Design and in vivo Characterization of A₁ Adenosine Receptor Agonists in the Native Ribose and Conformationally-Constrained (N)-Methanocarba Series

Abstract

(N)-Methanocarba ([3.1.0]bicyclohexyl) adenosines and corresponding ribosides were synthesized to identify novel A₁ adenosine receptor (A₁AR) agonists for CNS or peripheral applications. Human and mouse AR binding was determined to assess the constrained ring system’s A₁AR compatibility. N⁶-Dicyclobutylmethyl ribose agonist (9, MRS7469, >2000-fold selective for A₁AR) and known truncated N⁶-dicyclopropylmethyl methanocarba 7 (MRS5474) were drug-like. The pure diastereoisomer of known riboside 4 displayed high hA₁AR selectivity. Methanocarba modification reduced A₁AR selectivity of N⁶-dicyclopropylmethyl and endo-norbornyl-adenosines, but increased ribavirin selectivity. Most analogues tested (ip.) were inactive or weak in inducing mouse hypothermia, despite mA₁AR full agonism and variable mA₃AR efficacy, but strong hypothermia by 9 depended on A₁AR, which reflects CNS activity (determined using A₁AR or A₃AR null mice). Conserved hA₁AR interactions were preserved in modeling of 9 and methanocarba equivalent 24 (~400-fold A₁AR-selective). Thus, we identified, and characterized in vivo, ribose and methanocarba nucleosides, including with A₁AR-enhancing N⁶-dicyclobutylmethyl-adenine and 1,2,4-triazole-3-carboxamide (40, MRS7451) nucleobases.

Graphical Abstract

Introduction

The A₁ adenosine receptor (A₁AR) is coupled to the inhibition of adenylyl cyclase through the G_i protein and also acts on ion channels and MAP kinases.^1,2 A₁AR agonists are desired for their antiarrhythmic, antilipolytic, antinociceptive, cerebroprotective, cardioprotective, antidepressant, sleep-enhancing and antiseizure properties.^1–7 Beneficial peripheral effects of A₁AR agonists include protection in acute and chronic pain^8,9 or as antilipolysis.¹⁰ We recently described the induction of mouse hypothermia through brain A₁AR activation that is independent of peripheral A₃AR activation.³⁰ Pharmacologically-induced hypothermia has potential clinical applications, such as cerebroprotection in brain ischemia, but most A₁AR agonists are unsuitable as clinical candidates because of their cardiovascular effects.^1,4,11–13 Thus, there is a need for novel A₁AR agonists that are well-tolerated for a variety of clinical applications, but with diminished side effects.

The first A₁AR X-ray crystallographic structures, in complex with antagonists but not agonists, were reported and show promise to enable structure-based drug design (SBDD).^14,15 More recently, a high resolution structure of an agonist-bound A₁AR containing a G_i protein was determined using cryo-electron microscopy (cryo-EM).¹⁶ However, most potent A₁AR agonists so far have been discovered through extensive empirical probing of their structure activity relationship (SAR), for example, leading to nucleosides 1–7 (Chart 1).^1,2 Thus, adenosine derivatives variously substituted at the N⁶, 5′ and C2 positions have been reported as potent and selective A₁AR agonists. Widely used modifications of adenosine to increase A₁AR selectivity consist of introducing an N⁶-phenyl or an N⁶-cycloalkyl group, such as N⁶-cyclopentyl (1, 2, 5 and 6), N⁶-cyclohexyl or N⁶-endo-norbornyl (3, 4).^8,17–19 (2S)-N⁶-(endo-norbornyl)-adenosine 3 is among the most selective mono-substituted agonists of A₁AR.¹⁵ Recently, additional N⁶-bicycloalkyl derivatives containing heteroatoms were reported to be highly A₁AR-selective agonists.²⁰ Other commonly used modifications in AR agonists, such as 5′-Cl (4) and 5′-N-alkylamides (6), but not extended or bulky C2 substituents, are compatible with N⁶-cycloalkyl groups for retention of A₁AR affinity.^21–23 Alternative 5′ substituents, such as aryl thioethers, e.g. 5, and other ethers, were used to provide partial agonist activity at the A₁AR, especially as applied to antiarrhythmic agents.^1,24 The SAR of 5′-deoxyadenosine 5′-oxadiazole or 5′-tetrazole derivatives led to A₁AR-selective agonists.^1,18,19,25 However, the C8 position of adenosine was amenable to only limited derivatization with the retention of AR binding affinity, leading to partial A₁AR agonists and A₃AR antagonists.^26,27

Chart 1.

Structures of reference AR agonists. All nucleosides shown are A₁AR-selective,^1,2 except 6, which is a mixed A₁AR/A₃AR agonist.³²

Compound 4 is often used as a highly A₁AR-selective agonist with a 12,000-fold window of selectivity in mouse (m) A₁AR vs. mA₃AR binding, which was shown to induce hypothermia via the A₁AR.^28–30 When administered peripherally in the mouse, it passes the blood-brain barrier (BBB) sufficiently to activate a central mA₁AR, which lowered core body temperature (Tb). Furthermore, at intraperitoneal (ip.) doses up to 3 mg/kg, 4 did not cause intense activation of the peripheral mast cell mA₃AR, as seen with other A₁AR agonists. In addition to its high A₁AR binding selectivity, this in vivo selectivity of 4 might be partly due to a possible lower A₃AR efficacy, typical of adenosine derivatives with large N⁶ groups.¹⁷ Other A₁AR agonists, such as 1a, although 2400-fold selective in mA₁AR binding, did not achieve in vivo agonist selectivity due to activation of a peripheral mA₃AR, consistent with its full efficacy at the human (h) A₃AR.¹⁷ The greater propensity of AR agonists to activate mA₃AR rather than brain A₁AR after peripheral administration relates to the low BBB penetration of adenosine derivatives. This characteristic is partially a function of the multiple hydroxyl groups that reduce diffusion across biological membranes.³¹

The North (N)-methanocarba ([3.1.0]bicyclohexyl) modification increases the potency and selectivity of adenosine derivatives at the A₃AR and tends to maintain affinity at the A₁AR, but not A_2AAR.³² We previously introduced a moderately selective hA₁AR full agonist 7 that has a N⁶-dicyclopropylmethyl group, (N)-methanocarba substitution of ribose and only two hydroxyl groups due to 4′-truncation.³³ This compound has demonstrated antiseizure and antidepressant activity.^6,33 Here, we enlarged the pharmacological characterization of 7 and demonstrated its favorable absorption, distribution, metabolism, excretion and toxicological (ADME-tox) properties, suggesting further derivatization to expand the range of useful A₁AR agonists.

Thus, we introduced novel ribose and (N)-methanocarba adenosine derivatives with the goal of enhancing their A₁AR selectivity. One objective was to extend the SAR of (N)-methanocarba adenosine derivatives in the direction of A₁AR in two species (h and m). We examined if it is feasible to overcome the tendency of this rigid ring system to produce A₃AR selectivity, by appropriately substituting other positions. Previously, A₁AR-enhancing N⁶-cycloalkyl groups were combined with the A₃AR-enhancing (N)-methanocarba ring system to produce an agonist 6 of mixed A₁AR/A₃AR selectivity.³² Further modification could tune the selectivity, e.g. as seen in certain adenosine derivatives where a 2-chloro substitution increased hA₁AR selectivity³² or affinity.³⁴

Results

We have modified the 5′, C2 and other positions, and compared the receptor binding of some (N)-methanocarba nucleoside analogues directly with the corresponding ribosides. We have modeled the interaction of several novel A₁AR-selective N⁶-dicyclobutylmethyl agonists at the hA₁AR. We examined many of these A₁AR agonists in an in vivo assay of hypothermia in the mouse, which can delineate peripheral A₃AR and central A₁AR components.³⁰

Selection of target nucleosides

The nucleoside analogues examined in AR binding assays are shown in Table 1. Based on known AR agonists (e.g. 1–8, 14–18, 21, 23 and 38), related adenine substitutions were made in novel nucleosides in both the ribose (e.g. 9–13) and (N)-methanocarba (e.g. 22, 24–31 and 32–37, 39, 40) series. Adenosine derivatives with N⁶-cycloalkyl (such as N⁶-endo-norbornyl) and symmetric N⁶-dicycloalkylmethyl (N⁶-dicyclopropylmethyl and larger) groups were included as a potential means of enhancing selectivity for the A₁AR. N⁶-endo-Norbornyl derivative 13 is a pure diastereoisomer of 4, corresponding to the predicted preferred stereochemistry for A₁AR recognition from an earlier rA₁AR study.²⁸ The inclusion of N⁶-dicyclopropylmethyl analogues was based on the moderate A₁AR selectivity of 7 and 8.^17,33,34 In addition to known N⁶-dicyclopropylmethyl and N⁶-dicyclopentylmethyl substituents,^33,34 here we introduced the first N⁶-(dicyclobutylmethyl)-adenosine analogues (9–12, 19, 24–28 and 35) to be evaluated at ARs. We also compared 2-Cl and 2-H on the adenine moiety.

Table 1.

Structures and binding affinities^a of AR ligands in the species indicated, including reference compounds 1 - 7 and 15 - 18. R³ = OH, R⁴ = NHCH₃, Y = N, unless noted.

