Discovery of 1,2,4-triazine derivatives as adenosine A(2A) antagonists using structure based drug design.

Potent, ligand efficient, selective, and orally efficacious 1,2,4-triazine derivatives have been identified using structure based drug design approaches as antagonists of the adenosine A(2A) receptor. The X-ray crystal structures of compounds 4e and 4g bound to the GPCR illustrate that the molecules bind deeply inside the orthosteric binding cavity. In vivo pharmacokinetic and efficacy data for compound 4k are presented, demonstrating the potential of this series of compounds for the treatment of Parkinson's disease.

Introduction

The adenosine A2A receptor is expressed in the basal ganglia where it functionally opposes the actions of the dopamine D2 receptor, i.e., inhibition of the A2A receptor leads to enhancement of D2 receptor function. Given that the primary pathology in Parkinson’s disease is a loss of nigrostriatal dopamine and hence reduced dopamine D2 receptor activation, adenosine A2A receptor antagonism has emerged as a potential nondopaminergic therapy for this disorder. Preclinically, adenosine A2A receptor antagonists are effective in animal models of Parkinson’s disease, ranging from the reversal of haloperidol-induced catalepsy through to efficacy in more disease-relevant models such as 6-hydroxydopamine lesioned rats and MPTP-lesioned primates.[1] Furthermore, a number of these compounds have progressed into clinical development, the most advanced of which is currently preladenant. This compound was shown to be effective in a phase IIa trial of patients with moderate-to-severe Parkinson’s disease when administered in conjunction with levodopa, increasing on-time with no concomitant increase in dyskinesias.[2]

In the preceding publication, hit molecules derived from a virtual screening strategy were described.[3] One series of 1,3,5-triazine derivatives was identified and optimized to give potent and selective adenosine A2A receptor antagonists. As part of the optimization of this chemotype and by considering the proposed binding mode to the receptor, we hypothesized that the alternative 1,2,4-triazine isomers might also bind to the receptor but could sit more deeply in the receptor pocket accessing the region normally occupied by the ribose group of the natural ligand adenosine, in addition to mimicking the adenine ring itself. Upon testing of the commercially available parent 5,6-diphenyl-1,2,4-triazine-3-amine 4a, we discovered that the molecule was indeed an antagonist of the receptor (Table 1, compound 1; pKi = 6.93). In this paper, we outline our studies in this isomeric chemical series.

Table 1

SAR and in Vitro ADME Data for Compounds 4a to 4la

								SPR data
ID	formula	A2ApKi	A1pKi	LE[17]	RLM (min)	PPB (%)	kinetic solubility (μM)	ka	kd	KD	pKD
4a	X=C;R1–6=H	6.93	6.56	0.50	23	ND	>100	>5 × 107	>1 × 100	9.03 × 10–6	5.0
4b	X=C;R1=Cl;R2–6=H	7.29	7.25	0.50	29	97.9	13	3.79 × 105	1.68 × 10–1	4.42 × 10–7	6.4
4c	X=C;R1=R3=Cl;R2,4,5,6 =H	8.40	7.36	0.55	108	99.0	38	5.32 × 105	2.43 × 10–2	4.57 × 10–8	7.3
4d	X=C;R1=R3=Me;R2,4,5,6=H	7.67	6.71	0.50	9	98.0	20	ND	ND	ND	ND
4e	X=C;R1=Cl;R2=OH;R3,4,5,6 =H	8.85	9.79	0.57	69	98.0	45	4.07 × 106	1.01 × 10–3	2.48 × 10–10	9.6
4f	X=C;R1=R3=Me;R2=OH;R4,5,6=H	8.39	7.78	0.52	75	93.3	43	8.57 × 106	1.36 × 10–3	1.59 × 10–10	9.8
4g	X=N;R1=R3=Me;R4,5,6=H	8.11	7.07	0.53	100	82.1	40	9.92 × 106	1.15 × 10–2	1.16 × 10–9	8.9
4h	X=N;R1=R3=Me;R5=F;R4,6=H	7.81	6.40	0.48	100	69.0	43	1.13 × 107	1.15 × 10–1	1.02 × 10–8	8.0
4i	X=N;R1=R3=Me;R4,6=F;R5=H	7.56	6.77	0.45	100	ND	45	9.44 × 106	8.84 × 10–2	9.37 × 10–9	8.0
4j	X=N;R1=R3=Me;R4=F; R5,6=H	7.98	6.96	0.49	78	87.0	48	1.41 × 107	4.27 × 10–2	3.03 × 10–9	8.5
4k	X=N;R1=Me;R3=CF3;R4,5,6=H	8.46	7.50	0.48	86	92.0	35	1.08 × 106	3.73 × 10–3	3.45 × 10–9	8.5
4l	X=N;R1=Me;R3=CF3;R5=F;R4,6=H	8.34	6.93	0.45	97	93.0	34	1.55 × 106	4.09 × 10–2	2.63 × 10–8	7.6

