Discovery of Pyridone-Substituted Triazolopyrimidine Dual A2A/A1 AR Antagonists for the Treatment of Ischemic Stroke.

Ischemic stroke is a complex systemic disease characterized by high morbidity, disability, and mortality. The activation of the presynaptic adenosine A2A and A1 receptors modifies a variety of brain insults from excitotoxicity to stroke. Therefore, the discovery of dual A2A/A1 adenosine receptor (AR)-targeting therapeutic compounds could be a strategy for the treatment of ischemic stroke. Inspired by two clinical phase III drugs, ASP-5854 (dual A2A/A1 AR antagonist) and preladenant (selective A2A AR antagonist), and using the hybrid medicinal strategy, we characterized novel pyridone-substituted triazolopyrimidine scaffolds as dual A2A/A1 AR antagonists. Among them, compound 1a exerted excellent A2A/A1 AR binding affinity (K i = 5.58/24.2 nM), an antagonistic effect (IC50 = 5.72/25.9 nM), and good metabolic stability in human liver microsomes, rat liver microsomes, and dog liver microsomes. Importantly, compound 1a demonstrated a dose-effect relationship in the oxygen-glucose deprivation/reperfusion (OGD/R)-treated HT22 cell model. These findings support the development of dual A2A/A1 AR antagonists as a potential treatment for ischemic stroke.

Adenosine, the naturally occurring purine nucleoside, acts as an endogenous modulator in both the central and peripheral nervous systems by interacting with four subtypes of specific G-protein-coupled receptors (GPCRs): A1, A2A, A2B, and A3 adenosine receptors (ARs). The A1 and A3 ARs are negatively coupled to adenylyl cyclase and exert an inhibitory effect on cyclic adenosine monophosphate (cAMP) production by recruiting the Gi protein, whereas the A2A and A2B ARs promote adenylyl cyclase activation and subsequent cAMP production by recruiting the Gs protein.[1−4] The A2A and A1 ARs are highly enriched in specific parts of the central nervous system (CNS) and are associated with motor activity, psychiatric behaviors, and neuronal cell death. Many potent selective A2A AR antagonists have been designed as promising candidates for their beneficial effects on Parkinson’s disease (PD),[5] ischemia,[6] epilepsy,[7] Huntington’s disease (HD),[8] and Alzheimer’s disease (AD).[9]

Over the past three decades, the search for novel A2A AR antagonists has been greatly expanded, and a large number of drug candidates have entered clinical trials (Figure ). Unfortunately, only istradefylline (KW-6002) has been licensed as an antiparkinsonian drug in Japan (2013) and the United States (2019);[10] other A2A AR antagonists (preladenant,[11] vipadenant,[12] tozadenant,[13] etc.) were terminated because of a lack of efficacy in vivo or toxicity.[14] Recently, the A2A AR emerged as a novel immune checkpoint for the development of a cancer immunotherapy drug in combination with PD-1/PD-L1 or anti-CTLA-4 monoclonal antibody (mAb).[15,16] ZM241385 significantly inhibited tumor growth in a lung metastasis model and induced a remarkable delay in tumor growth in melanoma-bearing mice when combined with anti-CTLA-4 mAb.[17,18] Moreover, the discovery and development of dual A AR therapeutic compounds is an attractive and alternative therapeutic strategy for improving the in vivo efficacy of a single target. For example, AB928, a potent and selective dual A2A/A2B AR antagonist discovered by Arcus Biosciences, is currently undergoing clinical trials in multiple cancer settings.[19,20]

Figure 1

Reported selective A2A and dual A2A/A1 AR antagonists.

Ischemic stroke is a complex systemic disease characterized by high morbidity, disability, and mortality.[21] Increasing substantial evidence has shown a protective role for A2A AR antagonists in striatal and nigral neurons through the prevention of glutamate-dependent neuronal death, thereby reducing cortical damage in a variety of ischemic stroke models.[22−24] In A2A AR knockout (KO) mice, transient focal ischemia causes less neuronal damage compared with that in wild-type (WT) mice. The selective A2A AR antagonist SCH58261 reduced ischemic brain damage in an adult rat model of focal cerebral ischemia.[25−27] Meanwhile, the activation of the A1 AR was able to induce ischemic damage protection and the reduction of both reactive and proliferative microglia/macrophages after experimental stroke in rats.[28,29] These results demonstrate that the activation of the presynaptic A2A and A1 ARs modifies a variety of brain insults from excitotoxicity to stroke.