Compd. (other changes)	R1	R2	A1AR % inhibition or Ki (nM)a	A2AAR % inhibitiona	A3AR % inhibition or Ki (nM)a	Off-target, receptor (h, unless noted) (Ki, μM)a
Ribosides, 5′-OH or 5′-Cl
1ac,d		H	2.3 (h), 0.22±0.01 (m)	794 (h), 808±89 (m)	72±12 (h), 534±14 (m)	ND
2c,d		Cl	0.83 (h), 0.21±0.10 (m)	2270 (h), 988 (m)	38±6 (h), 17±5 (m)	ND
5f (R3 = S-(2-F)-Ph)		H	12 (h)	>10,000 (h)	>1000 (h)	ND
8b		H	0.85±0.27 (h), 0.8 (r)	1470±380 (h), 1370 (r)	41.3±5.3 (h)	none
9		H	2.14±0.52 (h), 0.37±0.02 (m)	3550±440 (h)	10,600±1400 (h), 897±45 (m)	DAT 1.78, NET 4.82
10 (R3 = Cl)		H	4.90±0.87 (h), 0.71±0.12 (m)	5200i (h)	16,600±2000 (h), 2110±50 (m)	TSPO 3.98, σ1 0.69, σ2 0.662
11 (R3 = SCH2CH3)		H	63.4±12.1 (h), 3.58±0.01 (m)	30±5% (h)	6220±1480 (h), 718±150 (m)	TSPO (90% inhib.j) σ1 0.626, σ2 0.642
12		Cl	17.8±8.7 (h), 0.65±0.06 (m)	2550±540 (h)	13,200±2500 (h), 653±85 (m)	5HT2C 1.45
3c,d		H	0.38±0.19 (h), 0.34 (r), 0.14±0.02 (m)	>10,000 (h), 477 (r)	915±299 (h), 282 (r), 424±41 (m)	ND
4c (R3 = Cl)	(±)	H	0.51 (h), 0.20±0.01 (m)	1340 (h), 3990±360 (m)	1290 (h), 2410±330 (m)	ND
13c (R3 = Cl)		H	0.76±0.42 (h)	2050±570 (h)	355±117 (h), 1560±140 (m)	σ1 0.484, σ2 0.432, 5HT7 1.24
Ribosides, 4′-carbonyl derivatives
14 (R4 = OH)	H	H	24±4% (h)	11±4% (h)	32±9% (h)	BZP (r) 7.34, σ1 1.48, σ2 0.796
15d (R3 = NH-Me)	H	H	36.7±9.4 (h), 84 (r)	466±95 (h), 67 (r)	24.4±7.9 (h), 63 (r)	none
16c,d,h (R4 = NH-Et)	H	H	6.8±2.4 (h),g 3.00±0.10 (h), 0.45±0.13 (m), 63 (r)	2.2±0.6 (h),g 35.0±14.0 (h), 12 (r)	35±12 (h), 14.1±6.8 (m), 113 (r)	none
17d (R4 = NH-cPr)	H	H	1.90±0.60 (h), 6.4 (r)	50±10 (h), 13.4 (r)	180±50 (h), 1600 (r)	H1 7.88, σ1 4.12 (gp), σ2 4.35
18d,h (R4 = NH-(CH2)2OH)	H	H	12.8±3.1 (h)	505±30 (h)	9450±1760 (h)	ND
(N)-Methanocarba, 4′-truncated
7c		Cl	47.9±10.5 (h), 5.20±0.05 (m)	3950±410 (h), 34±9% (m)	470±15 (h), 1060±250 (m)	5HT2B 0.641, 5HT2C 1.85
19		Cl	961±639 (h)	46±6% (h)	822±449 (h), 385±52 (m)	5HT2B 2.90, 5HT2C 8.32, α2A 6.96, TSPO 2.38
20b		Cl	34±3% (h)	13±3% (h)	48±2% (h)	5HT2C 2.92, σ1 5.09, σ2 2.01
(N)-Methanocarba, 5′-OH or 5′-Cl
21b	H	Cl	105±63 (h), 273 (r)b	3420±270 (h), 1910 (r)b	353±54 (h)d	none
22 (R3= Cl)	H	Cl	7.61±0.73 (h)	1750±290 (h)	253±148 (h)	σ1 1.97, σ2 2.55
23b		Cl	39±17 (h), 0.71±0.06 (m)	2200 (h), 41±9% (m)	1600±210 (h), 1030±40 (m)	5HT2B 0.641, 5HT2C 1.85, DAT 4.75
24		H	22.7±3.0 (h), 0.53±0.19 (m)	34±1% (h)	51±6% (h), 918±21 (m)	BZP (r) 4.08, 5HT2B 0.214, 5HT2C 1.32, σ1 1.79, σ2 0.753
25		Cl	8.96±1.02 (h), 2.47±0.26 (m)	55±5% (h), 26±3% (m)	25±2% (h), 612±58 (m)	5HT2B 0.153, 5HT2C 0.238, M5 3.00, DAT 4.75 TSPO 2.93
26 (R3 = Cl)		H	32.7±12.4 (h), 1.05±0.19 (m)	38±3% (h)	49±4% (h), 934±18 (m)	5HT2B 2.01, α2A 6.80, σ1 1.50, σ2 1.17, TSPO, 2.02
27 (R3 = Cl)		Cl	120±23 (h), 11.2±0.8 (m)	17±24% (h), 20±3% (m)	3820±1820 (h), 1560±60 (m)	5HT2B 1.52, 5HT2C 1.75, H2 2.60, σ2 0.349, TSPO, 2.89
28 (R3 = S-(2-F)-Ph)		Cl	45±4% (h), 2490±280 (m)	22±5% (h), 20±9% (m)	9750±4030 (h), 7440±660 (m)	5HT2B 0.334, 5HT2C 1.46, σ2 0.583
29		Cl	23.6±5.2 (h), 1.05±0.03 (m)	4260i (h), 15±2% (m)	288±54 (h), 574±23 (m)	5HT2B 0.472, NET 5.74
30 (Y = CH)		Cl	240±19 (h)	29±3% (h)	145±74 (h)	none
31 (R3 = Cl)		Cl	44.8±1.3 (h), 1.86±0.10 (m)	54±8% (h), 15±1% (m)	456±201 (h), 503±13 (m)	none
(N)-Methanocarba, 5′-carbonyl
32b (R4 = O-Et)		Cl	360±74 (h)	1570±180 (h)	236±41 (h)	5HT2B 0.015, 5HT2C 0.054, TSPO 2.50
33 (R4 = O-Et, 2′,3′-C(CH3)2)		Cl	49±9% (h)	15±2% (h)	41±6% (h)	5HT5A 8.69, H2 6.38
34b (R4 = NH-Me)		Cl	110±14 (h)	4320±870 (h)	34±11 (h)	5HT2B 0.023, 5HT2C 0.749
35 (R4 = O-Et)		Cl	73.8±14.2 (h)	57±12% (h)	1160±300 (h)	5HT2B 0.097, 5HT2C 0.089, M3 5.90, σ2 2.02
6e (R4 = NH-Me)		Cl	18.3±6.3 (h), 0.68±0.02 (m)	3250±300 (h)	3.7±0.9 (h), 5.8±1.6 (r), 3.46±0.13 (m)	5HT2B 0.012, σ1 1.55
36 (R4 = NH-cPr)		Cl	1.22±0.05 (h), 0.45±0.03 (m)	520±119 (h), 2650±560 (m)	59.0±17.8 (h), 29.5±0.6 (m)	5HT2B 1.57
37 (R4 = NH-(CH2)2OH)		Cl	17.6±5.3 (h), 1.56±0.09 (m)	5400i (h), 23±1% (m)	127±29 (h), 447±43 (m)	5HT2B 0.718, KOR, 2.63
Ribavirin analogues
38g	-	-	7410±100 (h), 4430±460 (m)	23±6% (h)	16±5% (h), 0% (m)	ND
39	OCH3	-	6±4% (h), 2990±80 (m)	14±5% (h)	22±4% (h), 15±2% (m)	β3 1.42, H3 4.52, σ2 1.72
40	NH2	-	495±35 (h), 25.2±2.8 (m)	11±3% (h)	41±6% (h), 47±2% (m)	5HT2B 1.91, 5HT2C 2.35, DAT 6.61, σ2 2.09

^aBinding in membranes of CHO or HEK293 (A_2A only) cells stably expressing one of three hAR subtypes, unless noted (n = 3–5). The binding affinity for hA₁, A_2A and A₃ARs was expressed as K_i values using agonists [³H]N⁶-R-phenylisopropyladenosine 51, [³H]2-[p-(2-carboxyethyl)phenyl-ethylamino]-5′-N-ethylcarboxamidoadenosine 52, or [¹²⁵I]N⁶-(4-amino-3-iodobenzyl)adenosine-5′-N-methyluronamide 53, respectively. A percent in italics refers to inhibition of binding at 10 μM. Nonspecific binding was determined using adenosine 5′-N-ethyluronamide 54 (10 μM). Values are expressed as the mean ± SEM (n = 3, unless noted). K_i values were calculated as reported.⁴² Off-target interactions determined by the PDSP. Receptor abbreviations are defined in the Supporting Information. Gp, guinea pig.

^bData from Tosh et al.^33,34

^cData from Carlin et al.³⁰ and van der Hoeven et al.⁴³

^dData from Gao et al.,¹⁷ Tosh et al.²³ and Gallo-Rodriguez et al., 1994.⁴⁵

^eData from Jacobson et al.³²

^fData from Kiesman et al., 2009; Müller and Jacobson, 2011.^1,35

^gData from Jacobson et al.⁴⁴ and Rodríguez et al.³⁶

^hFunctional EC₅₀ (nM) at hA_2BAR: 16, 140±19; 18, 948±38.⁴⁴

ⁱn = 1.

^jPercent inhibition at 10 μM.

ND, not determined.

