxicity due to reactive metabolite formation, we decided to reconsider carbon analogues, but this time introducing an additional methyl or ethyl substituent at the benzylic position as shown in Table 2.

Table 2. Functional Activity Data at hTAAR1 and Selectivity vs hα_2a for Compounds 16–19.

compd	R	hTAAR1 EC50 (nM)	hTAAR1 Emax	sel. ratio vs hα2A
16	(S)-Me	1540	79%	0.85
17	(R)-Me	730	84%	1.2
18	(S)-Et	18	98%	568
19	(R)-Et	2260	88%	3.5

Of these examples the (S,S)-ethyl-diastereomer 18 (= RO5256390) turned out to be a very potent full agonist at hTAAR1 and >500-fold selective vs α_2A. Evaluation in the Cerep¹¹ panel confirmed its selectivity against 155 other targets, and pharmacokinetic properties in mouse and rat were excellent making 18 an ideal tool for further behavioral tests.¹⁶

To access derivatives 16–19 a stereoselective synthesis was developed as depicted in Scheme 2 for compound 18. Stereoselective pseudoephedrine enolate alkylation according to the methodology of Myers¹⁷ and subsequent reduction yielded alcohol 26 in 91% ee, which was converted to iodide 27. The second stereogenic center was introduced using Schöllkopf bis-lactimether methodology¹⁸ (94% de for 29). Chromatographic purification yielded amino ester 30, which was converted in two steps to 18 (99.6% ee).

Scheme 2. Synthesis of 18 (RO5256390) by Application of Myers and Schöllkopf Methodology.

Reagents and conditions: (a) (−)-pseudoephedrine, Et₃N, CH₂Cl₂, 96%; (b) LDA, LiCl, THF, −78 °C–RT, 99%; (c) LiH₂NBH₃, THF, RT, 0 °C–RT, 74%; (d) I₂, PPh₃, imidazole, CH₂Cl₂, RT, 91%; (e) (i) 25, n-BuLi, HMPA, THF, −78 °C, (ii) add 24, THF, −78–0 °C, 91%; (f) TFA, MeCN, H₂O, RT, chromat., 62%; (g) NaBH₄, EtOH, 60 °C, 88%; (h) BrCN, NaOAc, MeOH, 0 °C–RT, 60%.

Having observed a strong influence of the linker region on activity at TAAR1 and on selectivity vs α_2A, we continued to make further variations in this part (see Table 3).

Table 3. Functional Activity Data at hTAAR1 and Selectivity vs hα_2a for Compounds 29–32.

compd	A	hTAAR1 EC50 (nM)	hTAAR1 Emax	sel. ratio
29	CH2–CH2–CH2	27	86%	217
30	CH2–O–CH2	360	102%	n.d.
31	CH2–CH2–O	9	98%	818
32	bond	67	104%	670

Interestingly, compounds with 3-atom-linkers were very active and selective as well (compounds 29 and 31, Table 3). However, mainly due to observed increased in vitro metabolic clearance¹⁰ of these rather flexible molecules we later abandoned this subseries in favor of the directly linked 2-amino-4-phenyloxazolines such as 32 (A = bond).

Figure 2 and Table 4 show a selection of substituents that we investigated for further optimization of these molecules.¹⁹ Lipophilic substituents in o- and m-position increased activity at hTAAR1 (33, 34), whereas substitution in p-position was less favorable (35, 37). However, for high activity at rat TAAR1 (rTAAR1) p-substitution was favorable, and a combination of m- and p-substituents led to derivatives that showed a reduced species difference, such as compound 36.

Figure 2.

Substituted 2-amino-4-phenyloxazolines.