Owing to our interest in ARs and the field of ischemic stroke, we set out to design and synthesize novel dual A2A/A1 AR antagonists based on the crystal structures of A2A AR (PDB: 3EML) and A1 AR (PDB: 5EUN) complexes.[30,31] We herein report the discovery and characterization of a new chemotype of dual A2A/A1 AR antagonists with a pyridone-substituted triazolopyrimidine scaffold, in which compound 1a demonstrated a remarkable dose–effect relationship in the oxygen-glucose deprivation/reperfusion (OGD/R)-treated HT22 cell model.

Our initial design was inspired by two known clinical phase III drug candidates, preladenant (selective A2A AR antagonist), with a triazolopyrimidine scaffold,[11] and ASP-5854 (dual A2A/A1 AR antagonist), with a pyrazine scaffold[32] (Figure ). Using insight from preladenant cocrystal structures with an A2A AR and an A1 AR, we noticed that the primary amide (ring A) and the triazole (ring B) with furan rings in preladenant established two bidentate hydrogen-bonding interactions with Glu169 and Asn253 (Figure A,B) with similar binding modes. However, preladenant is just a selective A2A AR antagonist, suggesting that the triazolopyrimidine scaffold was a key pharmacophore as the selective A2A AR antagonist. The primary amide in ASP-5854 also formed a bidentate hydrogen-bonding interaction with Asn253, in which the A2A AR was the key pharmacophore. Moreover, the pyridone in ASP-5854 formed an additional hydrogen bond with His278 in the A2A AR. By contrast, when docked with A1 AR, the pyridone in ASP-5854 formed a hydrogen bond with Thr90 in the A1 AR, which may be a key pharmacophore as the A1 AR antagonist. On the basis of these analyses, we intended to exploit the hybrid drug design approach to access all five aforementioned interactions with the goal of identifying a novel chemical scaffold dual A2A/A1 AR antagonist with better drug-like properties. Therefore, we designed and synthesized a series of compounds with the novel pyridone-substituted triazolopyrimidine chemotype and carried out a systematic study of structure–activity relationships (SARs).

Figure 2

Design strategy of novel chemotype dual A2A/A1 AR antagonists. (A) Interactions shown between preladenant with A2A AR (A1) and A1 AR (A2). (B) Interactions shown between ASP-5854 with A2A AR (B1) and A1 AR (B2). (C) Hybridization strategy of novel pyridone-substituted triazolopyrimidine scaffold.

As shown in Scheme , the synthetic strategy of 1a–1i involved a three-step sequence, including a nucleophilic substitution reaction, Dimroth rearrangement, and a Suzuki coupling process, starting from commercially available aryl formamide derivatives 2. Initially, the nucleophilic substitution of 5-bromo-4-chloropyrimidin-2-amine (3) by aryl formamide 2a–2i at 120 °C in n-butanol proceeded smoothly to deliver compounds 4a–4i. The subsequent Dimroth rearrangement of compounds 4a–4i was conducted in the presence of N,O-bis(trimethylsilyl) acetamide (BSA) and hexamethyldisilazane (HMDS) at 120 °C to give the desired cyclization triazolopyrimidine compounds 5a–5i. Finally, compounds 5a–5i were coupled to pyridone boronic esters (6) to afford the final products 1a–1i in an acceptable yield (4.1–6.5%) over three steps.

Scheme 1

Reagents and conditions: (a) n-BuOH, 120 °C, 8 h; (b) HMDS, BSA, 120 °C, 8 h; (c) Pd(dppf)Cl2, K2CO3, dioxane, H2O, 90 °C, 8 h, 4.1–6.5%, three steps.