On the ribose or pseudoribose moiety, adenosine-like 4′-CH₂OH analogues, 4′-CH₂Cl analogues, 4′-truncated analogues and 5′-N-alkyluronamides were included. Monosubstituted adenosine 5′-N-alkyluronamides 15–18 are known AR agonists upon which various A₃AR-selective (N)-methanocarba analogues were based,^32,35 although their selectivity across species has not been fully characterized.^17,36 Small 5′-N-alkyluronamides of adenosine, 15 and 16, are potent but nonselective AR agonists. Certain ribose analogues with additional 5′-amide modifications were previously noted to increase hA₁AR selectivity, including the 5′-N-cyclopropyl 17 and 5′-N-(2-hydroxyethyl) 18 uronamide derivatives of adenosine.²³ The same 5′ substituents were incorporated into (N)-methanocarba derivatives 36 and 37, but their A₁AR selectivity was reduced. 5′-Ester derivatives 32, 33 (2′,3′-protected intermediate) and 35 were also included among these prospective A₁AR agonists, although ethyl ester 32 and the corresponding methyl ester displayed only partial agonism at the A₃AR.³⁴ Nevertheless, at the hA₁AR 5′-ethyl ester 32 was already shown to be a full agonist (89% of maximal efficacy (E_max) of 1), therefore making esters relevant to the study of A₁AR agonists.³⁴ Although alkyl esters in general have limited in vivo half-lives, an (N)-methanocarba nucleoside 5′-ester derivative, similar to 32, was shown to be stable with respect to hydrolysis in vivo and in liver microsomes.³⁴

Several modified nucleobases were studied to expand the range of A₁AR agonists. Among simple adenosine derivatives, a 3-deaza modification was found to increase hA₁AR selectivity.³⁶ Therefore, we prepared compound 30, the 3-deaza analogue of 29, to see if this selectivity is also enhanced in the methanocarba series. One group of analogues containing a nonadenine nucleobase (a 1,2,4-triazole) was additionally included. The antiviral therapeutic nucleoside ribavirin 38 was reported to be a weak, but possibly selective agonist (K_i 86 μM at bovine A₁AR; 7.4 μM at hA₁AR).^36,37 Therefore, we prepared two (N)-methanocarba analogues of ribavirin, i.e. methyl ester 39 and its precise equivalent, carboxamide 40.

Synthesis of target nucleosides

The nucleoside derivatives were prepared as shown in Schemes 1 and 2 and Schemes S1 – S3 (Supporting information) from known starting nucleoside intermediates.^32,34 The route to (N)-methanocarba derivatives substituted at the N⁶ and C2 positions with typical A₁AR-enhancing groups and at the 5′ position with an aryl thioether is shown in Scheme 1 and closely parallels previous syntheses.³⁴ After functional group replacement at the 6-position and deprotection at the 5′ position, intermediates 60–62 were either fully deprotected, to give 24, 25 and 29, or substituted with 5′-chloro to give subsequently 26, 27 and 31 or progressed to the 5′-thioether 28. Scheme 2 shows the synthesis of ribose derivatives 9–12 containing the novel N⁶-dicyclobutylmethyl group and substituted at the 5′ position with an alkyl thioether. Starting with tribenzoyl-protected nucleoside precursors 67 and 68, synthesized as reported,^38,39 nucleophilic substitution at the 6 position with dicyclobutylmethylamine was followed by protecting group replacement to yield 2′,3′-isopropylidene intermediate 71, which was subjected sequentially to 5′-Cl substitution, thiol substitution and isopropylidene deprotection to yield 10 and 11. Alternatively, compound 9 was prepared in one step from 6-chloropurine-9-riboside using published methods.²⁸

Scheme 1.

Synthesis of (N)-methanocarba derivatives bearing typical A₁AR-enhancing groups at the N⁶ and C2 positions with and optionally substituted at the 5′ position with chloro or an aryl thioether. Reagents and conditions: (i) R¹NH₂, DIPEA, 2-propanol, rt; (ii) TBAF, THF, rt; (iii) 10% TFA, MeOH, 70 oC; (iv) SOCl₂, pyridine, CH₃CN, −5 oC-rt (v) 2-F-PhSH, NaH, DMF, 0 oC-rt.

Scheme 2.

Synthesis of ribose derivatives containing the novel N⁶-dicyclobutylmethyl group and optionally substituted at the 5′ position with chloro or an alkyl thioether. Reagents and conditions: (i) dicyclobutylmethyl-amine, DIPEA, 2-propanol, rt; (ii) NH₃/MeOH, rt; (iii) 2,2-dimethoxy propane, p-TSA, acetone, rt; (iv) SOCl₂, pyridine, CH3CN, 0 oC-rt; (v) EtSNa, DMF, rt; (vi) 10% TFA, MeOH, 70 oC.

Scheme S1 (Supporting information) shows the route to (N)-methanocarba derivatives containing an N⁶-endo-norbonyl group. The synthesis of additional (N)-methanocarba or 5′-ester derivatives, including those bearing alternative nucleobases are shown in Schemes S2 and S3 (3-deaza-adenine in A or 1,2,4-triazole in B).

In vitro pharmacological characterization

The nucleoside analogues were examined in radioligand binding assays (Table 1) at four hARs and two or three mARs as previously described.^30,33,40 In cases where there was only ~50% A₃AR binding inhibition at 10 μM, an estimated K_i of ~10 μM was used in approximating the selectivity ratio. The hA_2BAR affinities of selected compounds were measured in a radioligand binding assay using [³H]54,⁸¹ and most analogues tested showed <50% binding inhibition at 10 μM. Values at hA_2BAR were (K_i, μM or % at 10 μM): 9, 8.62 μM; 19, 24%; 35, 38%; 38, 39%.

Comparison of ribose 5′-N-alkyluronamides.

In the ribose series, 5′-N-alkyluronamides (14–18) displayed variable AR selectivity, and several of these groups were later applied to (N)-methanocarba analogues. The 5′-N-cyclopropyluronamide 17 was notably 95-fold hA₁AR-selective compared to hA₃AR.⁴⁴ Thus, for achieving A₁AR selectivity in 5′-N-alkyluronamides, neither the commonly applied 5′-N-methyl- or 5′-N-ethyl-uronamide was optimal. At the hA₃AR the 5′-N-methyl-uronamide 15 was best suited for optimizing affinity and selectivity, similar to previous results with the rat (r) ARs.⁴⁵ In contrast, the corresponding carboxylic acid 14 was inactive or only weakly active in hAR binding, suggesting that a negative charge in that polar ribose-binding region of the AR binding site is not tolerated.

Introduction of other ribose modifications.

Riboside and (N)-methanocarba derivatives were initially compared with the known N⁶-endo-norbornyl (3, 4, 13, 29–31, 36 and 37) substitution. In general, for this bulky N⁶ group, the selectivity for hA₁AR or mA₁AR compared to A₃AR was greater among riboside derivatives than among similar methanocarba derivatives. For example, the mA₁AR selectivity for N⁶-endo-norbornyl derivatives was 3000–12,000 fold in the ribose series and 70–500 fold in the methanocarba series, and the hA₁AR selectivity showed an even greater divergence in each case. Each of the following three N⁶-dicycloalkylmethyl ribose analogues bound with similar or higher hA₁AR selectivity (fold) compared to hA₃AR than the equivalent methanocarba analogues: riboside 9 (4950) compared to methanocarba 24 (~400), 10 (3390) compared to 26 (~300), 12 (742) compared to 25 (>1000). Similarly, at the mA₁AR, there tended to be lower selectivity (fold) compared to mA₃AR in the methanocarba series: 9 (2420) compared to 24 (1730), 10 (2970) compared to 26 (890), 12 (1000) compared to 25 (248). Thus, among adenine derivatives examined, the ribose series generally resulted in comparable or greater A₁AR selectivity.

The ribose hydroxyl groups favored A₁AR selectivity. The N⁶-dicyclopropylmethyl analogues’ A₁AR selectivity that was present in the ribose 8 and (N)-methanocarba 5′-OH 23 analogues was lost in the 5′-ester 32 and 5′-N-methyluronamide 34 series. As expected from structural information and previous SAR studies,⁴⁵ the ribose 2′- and 3′-OH groups are important for AR recognition (cf. weakly binding isopropylidene derivatives 33).

Other ribose 5′ position modifications were compared. As with the known selective agonist 4, a 5′-deoxy-5′-Cl modification in ribose derivative 10 was conducive to high hA₁AR (3390-fold) and mA₁AR (2970-fold) selectivity compare to A₃AR. The simple 5′-Cl analogue 22 exceeded the hA₁AR affinity of the parent (N)-methanocarba 2-Cl-adenosine 21 by 14-fold with only minor effects on hA_2AAR and hA₃AR affinity, thereby increasing A₁AR selectivity. However, in the 2-Cl methanocarba series with a large N⁶ group, both h and mA₁AR selectivity was most pronounced in 5′-OH derivative 25, which displayed K_i < 10 nM at both h and mA₁AR. The corresponding 5′-Cl derivative 27 was less potent and selective than 25 at the mA₁AR (140-fold selective compared to mA₃AR) and at the hA₁AR (32-fold compared to hA₃AR). An inactive methanocarba analogue 28 resembled known partial A₁AR agonist 5 in having a large 5′-thioether. The A₁AR selectivity of (N)-methanocarba derivatives 36 and 37 was reduced compared to the ribose series. Thus, a bulky 5′ position substitution was incompatible with a methanocarba modification using the criterion of A₁AR affinity, and either a 4′-CH₂OH or 5′-Cl group (depending on N⁶ substitution) was the best suited for maintaining A₁AR selectivity in the methanocarba series.

Comparison of N⁶ substituents.