RLM rat liver microsome half-life in mins; PPB rat plasma protein binding; SPR kinetics using A2A–StaR (see main text).

Results and Discussion

Design and Synthesis

The binding mode of 1,2,4-triazine derivatives was initially derived from modeling of representative compounds in an “experimentally enhanced” homology model of the adenosine A2A receptor (described in the preceding paper), refined using site directed mutagenesis data both from the literature and our own Biophysical Mapping (BPM) approach.[3−6] Parts A and B of Figure 1 illustrate the proposed binding mode of two analogues 4g and 4e (Table 1) in the orthosteric binding site of the receptor and also show the BPM binding fingerprint around these example ligands, used to refine and improve the model. The residues that, when mutated to alanine, reduce binding of each ligand are colored in red for nonbinding, dark-orange for the largest effect (tier 1), orange for the next largest effect (tier 2), yellow for the smallest effect (tier 3), and in green if the mutation caused an increase in binding of the ligand. The BPM studies have been reported elsewhere and analogues 4e and 4g here equate to examples 3b and 3d in the earlier publication.[4] As well as rationalizing the role of the aminoheterocyclic scaffold binding to Asn2536.55 (superscripts refer to Ballesteros–Weinstein numbering),[7] in particular the BPM fingerprints and modeling suggested that presence of a hydrogen bond acceptor at the para position of ring A of the ligands to His2787.43 with addition of one or more flanking lipophilic substituents on the same ring to interact with Ile662.64 and Ser2777.42 should be one focus of the SAR program (R1, R2, R3 positions in Scheme 1 and Table 1). In addition, careful 3D analysis of the GRID maps, calculated for both A2A and an A1 homology model, enabled the pharmacophoric preferences (hydrophobic, hydrogen-bond donor and acceptor) and shape constrictions/differences to be identified to allow an enhanced evaluation of each ligand docked into the binding site.[8,9] Very subtle binding site differences, such as A2A–Ser7.42 to A1–Thr7.42 and A2A–Ala2.57 to A1–Val2.57 and also ligand preferences between the two receptor models were exploited to overall enable the design of small, polar, selective, and ligand-efficient compounds.