Table 4. Functional Activity Data at Human TAAR1, Rat TAAR1, and Selectivity vs hα_2a for Compounds 33–48.

compd	R1	R2	R3	R4	configuration	hTAAR1 EC50 (nM)	hTAAR1 Emax	sel. ratio hTAAR1 vs hα2A	rTAAR1 EC50 (nM)	rTAAR1 Emax
33	H	Cl	H	H	S	23	100%	115	87	105%
34	H	H	Cl	H	S	21	73%	333	246	76%
35	H	H	H	Cl	S	143	64%	n.d.	52	64%
36	H	H	Cl	Cl	S	31	72%	94	8	58%
37	H	H	H	Br	S	150	93%	66	29	83%
38	H	H	H	Br	R	>10000		n.d.	n.d.
39	Me	Cl	H	H	S	165	76%	6.4	560	62%
40	Me	H	H	Br	S	41	53%	131	18	36%
41	H	H	H	Ph	S	2670	24%	1.9	428	80%
42	H	H	OPh	H	R/S	1950	34%	n.d.	n.d.
43	H	H	F	H	S	490	83%	n.d.	1920	68%
44	H	Me	H	H	S	67	78%	30	63	67%
45	H	Me	H	Cl	S	11	78%	405	1	93%
46	H	Et	H	Cl	S	26	83%	115	1	79%
47	H	cPr	H	Cl	S	12	77%	400	0.6	78%
48	H	Me	F	H	S	17	82%	1800	35	69%

Activity was strongly dependent on stereochemistry, the (S)-isomer was always more active (37 vs 38). Additional substitution R¹ = Me on the aminooxazoline ring made the compounds more partial agonists at hTAAR1 (39, 40). Larger substituents in m- and p-position were detrimental (41, 42); fluorination was less tolerated as well (43). ortho-Alkyl substituents, however, especially in combination with p-chloro-substitution, led to the most active and selective compounds (45–47). The most selective compound was methyl fluoro derivative 48. Compounds 32–48 were synthesized from phenylglycinols in a similar way to that described in Scheme 1.

We first selected the partial agonist 36 (= RO5203648) for further studies. A Cerep screen¹¹ showed high selectivity across a panel of 149 enzymes and receptors, and pharmacokinetic analysis revealed that this compound was well suited for in vivo studies in rat. To our knowledge, 36 was the first selective partial TAAR1 agonist tested in behavioral studies, and it clearly demonstrated antipsychotic, antidepressant, and antiaddictive activities in a number of animal models.^20,21

Unfortunately, additional in vitro testing of 36 revealed a very high metabolic clearance of this compound in human hepatocytes, which was not obvious in prior testing in human liver microsomes, and this led to a deselection of 36 for further development. These observations prompted us to compare in vitro microsomal clearance with in vitro hepatocyte clearance for a number of other compounds from the same series, whereby a surprisingly big discrepancy between the clearance data in the two assay systems was apparent for most of the compounds (Table 5).

Table 5. Comparison of in Vitro Clearance in Human Liver Microsomes (HLM) and in Human Hepatocytes (Hhep) for Compounds 36, 40, 45, 46, 47, 48, and 18.

compd	CL HLM [mL/min/kg]	CLm class	CL Hhep [mL/min/kg]	CLh class
36	1.8	low	17.5	high
40	0	low	14.6	med
45	3.6	low	15	high
46	4.0	low	13	med
47	4.2	low	10.1	med
48	7.7	med	14	med
18	1.1	low	2.9	low

In order to better understand the reasons for this discrepancy, we decided to elucidate unequivocally the metabolic clearance pathways by metabolite identification studies. The results indicated that N-glucuronidation, a comparatively rare metabolic transformation, occurred to a significant extent in human hepatocytes, being in most cases the primary clearance pathway for our 2-amino-4-phenyloxazoline compounds.²²

N-Glucuronidation did not take place under the standard assay conditions for our routine in-house liver microsome clearance screening assay because the necessary cofactors for glucuronidation were not present. This appeared to be much less of an issue for metabolic clearance in rodents, where the primary route of metabolism was found to be oxidation, with the consequence that clearance data from the hepatocyte and microsome assays was generally in agreement.¹⁰ Interestingly, for compound 18 from the other subseries very low clearance was confirmed in human hepatocytes, which we attributed to the increased steric demands of the branched linker present in compound 18.