The N⁶-dicyclopropylmethyl group was incorporated into six analogues (7, 8, 23, 32–34). As previously reported,^30,33 the N⁶-dicyclopropylmethyl group in the 4′-truncated methanocarba analogue 7 was associated with low (hA₁AR) or high (mA₁AR) selectivity.³⁰ Among other N⁶-dicyclopropylmethyl analogues only 4′-CH₂OH derivative 23 was highly mA₁AR selective (1450-fold compared to mA₃AR) but only moderately hA₁AR selective (41-fold compared to hA₃AR). However, the bulky N⁶-endo-norbornyl group was particularly suited for A₁AR-selectivity in the ribose series. N⁶-endo-Norbornyl derivative 13 is a pure diastereoisomer of known agonist 4 and was 467-fold hA₁AR-selective compared to hA₃AR

In the (N)-methanocarba series, the 2-chloro-N⁶-endo-norbornyl derivatives 29 and 31 are selective at the mA₁AR (550- and 270-fold, respectively, compared to mA₃AR), but only slightly selective at the hA₁AR (12- and 10-fold, respectively, compared to hA₃AR). Thus, the high A₁AR selectivity across several species associated with the N⁶-endo-norbornyl group in 5′-OH analogue 3 and 5′-Cl riboside 4 (±, a diastereomeric mixture)¹⁷ was generalizable to the methanocarba series.

In the series of 4′-truncated cycloalkyl homologues of methanocarba derivative 7, the corresponding N⁶-dicyclopentylmethyl analogue 20 was inactive in binding at the hARs.³³ The intermediate N⁶-dicyclobutylmethyl homologue 19 was weak and nonselective. Thus, in this truncated methanocarba series, the N⁶-dicyclopropylmethyl group of 7 provided the highest A₁AR affinity.

However, when a 4′-CH₂OH was present, the preferred ring size differed from the truncated series. The novel N⁶-dicyclobutylmethyl ribose derivative 9 (2-H-adenine) was highly potent in A₁AR binding. Compound 9 was 2400-fold mA₁AR selective compared to the mA₃AR; thus, it displayed equivalent selectivity and nearly the same mA₁AR affinity (K_i 0.37 nM) as reference compound 1a, while at the hA₁AR 9 was ~5000-fold and 1660-fold selective compared to hA₃AR and hA_2AAR, respectively. The affinity of 9 at hA_2BAR was also very weak. Thus, the N⁶-dicyclobutylmethyl group was identified as promoting multispecies A₁AR selectivity in adenosine derivatives.

Nucleobase modifications.

Only two variations at the adenine C2 position, H and Cl, were compared. With a 2-Cl present, the corresponding N⁶-dicyclobutylmethyl riboside 12 was 1000-fold mA₁AR selective compared to mA₃AR, but with slightly lower affinity at the mA₁AR than its 2-H analogue 9. Similarly, at the hA₁AR, the presence of 2-Cl led to reduced selectivity. Compound 12 bound at the hA₁AR with 8-fold lower affinity than 9. Thus, in this series, 2-Cl was not advantageous for achieving A₁AR selectivity over A₃AR.

The 3-deaza-adenine methanocarba analogue 30 (5′-OH) was balanced in hA₁AR/hA₃AR affinity, i.e. the A₁AR, but not the A₃AR affinity was reduced by an order of magnitude by the loss of N3 when compared to its adenine equivalent 29. Thus, it was not advantageous to omit N3 for achieving hA₁AR selectivity in the methanocarba series. Other deaza-adenine modifications were not examined.

With an alternative nucleobase, i.e. 1,2,4-triazole-3-carboxamide, the methanocarba analogue 40 was ~400-fold selective for the mA₁AR and showed a 15-fold greater hA₁AR affinity compared to known riboside 38. Thus, while the methanocarba modifcation reduced A₁AR selectivity in the adenine series, with this alternative nucleobase, mA₁AR selectivity was increased.

Functional assays of selected derivatives.

Truncated nucleoside 7 was previously shown to be a full agonist at the hA₁AR in inhibition of cAMP accumlation.³³ Here we demonstrated that it is a full agonist at the rA₁AR expressed in CHO cells with an EC₅₀ of 6.65±0.32 nM in stimulation of [³⁵S]GTPγS binding (Figure 1A, E_max = 119% of 1 μM 16), determined as reported.⁴¹ In comparison, 2 displayed an EC₅₀ of 1.65±0.47 nM, E_max = 103%. At the hA₃AR, compound 7 proved to be a partial agonist with an EC₅₀ value of 1.37 μM with ~half the E_max of 16 (Figure 1B). Functional assays using [³⁵S]GTPγS binding evaluated mA₁AR and mA₃AR agonism (Table 2). All nucleosides tested were full agonists at mA₁AR, with N⁶-endo-norbornyl (N)-methanocarba derivative 31 displaying a maximal 123% efficacy compared to standard 16. Compound 24 was confirmed to be a full mA₃AR agonist at 10 μM with E_max of 92±3%, compared to 16. Compound 9 displayed 70% E_max at 10 μM, while other nucleosides (12, 24, 29 and 31) were partial mA₃AR agonists with E_max of only 31 – 63%.

Figure 1.

A. Functional activity of agonists 2 and 7, in stimulation of guanine nucleotide binding at the rA₁AR (recombinant A₁AR membrane preparations from CHO-K1 cells, Perkin Elmer, compared to 2) B. Effects of agonists 7 and 16 in inhibition of cAMP accumulation at hA₃AR (in A₃AR-expressing CHO cells, treated with 10 μM forskolin, compared to 16). 100% value defined as effect of 1 μM 16. Also, functional assays at the hA₁AR are shown for several derivatives (EC₅₀ or IC₅₀ in nM): stimulation of [³⁵S]GTPγS binding (C, 9, 28.0±9.0; 16, 0.12±0.05; 40, 758±175); inhibition of forskolin-stimulated cAMP production (D, 9, 0.14; 40, 87); β-arrestin2 recruitment (E, 9, 209±90; 16, 5.03±2.84; 40, 2460±800).

Table 2.

Functional activity of selected nucleoside derivatives (%E_max at 10 μM) at mA₁AR and mA₃AR in a guanine nucleotide binding assay.

Compound (MW, cLogPa)	Binding Ki (and selectivityb) for mA1AR	Emax, mA1AR, %c	Emax, mA3AR, %c
9 (389, −1.10)	0.37 nM (2420)	101±7	70±9
12 (424, −0.39)	0.65 nM (1000)	90±19	63±5
24 (400, −1.62)	0.53 nM (1730)	97±11	92±3
29 (406, 1.18)	1.05 nM (547)	104±6	62±5
31 (424, 2.38)	1.86 nM (270)	123±16	50±11
40 (254, −4.52)	25.2 nM (~400)	ND	31±3

^aCalculated using ChemDrawProfessional, v. 16.0.

^bFold, compared to mA₃AR (Data from Table 1).

^cMean (effect of 10 μM of test compound as a % of 10 μM 16) ± SEM in increasing AR-mediated [³⁵S]GTPγS binding, n = 3–6.

ND, not determined.

Several compounds were compared in functional assays at the hA₁AR: stimulation of [³⁵S]GTPγS binding; inhibition of forskolin-stimulated cAMP production; β-arrestin2 recruitment (Fig. 1C–E).⁴¹ Varying agonist efficacies are shown for 9, 38 and 40 in comparison to standard agonist 16. The (N)-methanocarba analogue 40 appears to be more potent and efficacious than the parent antiviral drug 38.

Off-target interactions.

Table 1 shows other weak (μM) interactions detected for various nucleosides, at serotonin receptors, histamine receptors, other G protein-coupled receptors (GPCRs), sigma receptors and transporters. Key agonists 9 and 40 had several off-target interactions at transporters and receptors with K_i values > 1 μM, i.e. considerably weaker than their mA₃AR affinity. However, previously identified³² mixed agonist 6 bound to the 5HT_2C serotonin receptor with a K_i of 12 nM. Therefore, minimizing the off-target activity of the present series at 5HT_2BR and other GPCRs, determined by the NIMH Psychoactive Drug Screening Program (PDSP, Figure S1, Supporting information),⁵⁶ was an important consideration in the SAR optimization. Broad screening data (DiscoverX) of truncated methanocarba derivative 7 (at 10 μM) showed that it is not promiscuous, as it interacts only at the A₃AR and not in other GPCR assays (Table S1, Supporting information) except for the 5HT_2B serotonin receptor (5HT_2BR), in agreement with our previous report.³⁴ In a kinase screen, only two weak hits at 10 μM were observed for 7 among 403 kinases (Figure S2, Supporting information).

In vivo biological characterization and ADME-tox studies

Selected analogues (9, 10, 12, 24–27, 29, 31 and 40, Table 3) were examined for the ability to induce hypothermia in the mouse when administered ip. and were compared with published data in the same assay for compounds 1a, 4 and 7.³⁰ With 1a and 7, the peripheral A₃AR agonist hypothermic effect predominated, while 4 had a clear central A₁AR agonist effect.³⁰ Compound 9 produced hypothermia and hypoactivity at 1 or 3 mg/kg i.p in wild-type mice. As expected,³⁰ hypothermia was generally accompanied by reduced activity for all compounds tested. There was a loss of 9-induced hypothermia in A₁AR KO and in A₁AR/A₃AR double KO mice, but not in A₃AR KO mice (Figure 2A–E). Thus, the hypothermia induced by 9 was dependent on A₁AR, without a contribution from A₃AR. 5′-Chloro derivative 31 at up to 3 mg/kg ip. did not significantly induce hypothermia in WT mice (Table 3). Furthermore, compounds 10, 12, 24–27, and 40 (at 3 or 10 mg/kg, ip.) also had no pronounced effect on Tb (Figure S3, Supporting information). In addition, there was a hypothermic effect in response to (N)-methanocarba 5′-OH analogue 29 (3 mg/kg, ip.). The hypothermia induced by 29 was slightly attenuated in mice lacking A₁AR, more diminished in mice without A₃AR, and completely lost in mice missing both A₁AR and A₃AR (Figure 2F–J). These results suggest that 29 caused hypothermia via both central A₁AR and peripheral A₃AR.³⁰ The nucleoside effects on locomotor activity depression usually paralleled the effects in hypothermia (Table 3). In Figure 3, dose response curves for hypothermia and locomotor depression induced by 9 are compared.