Figure 1

(A,B) BPM fingerprint of 1,2,4-triazine adenosine A2A antagonists. Compounds 4g (A) and 4e (B) are illustrated bound to the orthosteric pocket of the receptor and the residues lining the pocket that interact with the ligands are labeled. The tier 1, 2, and 3 designation is described in the main text. The key hydrogen bonding to Asn2536.55 of the scaffold is highlighted by green dotted lines. (C,D) Illustration of the A2A–StaR2 ligand binding site in complex with compound 4g (C) and 4e (D). TM helices and visible extracellular regions are depicted in the rainbow format. Ligands are represented as stick models, carbon and chlorine atoms are green, oxygen atoms red, and nitrogen atoms blue. Residues involved in ligand binding are labeled and represented as gray sticks, oxygen atoms are red, and nitrogen atoms are blue. Extracellular loop 2, the key binding site residues and TM’s 1, 2, 5, and 6 are labeled for reference. Potential H-bonds between the ligand and receptor are represented as dashed blue lines. TM3 and TM4 have been omitted for clarity. (E) WaterMap calculation on the binding site of compound 4e (ligand removed for the calculation). Waters calculated are color coded to show the most “unhappy” vs bulk solvent as red (>3.5 kcal/mol), then yellow (2.2–3.5 kcal/mol), with gray intermediate (−1 to 2.2 kcal) and blue “happy” (<−1 kcal/mol). The CPK surface of the ligand 4e is shown as a red dot surface, clearly illustrating that the cluster of red and yellow “unhappy” waters deep in the binding site have been displaced. GRID maps are also shown that highlight the shape (Csp3 (C3) at 1 kcal/mol in light-gray), the lipophilic hotspots (aromatic CH probe (C1=) in yellow at −2.5 kcal/mol), and the water probe hotspots (in green wire mesh at −6.6 kcal/mol). (F) Alignment of the A2A homology model with 4e docked (cyan carbons) onto the crystal structure of A2A–4e complex (green carbons). The alignment was generated by the align algorithm in Pymol utilizing only helices where hydrogen bonds are formed with the ligand, helices 6 and 7. Helices 2, 3, and 4 are removed for clarity.

Scheme 1

Synthesis of 5,6-Biaryl-1,2,4-triazine-3-amine Derivatives (4) and 4-Pyridylboronic Acid Derivatives (6)

Reagents and conditions: (a) NBS, DMF, RT; (b) 3, Pd(PPh3)4, K2CO3, 1,4-dioxane/H2O, 150 °C; (c) [Ir(COD)OMe]2, DTBPY, [B(pin)]2, hexane, 50 °C.

During the SAR optimization process, a number of members of the series were successfully cocrystallized in the receptor using the published thermally stabilized adenosine A2A construct A2A–StaR2 (stabilized receptor or StaR) developed in our laboratories.[10] This allowed, for the first time to our knowledge, X-ray structure-directed optimization of a hit series to derive potent and selective leads for a G protein-coupled receptor. This reinforced our efforts focused on the optimization of the substitution patterns around the pendent aryl rings A and B in an atom efficient manner (Scheme 1 and Table 1) to engineer high potency and reduce affinity for the adenosine A1 receptor without significantly increasing molecular weight or lipophilicity. The X-ray structures also validated the BPM approach, which had successfully predicted the binding mode of the compounds but add a further level of understanding particularly with respect to teasing out selectivity for A2A over the A1 receptor.

Synthesis of the target compounds began with a set of purchased 3-amino-5-aryl-1,2,4-triazine derivatives 1, Scheme 1. These building blocks were further elaborated following treatment with NBS at room temperature to afford the corresponding 6-bromotriazines 2. A set of conditions was optimized to enable Suzuki cross-coupling of a diverse range of bromotriazines with commercially available boronic acid derivatives. Thus 2 and 3 were coupled at 150 °C in a sealed vessel in the presence of catalytic quantities of Pd(PPh3)4 to afford the biaryl triazines of interest (4). Access to bespoke 4-pyridylboronic acid derivatives was required to synthesize molecules suggested by the proposed binding mode in the receptor. Using a method described by Hartwig et al., commercial 2,6-disubstituted pyridines 5 were borylated in high yield, under iridium catalysis, with bis(pinacolato)diboron.[11−13]