The binding affinity of the synthesized pyridone-substituted triazolopyrimidine derivatives (1a–1i) toward the A2A and A1 ARs, along with A2B and A3 ARs, was evaluated in competitive binding experiments using membrane preparation of the human recombinants A1, A2A and A3, and the A2B AR overexpressed from CHO, HeLa, and HEK-293 cells, respectively. [3H] DPCPX (A1), [3H]ZM241385 (A2A), [3H]DPCPX (A2B), and [3H]NECA (A3) were used as radioligands.[33] The binding affinity data of synthesized compounds are listed in Table , with pyrazine antagonist compound 7 chosen as the reference. Among them, compound 1a with a furan ring as the Ar group exhibited the most excellent binding affinity with a K value of 5.58 nM against the A2A AR and 24.2 nM against the A1 AR and a high degree of selectivity for the A2B AR (A2B/A2A 88-fold) and the A3 AR (A3/A2A 1575-fold), respectively. Compounds 1b (pyridine as the Ar group) and 1c (thiazole as the Ar group) showed moderate binding affinity against the A2A AR (K = 62.4 nM) and comparable binding activity against the A1 AR (K = 84.6 nM), along with good selectivity over the A2B and A3 ARs. Interestingly, compound 1d (5-methylthiazole as the Ar group) displayed the most potent binding affinity data, with a K value of 21.9 nM against the A1 AR and two-fold selectivity over the A2A AR, which can be used as a lead for the further optimization of selective A1 AR antagonists. However, compounds 1e–1i showed less binding affinity against A1 to A3 ARs compared with compounds 1a–1d because of the introduction of a methyl group beside the heteroatom, which may increase the steric hindrance and affect the binding to the target cavity of A1 to A3 ARs. These results suggested that the introduction of heteroatoms in the Ar group was crucial for binding to the A2A AR and increased its affinity, and furan as the Ar group was the most potent. Conversely, the introduction of a methyl group beside the heteroatom on the Ar group was fatal for binding to the target due to the steric hindrance.

Table 1

Binding Affinity (K, nM) of Compounds 1a–1i at the Adenosine Receptors

	binding affinity (Ki, nM)
compound	A1 AR	A2A AR	A2B AR	A3 AR
1a	24.2 ± 5.49	5.58 ± 0.11	491 ± 4.45	8790 ± 3.27
1b	37.6 ± 4.34	62.4 ± 3.43	>10000	>10000
1c	78.0 ± 6.75	84.6 ± 4.73	1586 ± 4.14	>10000
1d	21.9 ± 4.53	43.9 ± 2.75	9919 ± 6.56	2934 ± 4.11
1e	88.4 ± 6.36	448 ± 4.42	>10000	2964 ± 4.55
1f	926 ± 3.70	1032 ± 2.46	>10000	>10000
1g	722 ± 9.90	3427 ± 4.20	>10000	>10000
1h	1719 ± 6.12	3767 ± 2.41	>10000	>10000
1i	2973 ± 4.65	797 ± 2.44	>10000	>10000
7	88.6 ± 5.56	26.9 ± 2.71	58.9 ± 3.33	>10000

Furthermore, the calcium flux functional experiments were carried out to assess the antagonistic/agonistic activity of the most potent compound 1a at the A2A and A1 ARs, along with A2B and A3 ARs.[34] The functional assay data IC50 for 1a shown in Figure indicated that the excellent antagonist activity of 1a was consistent with its binding affinity, whereas the agonist activity of 1a was negligible.

Figure 3

Antagonistic effect (IC50, nM) of compound 1a against A1, A2A, A2B, and A3 ARs.

Many studies have proven that neuron apoptosis is involved in the pathological process of ischemia injury. Thus the OGD/R model (in vitro ischemic model) was used to damage HT22 cells to simulate ischemic injury to investigate the effect of A2A/A1 AR antagonist compound 1a on HT22 cell damage.[35] First, we investigated the effect of compound 1a on cell apoptosis induced by OGD/R. The results demonstrated that OGD/R significantly induced apoptosis of HT22 cells, and compound 1a reversed, in a concentration-dependent manner, the up-regulation of pro-apoptotic genes such as cleaved caspase-3, cleaved caspase-9, cleaved PARP1, p53, and Bax in OGD/R-treated HT22 cells (Figure A). Notably, immunofluorescence analyses revealed that the antiapoptotic gene Bcl-2 staining was enhanced by compound 1a (Figure B). Likewise, the mRNA expression in the apoptosis markers (p53, Bax, Bcl-2) was consistent with the protein expression (Figure C). These results provided support that compound 1a protected against cell apoptosis.