Table 3.

Hypothermia and locomotor activity parameters in wild-type C57BL/6J mice.

Compound	Dose (mg/kg)	Tb (°C)	Tb P value	Activity (counts)	Activity P value	ARs causing hypothermia
Vehicle		36.3±0.8		15.5±5.5
1a	1	32.4±0.5	<0.0001	1.5±0.5	0.0078	A3AR > A1AR30
4	3	32.2±1.3	0.0014	2.8±1.6	0.0055	A1AR > A3AR30
7	3	33.4±1.7	<0.0001	5.5±4.1	<0.0001	A3AR >> A1AR30
9	3	33.0±0.7	<0.0001	6.6±2.3	<0.0001	A1AR selective (Figure 2)
10	10	35.8±0.1	ns	20.5±4.6	ns
12	3	35.5±0.8	0.047	13.5±3.5	ns
24	3	36.2±0.4	ns	18.1±6.8	ns
25	10	36.6±0.2	ns	18.4±3.2	ns
26	3	35.9±0.5	ns	14.8±1.7	0.041
27	10	36.5±0.6	ns	18.9±10.2	ns
29	3	33.4±0.7	<0.0001	10.0±4.3	0.046	A3AR > A1AR (Figure 2)
31	3	36.6±0.5	0.055	16.8±3.6	0.040
40	10	35.7±0.7	ns	19.1±5.3	ns

Compounds were dosed ip. Tb and activity were measured by telemetry and the averages from dosing to 60 min were calculated. P value >0.05 was considered not significant (ns).

Data are for the highest dose used and are mean ±SD from 3 to 22 mice/group (usually 5–7).

P values are unpaired t-Test vs vehicle-treated controls assayed simultaneously.

Vehicle data are mean ±SD from all (n=83) vehicle treated mice.

Figure 2.

Effects of nucleoside derivatives on Tb in WT (C57BL/6J male) or AR-KO mice. Compound 9 (A-E) or 29 (F-J), both at 3 mg/kg, ip., and vehicle were dosed in wild-type mice (WT, A, F), or mice lacking A₁AR (B, G), A₃AR (C, H), or both A₁AR and A₃AR (D, I). E, J. Tb during the first 60 min after dosing, mean ±SEM, N =4 to 12/group. t-Test (2-tailed, paired or unpaired as appropriate) P values for vehicle vs compound, within genotype.

Figure 3.

Effect of compound 9 (ip) on body temperature and locomotor activity in wild-type C57BL/6J mice.

The pharmacokinetics of 9 (3 mg/kg, p.o.) in the male rat (Table S2, Figure S4, Supporting information) demonstrated a maximal concentration of 121±44 ng/mL, which corresponds to 310 nM. The oral bioavailability was 31–43 %F, depending on the dose (1 – 10 mg/kg), and the oral half-life in plasma was 0.9 – 1.3 h. Thus, peripheral concentrations of 9 in the hypothermia tests are expected to be above its A₁AR K_i value for hours after dosing.

A compound that failed to lower Tb when administered ip., 24, produced intense hypothermia when given intracerebroventricularly (icv., Figure S3C,D, Supporting information). Both compounds 9 and 24 administered by an icv. route caused robust hypothermia with Tb dropping to ~24 °C (5 μg injected, which is equivalent to 0.157 mg/kg). This indicated that compound 24 can cause hypothermia centrally but lacks sufficient concentration in the brain when administered peripherally. In the second experiment, with doses of 1.67 μg/mouse (= 0.052 mg/kg), 9 induced deep hypothermia lasting for ~10 h while 24 reduced Tb only slightly.

As we reported, compound 7 displayed the optimal A₁AR selectivity in the series of 4′-truncated N⁶-dicycloalkylmethyl analogues.³³ The pharmacokinetics of 7 was determined in the mouse, demonstrating elimination t_1/2 values of 2.5 to 4 h following ip. (3, 15, or 30 mg/kg) administration and an elimination t_1/2 of 5 to 11 h with oral administration (dose-dependent; 1, 3, 10 mg/kg) (Figures S5 and S6, Supporting information). The bioavailability was calculated to be 98 and 124 %F for 1 and 3 mg/kg p.o., respectively. Thus, compound 7 has high oral bioavailability. Other ADME-tox measurements (Table S3, Supporting information) also demonstrated drug-like properties of 7: stable in simulated intestinal fluid and human liver microsomes, 51% remaining in simulated gastric fluid (120 min), IC₅₀ at hCYP P450 enzymes (2C9, 2D6, 3A4) >25 μM and CACO-2 cell efflux ratio of 0.91 (i.e. no efflux). Plasma protein binding was 90% (rat) and 94% (mouse).

The cardiovascular effects of 7 and prototypical A₁AR agonist 2 were examined preliminarily in the anesthetized rat (Figure S7, Supporting information), although heart rate is known to be lowered by anesthesia. A high dose of 7 (3 mg/kg, ip.) displayed modest effects on heart rate (slow onset of reduction) and blood pressure (a modest, transient rise). Although not directly comparable, a low dose of 2 (0.02 mg/kg, iv.) produced a sustained rise in blood pressure.

The pharmacokinetics of 7 in the monkey was also determined following oral and iv. administration (Figure S8, Supporting information). After consuming (in a food treat) compound 7 (1 or 3 mg/kg body weight), female rhesus monkeys (awake and fed) displayed no adverse reaction up to 4 h post-dosing. The pharmacokinetics of 7 was then determined in awake male cynomolgus monkeys following iv. (0.5 mg/kg) and p.o. (1.0 mg/kg) administration. The bioavailability was estimated as 24%, and there were no clinical observations noted throughout the one-day observation period. Thus, this A₁AR agonist was well tolerated in non-human primates.

Beyond the 4′-truncated series, the highly A₁AR-selective agonist 9 was examined in other ADME-tox tests (Supporting information). The ligand efficiency (LE) of 9 is a favorable 0.43, and other parameters predict drug-likeness. 9 showed no inhibition against 1A2, 2C19, 2D6, 3A4 CYP isozymes and weak inhibition at the 2C9 isozyme. It was nontoxic in HEP-G2 hepatocytes, displayed a hERG IC₅₀ of >30 μM (fluorescence polarization assay) and was highly soluble in aqueous medium containing 1% DMSO at H 7.4 (>193 μg/kg, Supporting information). The compound showed high permeability in CACO-2 cells, with a low efflux ratio of 3.68. It was completely stable in simulated gastric and intestinal fluids and in plasma of three species (h, m and rat), but showed medium clearance in liver microsomes of the same species.

Another factor to consider in light of the in vivo activity profiles is that the 2-Cl group, as in 12, decreases the polarity (cLogP 9 = −1.1; cLogP 12 = −0.39) and might increase plasma protein binding. If so, this would reduce the free fraction of drug and influence its ability to induce mast cell degranulation and to freely diffuse into the CNS. Therefore, the plasma protein binding of 9 and five other analogues (Table 4) was determined in three species (m, r, h). The analogue 40 of ribavirin was estimated to be largely free in plasma, unlike other analogues. Curiously, substituting ribose 9 with (N)-methanocarba 24 tended to increase the free fraction. Conversely, substituting a chlorine atom at the 2 position (12 vs. 9, for m and r) or 5′ position (31 vs. 29, for all three species) decreased the free fraction. In mouse plasma the free fractions of 24 (~28%) was greater than for 9 (~21%). Therefore, a comparison of plasma protein binding did not explain the inactivity of 24 in inducing hypothermia.

Table 4.

Plasma protein binding at 10 μM in three species, expressed as unbound percent.^a

Compound	human, % freeb	mouse, % freeb	rat, % freeb
9	6.05±1.60	20.6±1.2	14.6±0.01
12	5.81±0.39	4.74±0.67	4.64±0.06
24	14.6±1.3	27.5+1.7	17.8±1.2
29	9.75±2.03	7.68±0.39	6.59±0.49
31	4.04±0.60	2.95±0.84	2.53±0.45
40	c	c	c

^aProcedure is in the Supporting information. Verapamil and warfarin served as positive controls.

^bMean ± SD.

^cEstimated >80% free.

Molecular Modeling

Although we previously based hA₁AR homology modeling on an agonist-bound hA_2AAR, we sought a closer template for this study.⁴⁶ The current hA₁AR antagonist-bound X-ray structures were found to be unsuitable as templates for agonist docking by us and others.²⁹ Thus, modeling hA₁AR interactions was based on the recently reported cryo-electron microscopic (cryo-EM) structure.¹⁶ We modeled the hA₁AR interactions of two selective A₁AR agonists: riboside 9 and its methanocarba equivalent 24.

No water molecules were resolved in the cryo-EM structure of hA₁AR, but in the adenosine-bound hA_2AAR X-ray structure (2YDO⁵⁹) a water molecule mediates a H-bond network involving the adenosine 5′-OH group, the conserved His250 (6.52, His251 in hA₁AR) and Asn181 (5.43, Asn184 in hA₁AR). Assuming that a water molecule could have the same role in hA₁AR, this molecule was copied to the hA₁AR-adenosine complex and minimized to result in formation of a similar H-bond network with 5′-OH, His251(6.52) and Asn184(5.43).