X-Ray Crystallography

The overall structure of the A2A–StaR2 in complex with compounds 4e and 4g is in close agreement to the previously solved structures in our laboratory and methods for crystallization and a description of the general receptor architecture are described elsewhere.[10] Statistics for data collection and refinement are given in Supporting Information Table S1. The cocrystal structures of the A2A–StaR2 in complex with compounds 4g and 4e (Figure 1C,D) show clear positive omit density at 3.0σ (data not shown) for the presence and position of the ligands in the receptor binding pocket. The structure of A2A–StaR2–4g shows the amino-triazine core makes two critical donor and acceptor H-bonding interactions with the side chain of Asn2536.55 with bonding distances of 2.85 and 2.76 Å, respectively. In addition, the helical portion of extracellular loop (ECL) 2 is positioned for Phe168 to form a perpendicular π–π stack on one side of the core, while the side chain of Met2707.35 makes a hydrophobic interaction on the opposite side, completing the receptor interactions around this region of the ligand. The phenyl substituent from the C5 position of the triazine core occupies a hydrophobic pocket deeper inside the receptor flanked by Leu843.32, Leu853.33, Met1775.38, Asn1815.42, Trp2466.48, Leu2496.51, and His2506.52. The second substituent, dimethyl-pyridine, from the C6 carbon of the triazine core occupies the ribose binding pocket (of the natural agonist adenosine) defined by His2787.43 and Ser2777.42, with one methyl substituent pointing toward a hydrophobic region defined by Ala632.61 and Ile662.64 and the other pointing toward the stabilizing mutation Ser2777.42Ala.[14] Additionally, the side chain of His2787.43 is positioned 4.03 Å away from N4 of the dimethyl-pyridine substituent, precluding a direct H-bond between the ligand and this receptor residue. Compound 4e occupies a similar position overall to that of 4g in the A2A receptor ligand binding site but with specific differences. The amino-triazine core makes a similar set of interactions to the receptor as 4g, however, the bond lengths to Asn2536.55 increase to 3.12 and 3.1 Å in comparison. The driving force for this is an additional hydrogen bond, 2.87 Å in length, formed between the phenolic hydroxyl on the chloro-phenol substituent to Nε of His2787.43. Additionally, Glu131.37 has switched rotamer and is now poised to H-bond to Nδ1 of His2787.43. The additional conjugation of the ligand through the receptor pulls compound 4e ∼1.2 Å deeper into the ribose binding pocket in comparison to 4g (pivoting on the common phenyl group) and perhaps provides a basis for the slow off-rate receptor kinetics of these phenolic compounds (see Table 1). It is noted that two potential conformations for the phenyl and dimethyl-pyridine substituents (compound 4g), and the phenyl and chloro-phenol substituents (compound 4e), involving concomitant ∼50° rotations of each around the bond to the amino-triazine core, can exist in nature. In the structures presented here the conformations with the lower b-factors were submitted to the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB).

An analysis of the A2A binding site (with ligand 4e removed) using the WaterMap software (Schrödinger)[15] shows that these highly ligand efficient ligands (LE = 0.57 for 4e) occupy exactly the region where there is a cluster of “unhappy” waters (shown as red and yellow balls in Figure 1E) and not other less favorable regions (as for example less ligand efficient compounds such as ZM241385 bind, reported elsewhere).[16] The WaterMap software uses a molecular dynamics simulation on a full explicit water network to calculate the enthalpic and entropic energies of waters compared to bulk solvent. Finally, in Figure 1, as the initial SAR work was based on our BPM-optimized homology model, we include a comparison of a docked structure of 4e into the homology model and the protein–ligand X-ray structure, showing that the overall orientation and key hydrogen bond interactions were correctly predicted (Figure 1f). The biophysical mapping fingerprints of compounds 4e and 4g (published previously by Zhukov et al, as compounds 3b and 3d in Table 1), respectively, are in excellent agreement with the crystal structures, highlighting the significant interactions with Ile662.64, Leu853.33, Asn1815.42, and Asn2536.55 and confirming the initial deep placement of the ligand in the ribose sugar pocket of the A2A receptor.[4]