Next we evaluated the propensity for the remaining compounds on our shortlist, namely, the compounds 40, 46, 47, and 48, which had medium clearance in human hepatocytes, to undergo covalent binding (CVB) to proteins during metabolism. Thus, in addition to the standard GSH adduct assay we measured covalent binding with ¹⁴C-labeled material after metabolic activation in human liver microsomes and (since differences in metabolic clearances had been observed) in human hepatocytes as well (Table 6). The methyl-substituted aminooxazoline 40 showed the lowest CVB values, which was in agreement with the classification by the in silico prediction tool MCASE²³ and consistent with the hypothesis that nucleophilic attack leading to opening the aminooxazoline ring can be diminished by introducing steric hindrance. For the other compounds including 18, varying degrees of covalent binding were detected, with all compounds, however, well below 100 pmol/mg protein that constitutes a development concern in conjunction with a high clinical dose.²⁴

Table 6. Measuring of Covalent Binding (CVB) after Metabolic Activation at Human Liver Microsomes (HLM) and Human Hepatocytes (Hhep) for Compounds 40, 46, 47, 48, and 18.

compd	MCASE structural alert	GSH HLM assay (adduct)	CvB HLM [pmol/mg protein]	CvB Hhep [pmol/106 cells]
40	NEG	none	5	9
46	POS	none	2	42
47	POS	M + GSHa	52	51
48	POS	M + GSHa	29	32
18	POS	M + GSHa	9	22

^aSlightly above detection limit.

After considering activity at hTAAR1, human hepatocyte clearance and applying further selection criteria such as inhibition of cytochrome isoforms (especially CYP2D6 with a 10-fold higher IC₅₀ compared to the other candidates) and selectivity vs other receptors we finally selected compound 48 (= RO5263397) as the most promising development candidate for a partial TAAR1 agonist and confirmed the selectivity of 48 against 155 target proteins by performing a CEREP screen.¹¹

Table 7 summarizes physicochemical and in vitro safety data for compounds 18 and 48, which readily supported further studies with these compounds. Pharmacokinetic analyses in rat, mouse, and cynomolgous monkey revealed very favorable in vivo properties, which have already been reported elsewhere.¹⁶

Table 7. Physicochemical and in Vitro Safety Characterisation of Compounds 18 (RO5256390) and 48 (RO5263397).

parameter	RO5256390	RO5263397
basic pKa	8.98	8.07
Aq. solubilitya (μg/mL)	>4980	5830
logD at pH 7.4	1.29	1.12
PAMPA25 Peff (10–6 cm/s)	10.7	14.5
hERG IC20 (μM)	9.2	10.0
CYP 3A4/2D6/2C9 IC50 (μM)	>50/3.6/>50	>50/32/>50
AMES/MNT	NEG/NEG	NEG/NEG
in vitro phototoxicity26	NEG	NEG

^a0.05 M phophate buffer, pH = 7.2–7.5, thermodynamic sol.

Testing both compounds in a variety of preclinical in vivo models revealed very interesting antipsychotic-like profiles.¹⁶ The TAAR1 partial agonist 48 in addition increases wakefulness in rats and is active in the forced-swim test (FST) in rats indicative of potential antidepressant activity.¹⁶ In addition, efficacy in reducing cocaine-mediated behaviors in animal models of substance abuse has recently been reported.²⁷⁻²⁹

In summary, we report here the discovery and optimization of 2-aminooxazolines as novel, selective, full and partial TAAR1 agonists. Starting from the known adrenergic ligand S18616 (1) and modifying the linker region and exploring additional SAR, we investigated several subseries of TAAR1 ligands. Besides functional activity at hTAAR1 and selectivity vs adrenergic α_2A receptor, metabolic stability measured in hepatocytes was used as a key parameter to finally select two molecules, 18 (RO5256390) and 48 (RO5263397), for further studies. Both compounds are active in a variety of behavioral models for schizophrenia and drug addiction and have been selected as candidates for GLP toxicity studies.

Glossary

ABBREVIATIONS

TA

trace amine

TAAR1

trace amine associated receptor 1

h

human

r

rat

SAR

structure–activity relationship

SOSA

selective optimization of side activities

MCASE

Multiple Computer Automated Structural Evaluation

GSH

glutathion

CvB

covalent binding

HLM

human liver microsomes

Hhep

human hepatocytes

CL

clearance

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.5b00449.

• Experimental details for the synthesis of compounds 2–48 (PDF)

Author Contributions

The manuscript was written through contributions of all authors.

The authors declare no competing financial interest.