Figure 4

Compound 1a prevented HT22 cell apoptosis after oxygen-glucose deprivation (OGD). (A) HT22 cells were pretreated with or without compound 1a (1–10 μM) for 4 h and then were treated with OG, then 4 and 24 h of reoxygenation. Cell lysates were prepared and blotted with antibodies to Bax, p53, cleaved PARP1, cleaved caspase-3, and cleaved caspase-9; ***p < 0.001, all data are presented as the mean ± SD of three independent experiments. (B) HT22 cells were pretreated with or without 10 μM compound 1a for 4 h and then were treated with OGD, then 4 and 24 h of reoxygenation. Cells were immunostained with antibody Bcl-2 (magnification, 200×). (C) HT22 cells were pretreated with or without 10 μM compound 1a for 4 h and then were treated with OGD, then 4 and 24 h of reoxygenation. mRNA levels of Bax, p53, and Bcl2 were quantified using qRT-PCR. *p < 0.05, ***p < 0.001. All data are presented as the mean ± SD of three independent experiments.

Currently, there is strong evidence that inflammatory processes may contribute to secondary brain damage after ischemic stroke. Indeed, inflammation modulators including iNOS, COX-2, and VCAM-1 were induced by OGD/R; on the contrary, increased inflammation mediators were significantly inhibited by compound 1a treatment (Figure A). Consistent with these findings, immunofluorescence staining revealed a lack of COX-2 in compound-1a-treated HT22 cells as compared with that in OGD/R-treated HT22 cells (Figure B). Recent findings identified that NLRP3 inflammasomes play a major role in neuronal cell death in stroke and further suggested that targeted inflammasome assembly and activity may ameliorate ischemic injury.[36,37] We found that OGD/R robustly induced the expression of inflammasome protein caspase-1 (p20) and mature pro-inflammatory cytokines IL-18 and IL-1β in HT22 cells; in turn, compound 1a reduced caspase-1 (p20), IL-18, and IL-1β expression (Figure C). Likewise, the mRNA expression in inflammation markers Nos2, Vcam-1, and II1b was consistent with the protein expression (Figure D). These results provide further support that compound 1a protected against neuron OGD/R injury.

Figure 5

Compound 1a reduced the inflammation response in OGD/R-treated HT22 cells. (A) HT22 cells were pretreated with or without compound 1a (1–10 μM) for 4 h and then were treated with OGD, then 4 and 24 h of reoxygenation. Cell lysates were prepared and blotted with antibodies to iNOS, VCAM-1, and COX-2. ***p < 0.001. All data are presented as the mean ± SD of three independent experiments. (B) HT22 cells were pretreated with or without 10 μM compound 1a for 4 h and then were treated with OGD, then 4 and 24 h of reoxygenation. Cells were immunostained with antibody COX-2 (magnification, 200×). (C) HT22 cells were pretreated with or without compound 1a (1–10 μM) for 4 h and then were treated with OGD, then 4 and 24 h of reoxygenation. Cell lysates were prepared and blotted with antibodies to IL-1β, IL-18, and caspase-1 (p20). ***p < 0.001. All data are presented as the mean ± SD of three independent experiments. (D) HT22 cells were pretreated with or without 10 μM compound 1a for 4 h and then were treated with OGD, then 4 and 24 h of reoxygenation. mRNA levels of Nos2, Vcam-1, and II1b were quantified using qRT-PCR. *p < 0.05, ***p < 0.001. All data are presented as the mean ± SD of three independent experiments.

The metabolic stability is a prime consideration when developing a candidate (Table ). The in vitro metabolic stability of compound 1a was measured using human liver microsomes (HLMs), rat liver microsomes (RLMs), mouse liver microsomes (MsLMs), dog liver microsomes (DLMs), and monkey liver microsomes (MkLMs). Compound 1a displayed good metabolic stability with a half life of 77.4, 56.0, and 83.9 min along with an intrinsic clearance (CL) of 16.1, 44.5, and 23.8 mL/min/kg in the HLMs, RLMs, and DLMs, respectively. After 60 min, 58.3, 48.6, and 60.6% of compound 1a remained in the HLMs, RLMs, and DLMs, respectively. However, compound 1a displayed less metabolic stability with a half life of 3.4 and 14.6 min along with an intrinsic clearance (CL) of 1601.7 and 128 mL/min/kg in the MsLMs and MkLMs, respectively. In addition, compound 1a showed moderate brain penetration (B/P ratio = 0.22) (Table S3), which is suitable for the lead compound of ischemic stroke.