Structure–Activity Relationship

The parent compound 4a was identified as a ligand efficient adenosine A2A receptor antagonist (Table 1).[17] Having examined the putative binding mode in silico, SAR focused on simple substitution of rings A and B (Table 1). During the optimization process, the binding mode predictions were first improved using our BPM approach and then by determination of the crystal structures of compounds 4e and 4g (discussed above). The models suggested only small groups would be tolerated, particularly in ring B, and this was quickly evident because addition of chlorine or methyl in ring A and fluorine in ring B was tolerated or increased affinity (compounds 4a–d and 4h–j), but larger groups tended to be detrimental to potency (data not shown). The models also suggested the potential to form hydrogen bonding interactions from ring A at the para position R2, and introduction of either a phenolic hydroxyl (compounds 4e and 4f) or a 4-pyridyl nitrogen (compound 4g) were found to increase potency (compare 4f and 4g with 4d). A challenging aspect of the optimization was that affinity for the adenosine A1 receptor was generally also observed, and it was thought desirable to reduce affinity against this target to avoid any potential side effects caused by cross reactivity with this receptor. It was quickly noted, consistent with small differences in the binding sites predicted from a GRID analysis of models of the two receptors (discussed above), that 3,5-disubstitution in ring A introduced a modest selectivity for A2A over A1 (compare 4b with 4c and 4e with 4f). Also, although a very subtle effect, introduction of fluorine at R5 on ring B reduced activity against both receptors but slightly improved the selectivity window for A2A over A1 (compare 4g with 4h). Introduction of fluorine in other substitution patterns tended to reduce affinity further or had no additional selectivity benefit (compounds 4i and 4j). Overall, the derivatives were found to be highly selective over the adenosine A3 receptor, e.g., compounds 4c, 4g, and 4h had antagonist affinity of >10 μM. The series was generally 10-fold selective over the A2B subtype, and inhibition of this target was not thought to be detrimental therapeutically; 4c, 4g, and 4h had pKi values of 7.4, 7.3, and 6.8, respectively (A3 and A2B data were generated by Ricerca Biosciences, Taipei). Fine tuning of affinity by various combinations of small lipophilic substituents quickly led us to pyridyl analogues 4k and 4l, incorporating a CF3 group at the 3-position of ring A with methyl at the 5-position and with or without fluorine at the para position of ring B. These two compounds have the best balance of potency and selectivity, derived by introduction of simple substituents on the scaffold, of the examples shown in the Table.

Table 1 also details the binding kinetics of compounds from the lead series as measured using the adenosine A2A–StaR by surface plasmon resonance on a Biacore instrument (see Supporting Information for full details). The pKD values were generally in good agreement with the radioligand binding data, and the method also allowed comparison of on rates (ka) and off rates (kd) to the receptor between related compounds. Of most note in the data presented here is that the phenolic analogues 4e and 4f had very slow off rates, consistent with the change in binding mode observed in the crystal structure of 4e discussed above. Consideration of relative receptor kinetics was part of the decision making process used for selection of compounds for in vivo efficacy experiments.

Pharmacokinetics and in Vivo Efficacy

The examples in Table 1 generally exhibited good physicochemical and in vitro ADME properties, having moderate to high aqueous solubility, good stability in rat liver microsomes (RLM) and, where a polar substituent had been introduced (such as a pyridyl nitrogen atom), relatively low plasma protein binding (PPB). Another general trend was a lack of inhibition of cytochrome P450 enzymes and the hERG channel, with only the occasional outlier (data not shown). A number of compounds also demonstrated good pharmacokinetic properties in rat dosed either orally or intravenously. Data is given here for example 4k in a rat pharmacokinetic (PK) experiment; PK parameters are shown in Table 2. Compound 4k displayed moderate clearance (42 mL/min/kg), although due to a relatively high steady-state volume of 4.6 L/kg, the compound had an acceptable half-life of 1.2 h. Compound 4k was rapidly absorbed after oral dosing (Tmax = 0.4 h) and displayed good exposure with AUC = 846 ng·h/mL, resulting in an estimated Fpo of 100%. The derivative also displayed excellent brain penetration, as measured by samples at 0.5 h post-IV dose (brain/plasma = 3.2). Furthermore, measured levels in the CSF at the same time point suggested that the unbound fraction of 4k in the brain was 3.6%, reasonably consistent with the measured plasma protein binding in vitro of 92% (unbound plasma fraction = 8%).