The binding pocket entrance in the cryo-EM hA₁AR structure is hindered by some bulky amino acid side chains, which interfere with computational docking of N⁶-substituted adenosine derivatives. For this reason, Glu172 (EL2), Met177 (5.37), Thr257 (6.58), Lys265 (EL3) and Thr270 (7.34) were temporarily mutated into Ala residues before performing docking simulations. After obtaining a docking pose for compound 9, the aforementioned residues were retro-mutated restoring the original coordinates and were minimized together with the docked ligand. To assess the stability of the proposed binding mode and to observe its temporal evolution, the complex obtained after molecular docking was subjected to MD simulations. Moreover, the methanocarba equivalent of compound 9, compound 24, was superposed by flexible alignment to compound 9’s selected pose, to allow comparison of the two analogs.

To better reproduce the planarity of the N⁶ nitrogen atom during MD, the dihedral angle C5-C6-N6-C1 (Figure S9) parameters were optimized by fitting the QM Potential Energy Surface (PES) obtained by a dihedral angle scan. 30 ns MD simulations were run (in triplicate) for both the compounds, and the trajectories were analyzed after aligning the protein to the Cα carbon atoms of the initial receptor structure.

Both compounds appeared stabilized in the suggested binding mode, as indicated by an average heavy atom root-mean-square deviation (RMSD) of 1.6 Å in the case of compound 9 and of 1.7 Å in the case of compound 24. Moreover, both compounds’ RMSD and ligand-receptor interaction energies tended to plateau (Fig. SI10–11, A-B) indicating both geometrical and energetical stability. The ligand was anchored to the receptor by a pattern of interactions that are typical for the AR family, involving His251 (6.52), Asn254 (6.55), Thr277 (7.41), His278 (7.42) (Fig. 4, panel A-C; Fig. SI10–11, panel D). The side chain amide of Asn254 (6.55) was engaged in a bidentate H-bond with each ligand, involving N⁶H as H-bond donor and N7 as acceptor, which was maintained during each entire simulation. The ribosyl 2′-OH group of 9 transiently interacted with His278 (7.42), while this direct contact was less prominent for 24. In the starting adenosine-bound conformation of hA₁AR, the hydroxyl group of Thr277 (7.41) pointed far away from 3′-OH group, contrary to the residue at equivalent position of hA_2AAR, Ser277 (7.41), which interacts with the 3′ group of ribose through a H-bond. However, a rotation of Thr277 was observed during the MD simulations, together with the formation of a H-bond with 3′-OH group of both 9 and 24. Another key residue was His251 (6.52), which interacts through a direct or water-mediated H-bonding with the 5′-OH group of each compound.

Figure 4.

Conformation of compounds 9 (panel A-B) and 24 (panel C-D) bound to hA₁AR (PDB ID: 6D9H) in the last frame of a 30 ns MD simulation. Compounds 9 and 24 are respectively rendered by orange and green sticks. In panels A and C the receptor is represented by a gray ribbon (TM7 transparent to allow the visualization of the pose), with significative residues rendered by gray sticks. Water molecules within 3 Å from the ligands are reported and the pattern of H-bonds is highlighted by green lines. Panels B and D report a different view of the complexes, with the binding site surface colored according to residue hydrophobic character (from purple to cyan for hydrophobic to hydrophilic contributions).

Another important contribution to the stabilization of the bound state was provided by the conserved π- π interaction between Phe171 (EL2) and the aromatic purine base over time. The two cyclobutyl substituents fit into two clefts, respectively between EL2, TM5 and TM6 in one case, and TM6 and TM7 in the other. The character of the two clefts is mainly hydrophobic, with one cyclobutyl surrounded (4 Å distance) by Phe171 (EL2), Glu172 (EL2), Met177 (5.37), Met180 (5.40), Asn254 (6.55) and the other by Leu250 (6.51), Leu253 (6.54), Thr257 (6.58), Lys265 (EL3), Thr270 (7.34), Ile274 (7.38) (Fig. 4, panel B-D, Fig. SI10–11, panel C).

During simulation of 9, an approach between Glu172 (EL2) and Lys265 (EL3), which in the cryo-EM structure are ~7 Å apart, is observed, resulting in a water-mediated interaction. This resembles the interaction between A_2AAR Glu169 (EL2) and His264 (EL3) reported in many crystal structures,⁷⁵ which is known to have a role in decreasing the dissociation rate of A_2AAR ligands,⁷⁶ hinting at a similar effect in hA₁AR.

Discussion

Additional AR agonists that display A₁AR selectivity in vivo and are well tolerated are still needed. Often the dose range of an agonist to produce a desired effect of A₁AR activation, such as antiseizure activity, without toxic side effects is narrow.^4,33 Few A₁AR agonists are suitable for dose escalation in vivo due to their cardiovascular effects,^1,4,10 but we have shown that (N)-methanocarba derivative 7 is an exception to this usual caveat of A₁AR agonists, being non-lethal in the mouse at 60 mg/kg. Compound 7 within the 4′-truncated series achieved the optimal A₁AR binding selectivity, as other N⁶-dicycloalkyl groups failed to increase A₁AR affinity.³³ We have further characterized its A₁AR full agonism and its weak A₃AR partial agonism. 7 is relatively stable and well tolerated in vitro and in vivo, in rodents and nonhuman primates. Along with 7 as a structural lead in this series, we patterned some of the target molecules on the endo-norbornyl analogue 4, a highly selective A₁AR agonist.

Unlike 4, which has a racemic substituent as the N⁶ group, compounds 3, 13, 29–31, 36 and 37 are stereochemically pure in the form that is more potent in rA₁AR binding,²⁸ rather than being a diastereomeric mixture. This defined endo-norbornyl group stereochemistry is distinctly advantageous for pharmacological studies. Another N⁶ group present in many analogues, dicyclobutylmethyl, has never before appeared in published reports on adenosine agonists. However, it is an important molecular lead toward the refinement of selective A₁AR agonists that are well tolerated in vivo.

In a comparison of non-truncated ribose and methanocarba derivatives, we found that A₁AR-selectivity is more generally associated with the former series. Furthermore, N⁶-dicyclobutylmethyl derivatives were identified as novel A₁AR agonists. Two ribose analogues, 5′-OH 9 and 5′-Cl-5′-deoxy 10 derivatives, containing the N⁶-dicyclobutylmethyl group were found to be >2000-fold selective for the mA₁AR compared to the mA₃AR. Furthermore, 9 and 10 are highly selective for the hA₁AR compared to other hAR subtypes. Our efforts to identify methanocarba derivatives that display a clear preference for the A₁AR did not reach the degree of selectivity seen in the ribose series. Nevertheless, N⁶-dicyclobutylmethyl derivatives 24–26 were selective at both hA₁AR and mA₁AR (≥200-fold selective compared to A₃AR). In the series of methanocarba N⁶-dicycloalkylmethyl analogues, the previously reported N⁶-dicyclopropylmethyl analogue 23 was less selective at the hA₁AR than the higher homologue 25 but more selective at the mA₁AR.

Focusing on nucleosides 9 and 24, we have suggested a hypothetical binding mode through MD simulations. Both compounds are stabilized by a network of conserved interactions typical of both the adenosine-bound A_2AAR X-ray crystal structure and cryo-EM A₁AR. The π-π interaction with Phe171 (EL2) as well as a H-bond network involving His251 (6.52), Thr277 (7.41), His278 (7.42) and Asn254 (6.55) and engaging respectively the ribosyl 2′, 3′, and 5′-OH groups, the exocyclic N⁶H and the endocyclic N7 nitrogen atom are maintained during the simulations. The two cyclobutyl substituents insert into two hydrophobic clefts and consequently bury receptor hydrophobic surfaces that contribute to binding through a hydrophobic effect.

Alternative nucleobases might be suitable for potent and selective A₁AR agonists.³⁶ The human plasma concentrations achieved with clinical doses of ribavirin 38 are equal to or greater than its K_i value at the hA₁AR.⁴⁷ Interestingly, coffee consumption was found to ameliorate some ribavirin side effects, including fatigue.^48,49 The AR antagonism associated with ingested caffeine doses raises the possibility that AR activation might be one mechanism responsible for a subset of the numerous side effects of ribavirin.⁴⁸ Dyspnea is associated with ribavirin use, and this is a known consequence of peripheral A₁AR activation.⁵⁰

The 1,2,4-triazole-3-carboxamide analogue 40 illustrates a means of circumventing the tendency of the (N)-methanocarba modification toward high A₃AR affinity by using an alternative nucleobase. In fact, compared to 38, its A₁AR affinity is greatly enhanced, to achieve ~400-fold mA₁AR selectivity compared to mA₃AR. However, not all non-adenine nucleobases have this enhancing effect, as was shown with the 1,3-dialkylxanthine-7-ribosides, which display only a small rA₁AR enhancement upon replacement of ribose with (N)-methanocarba (K_i 24 μM for the 1,3-di-n-butyl analogue).⁶³

In the (N)-methanocarba series, conversion of 5′-OH to 5′-Cl either slightly (26 vs. 24) or greatly reduced (27 vs. 25) the A₁AR-affinity. A comparison of 21 and 22 indicated a beneficial effect of 5′-Cl substitution in a simple (N)-methanocarba derivative. Also, a 3-deaza modification greatly reduced A₁AR affinity (30 vs. 29), but also reduced off-target interactions.