Table 2

In Vitro Metabolic Stability of Compound 1a in the Presence of Different Microsomes

parameters	T1/2 (min)	CLint(mic) (μL/min/mg)	CLint(liver) (mL/min/kg)	unchanged (T = 1 h, %)
HLMsa	77.4	17.9	16.1	58.3
RLMsb	56.0	24.7	44.5	48.6
MsLMsc	3.4	404.5	1601.7	0.00
DLMsd	83.9	16.5	23.8	60.6
MkLMse	14.6	94.8	128.0	5.9

Human liver microsomes.

Rat liver microsomes.

Mouse liver microsomes.

Dog liver microsomes.

Monkey liver microsomes.

Molecular docking modeling was performed to interpret the dual A2A/A1 AR binding affinity of compound 1a at the molecular level. The binding modes of compound 1a at the A2A and A1 AR cavities were analyzed by docking simulations using the Autodock software package, with the crystal structures of the A2A and A1 AR complexes as templates, respectively. The docking results (Figure ) revealed that compound 1a adopted the general binding mode at both the A2A and A1 AR binding sites. In this binding mode, the pyridone-substituted triazolopyrimidine scaffold was positioned in the depth of the binding pocket and underwent a p–p interaction with the Phe residue (Phe168 in the A2A AR, Phe171 in the A1 AR). In addition, compound 1a formed five hydrogen bonds with the A2A AR (Glu169, Asn253, and His278), whereas it formed only two hydrogen bonds with the A1 AR (Asn253). Therefore, compound 1a adopted a much more favorable binding pose at the A2A AR cavity than at the A1 AR cavity. Moreover, the Autodock docking results also indicated that the binding pose was associated with a better docking score at the A2A AR (−9.02 kcal/mol, K = 0.24 μM) than at the A1 AR (−7.79 kcal/mol, K = 1.94 μM). Hence, the molecular docking results explained the A2A AR affinity and the slight selectivity over the A1 AR (4.4-fold) of compound 1a.

Figure 6

Schematic description of the ligand–target interaction between compound 1a and (A) the A2A AR and (B) the A1 AR.

In the present study, we designed and synthesized a novel pyridone-substituted triazolopyrimidine scaffold dual A2A/A1 AR antagonist using a computer-aided rational drug design approach along with a hybrid medicinal strategy, inspired by two phase III drugs, ASP-5854 (dual A2A/A1 AR antagonist) and preladenant (selective A2A AR atagonist). Invitro evaluations of the A1, A2A, A2B, and A3 AR binding assays for the synthesized compounds showed promising results. Among them, compared with ASP-5854, the most potent compound 1a showed a better clog P and comparable A2A/A1 AR binding affinity (K = 5.58/24.15 nM). In addition, compound 1a showed excellent solubility at pH 1.2 (1.99 mg/mL), which is suitable for oral administration after some rational modification. Moreover, compound 1a showed an excellent antagonistic effect (IC50 = 5.72/25.93 nM), moderate brain penetration (B/P ratio = 0.22), and good metabolic stability with T1/2HLMs = 77.4 min, T1/2RLMs = 56 min and T1/2DLMs = 83.9 min. Importantly, compound 1a demonstrated a remarkable dose–effect relationship in the OGD/R-treated HT22 cell model, including the reduction of HT22 cells apoptosis and an alleviation of inflammatory modulator (iNOS, COX-2, and VCAM-1) and inflammatory cytokine (p20, IL-18, IL-1β, Nos2, Vcam-1, and II1b) release. With these encouraging results, we anticipate that this novel pyridone-substituted triazolopyrimidine scaffold could be an excellent starting point for the further development of dual A2A/A1 AR antagonists to benefit the field of ischemic stroke. The current effort is focused on further improving the potency along with good pharmacokinetics and pharmacodynamics both invivo and invitro, and these findings will be reported in due course.