Table 2

PK Parameters of Compound 4k in Rat

4k, 1 mg/kg(IV)		4k, 2 mg/kg(PO)
plasma clearance	42 mL/min/kg	Tmax	0.4 h
Vd(ss)	4.6 L/kg	Cmax	244 ng/mL
terminal t1/2	1.2 h	terminal t1/2	1.1 h
AUCinf	397 ng·h/mL	AUCinf	846 ng·h/mL
brain:plasma (0.5 h)	3.2	Fpo	100%
CSF:brain (0.5 h)	0.036

Given the good overall in vitro ADME and in vivo PK profile of 4k, especially with respect to brain penetration and oral bioavailability, the compound was tested for its ability to reverse haloperidol-induced catalepsy in rats, a simple and well validated in vivo pharmacodynamic model mimicking the loss of striatal dopamine receptor function observed in Parkinson’s disease.[1] Compound 4k was found to very potently reverse catalepsy induced by haloperidol, with ED50 values of 0.2 mg/kg at both 1 and 2 h post dose time points (Figure 2).

Figure 2

In vivo efficacy of 4k. Dose-dependent effect of 4k (0.1–1 mg/kg, po; 1 and 2 h pretreatment time) to reverse haloperidol-induced catalepsy in rats in comparison with the positive control, istradefylline (1 mg/kg, po).

Conclusions

The studies presented here have shown for the first time that biophysical mapping and cocrystal X-ray structures of ligands to a G protein-coupled receptor can be used to direct optimization of novel, low molecular weight hit molecules into highly potent and selective lead compounds. Compound 4k, described above, has desirable physicochemical and drug-like properties, including high oral bioavailability and very potent in vivo efficacy. Further optimization of this 1,2,4-triazine series of antagonists of the adenosine A2A receptor has subsequently allowed identification of a preclinical candidate for the potential treatment of Parkinson’s disease, and details of the development of this molecule will be the topic of future publications.

Experimental Section

Synthetic Methods

The purity of the final compounds was determined by HPLC or LC/MS analysis to be >95%. Full experimental details of all compounds in Table 1 are described in the Supporting Information. Synthesis of compounds 4g and 4k are described below.

6-Bromo-5-phenyl-1,2,4-triazin-3-amine

A solution of 5-phenyl-1,2,4-triazin-3-amine (1.50 g, 8.70 mmol) in DMF (15 mL) was cooled to −25 °C and treated with a solution of N-bromosuccinimide (4.50 g, 26.6 mmol) in DMF (10 mL) by dropwise addition. The reaction was warmed gradually to room temperature and stirred overnight with TLC monitoring. After completion of the reaction, the mixture was poured into saturated bicarbonate solution (50 mL) and extracted with diethyl ether (25 × 3 mL). The organic phases were combined, dried over Na2SO4, and concentrated in vacuo. The crude compound was purified by gradient flash chromatography, eluting with mixtures of ethyl acetate in hexane to afford 6-bromo-5-phenyl-1,2,4-triazin-3-amine (1.40 g, 64%). HPLC: 99%, 8.31 min (244 nm). Mass spectroscopy: m/z 250.9 [M + H]+. 1H NMR: (400 MHz, DMSO) δ: 7.49–7.57 (m, 5H), 7.72 (m, 2H).

2-(Trifluoromethyl)-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-6-methylpyridine

Methoxy(cyclooctadiene)iridium(I) dimer (30 mg, 0.062 mmol), 4,4′-di-tert-butyl-2,2′-bipyridine (33 mg, 0.124 mmol), and bis(pinacolato)diboron (4.09 g, 16.1 mmol) were added to a flask which had been thoroughly purged with nitrogen. The flask was once more purged before adding hexane via syringe (30 mL). The resulting mixture was heated at 50 °C for 10 min until the appearance of a dark-red solution was observed. 2-Trifluoromethyl-6-methyl pyridine (4.0 g, 24.8 mmol) was then added by syringe, and heating continued for a further 6 h. After cooling to room temperature, the crude reaction mixture was concentrated under reduced pressure. The resulting residue was purified by column chromatography, eluting with ethyl acetate/hexane mixtures to afford the target compound 2-(trifluoromethyl)-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-6-methylpyridine (5.9 g, 83%). HPLC: 96%, 7.57 min (210 nm). Mass spectroscopy: m/z 287.8 [M + H]+. 1H NMR: (400 MHz, DMSO) δ: 1.31 (s, 12 H), 2.51 (s, 3H), 7.70 (s, 1H), 7.76 (s, 1H).