As reported, either central mA₁AR or robust peripheral mast cell mA₃AR activation can lower body temperature.³⁰ We have focused on 9 as a novel lead, and we demonstrated its ability to induce centrally-mediated A₁AR hypothermia when administered ip. in the mouse, which presumably requires it to cross the BBB. Peripheral A₃AR activation by 9 also enhances A₁AR-induced hypothermia, but the A₃AR is not required to lower Tb. Similarly, compound 29 activated both receptors to greatly reduce Tb, although its A₁AR effect was less pronounced than for 9. However, compounds 10, 24–27 and 31 did not significantly lower Tb in WT mice, although all were considerably mA₁AR-selective. Therefore, these compounds failed to activate either central A₁AR or peripheral mast cell A₃AR when administered peripherally. The lack of an A₁AR effect to cause hypothermia is consistent with low brain entry of the nucleosides.³¹ However, the lack of a peripheral A₃AR effect was surprising and might be explained by a low agonist efficacy at this receptor. As noted, there is reason to expect that 4 is not a full agonist at the A₃AR, which might contribute to its favoring the central A₁AR effect vs. the peripheral A₃AR mast cell effect in Carlin et al.³⁰ Other data on adenosine derivatives with large N⁶ groups show that they tend to be partial agonists or antagonists at the A₃AR, while fully activating the A₁AR.¹⁷ This could be an important factor in the overall selectivity of 4 as seen in vivo, in addition to its 10,000-fold binding selectivity for mA₁AR vs mA₃AR. However, several compounds that are similar to 9 in efficacy measurements (Table 2) did not lower Tb. It is worth noting that the mA₃AR functional assay in Table 2 might overestimate the intrinsic efficacy of the nucleosides at that subtype, because it utilized HEK cells overexpressing the receptor. Also, the nucleoside concentration (10 μM) used to determine E_max in the guanine nucleotide binding assay is most likely higher than the in vivo plasma concentration achieved for each nucleoside, based on the rat pharmacokinetic data for 9. Thus, many compounds might lack robust activation of mA₃AR receptor in vivo.

We also have to consider a pharmacokinetic explanation for the lack of hypothermia of certain analogues, such as 10. Since many nucleosides do not appreciably cross the BBB,³¹ it is not surprising that most analogues administered ip. did not elicit an A₁AR-mediated hypothermia. There is no reason to expect that these compounds are rapidly metabolized or excreted in the mouse, as the relative stability of 9 was shown in the rat. The molecular weights of 4 and its dicyclobutylmethyl analogue 10 are roughly equivalent, so mass alone is not a factor, although shape might be a factor in the BBB permeability.⁵¹ There is no direct correlation between ability to induce hypothermia and molecular weight or cLogP of any of the nucleosides in Table 2.

The moderate in vivo potency of 9 to induce hypothermia in the mouse is not comparable to high potency of relatively nonselective AR agonists, such as 16 and N⁶-(R-phenylisopropyl)-adenosine.^52–54 A possible explanation might result from robust activation of multiple AR subtypes and/or less hindered distribution in the body by 16 and its congeners.

High plasma protein binding might also account for the reduced ability of a given nucleoside to enter the brain.⁸⁰ Although these novel nucleosides have closely related structures, knowing the free fraction of each nucleoside would help determine the available, free concentration of agonist to diffuse across the BBB and to act on the mA₃AR in mast cells. However, the data did not provide a plausible explanation for the different effects in hypothermia. A comparison of the hypothermic effects of 9 and 24, administered either ip. or icv., indicated a greatly reduced ability of 24 to cross the BBB compared to 9.

Conclusions

(N)-Methanocarba adenosine derivatives and corresponding ribosides were synthesized and compared pharmacologically to identify novel A₁AR agonists for CNS or peripheral applications. The N⁶-dicyclobutylmethyl ribose analogue 9 and the previously reported truncated N⁶-dicyclopropylmethyl (N)-methanocarba mA₁AR-selective agonist 7 were drug-like. In the methanocarba series, appropriate adenine or pseudoribose functionalization, or use of an alternative ribavirin nucleobase could provide selectively binding A₁AR agonists. Nevertheless, A₁AR selectivity, such as in stereochemically unambiguous N⁶-endo-norbornyl derivatives, was more consistently achieved with a ribose ring. Hypothermia induced by BBB-penetrant 9 was primarily driven by the A₁AR, which is evidence that the nucleoside is CNS-active. Other nucleosides that selectively bind at A₁AR and are full agonists did not induce hypothermia, indicating insufficient brain concentrations. However, it was surprising that only a few of these compounds elicited an A₃AR-mediated hypothermia. We modeled hA₁AR interactions of dicyclobutylmethyl derivatives 9 and 24 with MD, and conserved interactions were preserved. However, we have not studied the cardiovascular effects of these A₁AR agonists. These novel A₁AR agonists can now be evaluated in other A₁AR central and peripheral actions, such as in pain relief or inhibition of lipolysis.

Experimental section

Chemical synthesis

Materials and instrumentation

Most of the reagents and solvents were purchased from Sigma-Aldrich (St. Louis, MO). Other reagents were purchased from Small Molecules, Inc. (Hoboken, NJ), Anichem (North Brunswick, NJ), PharmaBlock, Inc. (Sunnyvale, CA), Frontier Scientific (Logan, UT) and Enamine (Monmouth Jct., NJ). ¹H NMR spectra were obtained with a Bruker 400 spectrometer using CDCl₃, CD₃OD and DMSO as solvents. Chemical shifts are expressed in δ values (ppm) with tetramethylsilane (δ 0.00) for CDCl₃ and water (δ 3.30) for CD₃OD. NMR spectra were collected with a Bruker AV spectrometer equipped with a z-gradient [¹H, ¹³C, ¹⁵N]-cryoprobe. TLC analysis was carried out on glass sheets precoated with silica gel F254 (0.2 mm) from Sigma-Aldrich. Purity of the final nucleoside derivatives was checked using a Hewlett−Packard 1100 HPLC equipped with a Zorbax SB-Aq 5 μm analytical column (50 × 4.6 mm; Agilent Technologies Inc., Palo Alto, CA). Mobile phase: linear gradient solvent system, 5 mM TBAP (tetrabutylammonium dihydrogen phosphate) −CH₃CN from 80:20 to 0:100 in 13 min; the flow rate was 0.5 mL/min. Peaks were detected by UV absorption with a diode array detector at 230, 254, and 280 nm. All derivatives tested for biological activity showed >95% purity by HPLC analysis (detection at 254 nm). Low-resolution mass spectrometry was performed with a JEOL SX102 spectrometer with 6-kV Xe atoms following desorption from a glycerol matrix or on an Agilent LC/MS 1100 MSD, with a Waters (Milford, MA) Atlantis C18 column. High resolution mass spectroscopic (HRMS) measurements were performed on a proteomics optimized Q-TOF-2 (Micromass-Waters) using external calibration with polyalanine, unless noted. Observed mass accuracies are those expected based on known instrument performance as well as trends in masses of standard compounds observed at intervals during the series of measurements. Reported masses are observed masses uncorrected for this time-dependent drift in mass accuracy.

(2R,3R,4R,5R)-2-(6-((Dicyclobutylmethyl)amino)-9H-purin-9-yl)-5-(hydroxymethyl)tetrahydrofuran-3,4-diol (9)

Saturated methanolic ammonia solution (25 mL) was added to compound 69 (425 mg, 0.60 mmol) and the solution stirred overnight at temperature in a sealed tube. Solvent was evaporated and the residue was purified on flash silica gel column chromatography (CH₂Cl₂:MeOH = 25:1) to give compound 9 (220 mg, 93%) as colorless powder. ¹H NMR (CD₃OD, 400 MHz) δ 8.28 (s, 1H), 8.20 (s, 1H), 5.97 (d, J = 6.4 Hz, 1H), 4.77 (t, J = 5.6 Hz, 1H), 4.41 (t, J = 7.2 Hz, 1H), 4.34 (d, J = 4.8 Hz, 1H), 4.19 (d, J = 6.0 Hz, 1H), 3.92 (d, J₁ = 2.4 Hz, J₂ = 12.4 Hz, 1H), 3.78 (d, J₁ = 2.4 Hz, J₂ = 12.4 Hz, 1H), 2.55–2.51 (m, 2H), 2.02–1.83 (m, 10H), 1.79–1.75 (m, 2H). HRMS calc. C₁₉H₂₈N₅O₄ (M + H)⁺: 390.2141; found 390.2144.

(2S,3S,4R,5R)-2-(Chloromethyl)-5-(6-((dicyclobutylmethyl)amino)-9H-purin-9-yl)tetrahydrofuran-3,4-diol (10)

Compound 10 (89%) was prepared from compound 72 following the same method as for compound 24. ¹H NMR (CD₃OD, 400 MHz) δ 8.27 (s, 1H), 8.24 (s, 1H), 6.05 (d, J = 5.2 Hz, 1H), 4.79 (d, J = 5.2 Hz, 1H), 4.41–4.39 (m, 2H), 4.31–4.27 (d, J₁ = 4.8 Hz, J₂ = 5.2 Hz, 1H), 3.97 (d, J₁ = 5.2 Hz, J₂ = 6.8 Hz, 1H), 3.87 (d, J₁ = 5.2 Hz, J₂ = 6.8 Hz, 1H), 2.57–2.50 (m, 2H), 2.01–1.83 (m, 10H), 1.79–1.75 (m, 2H). HRMS calc. C₁₉H₂₇N₅O₄Cl (M + H)⁺: 408.1802; found 408.1800.