6-(2,6-Dimethylpyridin-4-yl)-5-phenyl-1,2,4-triazin-3-amine 4g: A Typical Procedure for the Synthesis of 5,6-Biaryl-1,2,4-triazine-3-amine derivatives

A solution of 6-bromo-5-phenyl-1,2,4-triazin-3-amine, (90 mg, 0.358 mmol) in dioxane (2.0 mL) was treated with 2,6-dimethyl-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)pyridine (96 mg, 0.412 mmol) and K2CO3 (148 mg, 1.07 mmol). The resulting mixture was diluted with water (1.0 mL), degassed, treated with tetrakis triphenylphosphinepalladium(0) (21 mg, 0.018 mmol), and stirred for 2 h at 150 °C in a sealed vessel. Upon completion of the reaction, the mixture was diluted with water (20 mL) and extracted with ethyl acetate (3 × 20 mL); the combined organic extracts were then dried over Na2SO4 and concentrated under reduced pressure. The crude compound was purified by gradient flash chromatography, eluting with mixtures of ethyl acetate and hexanes to afford 6-(2,6-dimethylpyridin-4-yl)-5-phenyl-1,2,4-triazin-3-amine 4g (44 mg, 43%). HPLC: 98%, 6.09 min (281 nm). Mass spectroscopy: m/z 278.1 [M + H]+. 1H NMR: (400 MHz, DMSO) δ: 2.33 (s, 6H), 6.97 (s, 2H), 7.37–7.43 (m, 4H), 7.48 (m, 1H), 7.58 (bs, 2H).

6-[2-Methyl-6-(trifluoromethyl)pyridin-4-yl]-5-phenyl-1,2,4-triazin-3-amine 4k

6-[2-Methyl-6-(trifluoromethyl)pyridin-4-yl]-5-phenyl-1,2,4-triazin-3-amine 4k (0.32 g, 35%) was prepared from 6-bromo-5-phenyl-1,2,4-triazin-3-amine (0.70 g, 2.78 mmol) and 2-methyl-4-(4,4,5,5-tetramethyl-1,3,2-dioxaborolan-2-yl)-6-trifluoromethyl-pyridine (1.2 g, 4.1 mmol) according to the typical procedure described above. HPLC purity: 99%, 10.29 min (269 nm). Mass spectroscopy: m/z 332.0 [M + H]+. 1H NMR: (400 MHz, DMSO) δ: 2.48 (s, 3H), 7.38 (m, 5H), 7.47 (s, 1H), 7.58 (s, 1H), 7.73 (bs, 2H).

Biology Methods

Methods for determination of antagonist potency against human adenosine A2A and A1 receptors, binding constants, and receptor kinetics of compounds binding to the A2A–StaR by surface plasmon resonance and the procedure for determination of in vivo efficacy in rodents by reversal of haloperidol induced catalepsy are detailed in the Supporting Information.

Diffraction Data Collection of A2A–StaR2 in Complex with 4e and 4g

Diffraction data from crystals of A2A–StaR2 in complex with compounds 4e and 4g were collected at I24, Diamond Light Source, Oxford, UK. Statistics for data collection and refinement are given in Supporting Information Table S1. Atomic coordinates and structure factors have been deposited in the RCSB under accession codes 3UZC and 3UZA, respectively.

Computational Chemistry

Homology models were constructed from the avian β1 adrenergic GPCR crystal structure bound to cyanopindolol (PDB: 2VT4) using several computational approaches as detailed in the preceding paper and refined/validated using site-directed mutagenesis and BPM data and known ligands (see Supporting Information for full details).[3,4] Docking was done using Glide SP and XP (Schrödinger), and GRID analyses of the binding sites was used to evaluate potential docking poses (using the Csp3 (C3) for shape, aromatic CH probe (C1=) for lipophilic hotspots, carbonyl group (O) for hydrogen-bond acceptor hotspots, and amide NH (N1) for hydrogen-bond donor hotspots) and driving the designs.[8,9]