(2R,3R,4S,5S)-2-(6-((Dicyclobutylmethyl)amino)-9H-purin-9-yl)-5-((ethylthio)methyl)tetrahydrofuran-3,4-diol (11)

Compound 11 (91%) was prepared from compound 73 following the same method as for compound 24. ¹H NMR (CD₃OD, 400 MHz) δ 8.30 (s, 1H), 8.23 (s, 1H), 6.02 (d, J = 5.2 Hz, 1H), 4.78 (t, J = 6.0 Hz, 1H), 4.39–4.33 (m, 2H), 4.24–4.21 (m, 1H), 3.01 (d, J₁ = 5.6 Hz, J₂ = 8.8 Hz, 1H), 2.93 (d, J₁ = 5.6 Hz, J₂ = 8.8 Hz, 1H), 2.63–2.48 (m, 4H), 2.01–1.83 (m, 10H), 1.76–1.75 (m, 2H), 1.22 (t, J = 7.6 Hz, 3H). HRMS calc. C₂₁H₃₂N₅O₃S (M + H)⁺: 434.2226; found 434.2232.

(2R,3R,4R,5R)-2-(2-Chloro-6-((dicyclobutylmethyl)amino)-9H-purin-9-yl)-5-(hydroxymethyl)tetrahydrofuran-3,4-diol (12)

Compound 12 (91%) was prepared from compound 70 following the same method as for compound 9. ¹H NMR (CD₃OD, 400 MHz) δ 8.25 (s, 1H), 5.92 (d, J = 6.0 Hz, 1H), 4.70 (t, J = 5.2 Hz, 1H), 4.40–4.32 (m, 2H), 4.17 (d, J = 6.8 Hz, 1H), 3.92 (d, J₁ = 2.4 Hz, J₂ = 10.0 Hz, 1H), 3.78 (d, J₁ = 2.4 Hz, J₂ = 10.0 Hz, 1H), 2.55–2.47 (m, 2H), 2.01–1.84 (m, 10H), 1.77–1.76 (m, 2H). HRMS calc. C₁₉H₂₇N₅O₄Cl (M + H)⁺: 424.1752; found 424.1750.

(1R,2R,3S,4R,5S)-4-(6-((Dicyclobutylmethyl)amino)-9H-purin-9-yl)-1-(hydroxymethyl)bicyclo[3.1.0]hexane-2,3-diol (24)

A solution of compound 60 (30 mg, 0.02 mmol) in methanol (3 mL) and 10% TFA in water (3 mL) was heated at 70 °C for 6hrs. Solvent was evaporated under vacuum and the residue was purified on flash silica gel silica chromatography (CH₂Cl₂:MeOH = 25:1) to give compound 24 (23 mg, 86%) as a syrup. ¹H NMR (CD₃OD, 400 MHz) δ 8.47 (s, 1H), 8.22 (s, 1H), 4.82 (d, J = 6.4 Hz, 1H), 4.40 (t, J = 6.8 Hz, 1H), 4.31 (d, J = 11.6 Hz, 1H), 3.91 (d, J = 6.4 Hz, 1H), 3.33 (d, J = 11.6 Hz, 1H), 2.54–2.50 (m, 2H), 1.98–1.83 (m, 10H), 1.76–1.75 (m, 2H), 1.68–1.65 (m, 1H), 1.54 (t, J = 5.2 Hz, 1H), 0.78–0.75 (m, 1H). HRMS calc. C₂₁H₃₀N₅O₃ (M + H)⁺: 400.2349; found 400.2356.

(1R,2R,3S,4R,5S)-4-(2-Chloro-6-((dicyclobutylmethyl)amino)-9H-purin-9-yl)-1-(hydroxymethyl)bicyclo[3.1.0]hexane-2,3-diol (25)

Compound 25 (87%) was prepared from compound 61 following the same method as for compound 24. ¹H NMR (CD₃OD, 400 MHz) δ 8.44 (s, 1H), 4.81 (s, 1H), 4.79 (d, J = 6.8 Hz, 1H), 4.37 (d, J = 8.0 Hz, 1H), 4.29 (d, J = 11.6 Hz, 1H), 3.90 (d, J = 6.4 Hz, 1H), 2.53–2.47 (m, 2H), 1.99–1.88 (m, 10H), 1.84–1.82 (m, 2H), 1.65–1.62 (m, 1H), 1.56 (d, J = 4.8 Hz, 1H), 0.78–0.75 (m, 1H). HRMS calc. C₂₁H₂₉N₅O₃Cl (M + H)⁺: 434.1959; found 434.1964.

(1S,2R,3S,4R,5S)-1-(Chloromethyl)-4-(6-((dicyclobutylmethyl)amino)-9H-purin-9-yl)bicyclo[3.1.0]hexane-2,3-diol (26)

Compound 26 (88%) was prepared from compound 63 following the same method as for compound 24. ¹H NMR (CD₃OD, 400 MHz) δ 8.36 (s, 1H), 8.25 (s, 1H), 4.83 (d, J = 6.0 Hz, 1H), 4.40 (t, J = 7.2 Hz, 1H), 4.33 (d, J = 11.6 Hz, 1H), 4.01 (d, J = 6.4 Hz, 1H), 3.65 (d, J = 11.6 Hz, 1H), 2.54–2.51 (m, 2H), 1.99–1.83 (m, 10H), 1.79–1.75 (m, 4H), 0.99–0.95 (m, 1H). HRMS calc. C₂₁H₂₉N₅O₂Cl (M + H)⁺: 418.2010; found 418.2016.

(1S,2R,3S,4R,5S)-4-(2-Chloro-6-((dicyclobutylmethyl)amino)-9H-purin-9-yl)-1-(chloromethyl)bicyclo[3.1.0]hexane-2,3-diol (27)

Compound 27 (89%) was prepared from compound 64 following the same procedure as for compound 24. ¹H NMR (CD₃OD, 400 MHz) δ 8.29 (s, 1H), 4.83–4.80 (m, 2H), 4.38 (t, J = 8.0 Hz, 1H), 4.31 (d, J = 11.6 Hz, 1H), 3.99 (d, J = 6.8 Hz, 1H), 3.69 (d, J = 11.6 Hz, 1H), 2.53–2.47 (m, 2H), 1.99–1.88 (m, 10H), 1.84–1.74 (m, 4H), 0.99–0.95 (m, 1H). HRMS calc. C₂₁H₂₈N₅O₂Cl₂ (M + H)⁺: 452.1620; found 452.1624.

(1S,2R,3S,4R,5S)-4-(2-Chloro-6-((dicyclobutylmethyl)amino)-9H-purin-9-yl)-1-(((2-fluorophenyl)thio)methyl)bicyclo[3.1.0]hexane-2,3-diol (28)

Compound 28 (90%) was prepared from compound 66 following the same procedure as for compound 24. ¹H NMR (CD₃OD, 400 MHz) δ 8.13 (s, 1H), 7.49–7.45 (m, 1H), 7.20–7.15 (m, 1H), 7.07–7.02 (m, 2H), 4.84 (d, J = 6.0 Hz, 1H), 4.71 (s, 1H), 4.38 (t, J = 7.6 Hz, 1H), 4.05 (d, J = 6.8 Hz, 1H), 3.69 (d, J = 13.6 Hz, 1H), 3.31 (d, J = 13.6 Hz, 1H), 2.54–2.48 (m, 2H), 2.02–1.89 (m, 10H), 1.84–1.77 (m, 2H), 1.59–1.54 (m, 2H), 0.91–0.87 (m, 1H). HRMS calc. C₂₇H₃₂N₅O₂ClSF (M + H)⁺: 544.1949; found 544.1951.

(1R,2R,3S,4R,5S)-4-(6-(((1R,2S,4S)-Bicyclo[2.2.1]heptan-2-yl)amino)-2-chloro-9H-purin-9-yl)-1-(hydroxymethyl)bicyclo[3.1.0]hexane-2,3-diol (29)

Compound 29 (84%) was prepared from compound 62 following the same method as for compound 24. ¹H NMR (CD₃OD, 400 MHz) δ 8.44 (s, 1H), 4.80 (s, 1H), 4.77 (d, J = 6.8 Hz, 1H), 4.29 (d, J = 11.6 Hz, 1H), 4.30 (br s, 1H), 3.88 (d, J = 6.8 Hz, 1H), 3.38 (d, J = 11.6 Hz, 1H), 2.35–2.34 (m, 2H), 1.93–1.87 (m, 1H), 1.64–1.53 (m, 5H), 1.49–1.34 (m, 2H), 1.27–1.21 (m, 2H), 0.78–0.74 (m, 1H). HRMS calc. C₁₉H₂₅N₅O₃Cl (M + H)⁺: 406.1646; found 406.1647.

(1S,2R,3S,4R,5S)-4-(6-(((1R,2S,4S)-Bicyclo[2.2.1]heptan-2-yl)amino)-2-chloro-9H-purin-9-yl)-1-(chloromethyl)bicyclo[3.1.0]hexane-2,3-diol (31)

Compound 31 (93%) was prepared from compound 65 following the same procedure as for compound 24. ¹H NMR (CD₃OD, 400 MHz) δ 8.28 (s, 1H), 4.81–4.79 (m, 2H), 4.31 (d, J = 11.6 Hz, 1H), 4.03 (br s, 1H), 3.97 (d, J = 6.8 Hz, 1H), 3.68 (d, J = 11.6 Hz, 1H), 2.38–2.32 (m, 2H), 1.92–1.87 (m, 1H), 1.82–1.79 (m, 1H), 1.75 (t, J = 4.8 Hz, 1H), 1.61–1.55 (m, 3H), 1.48–1.34 (m, 2H), 1.30–1.23 (m, 2H), 0.98–0.96 (m, 1H). HRMS calc. C₁₉H₂₄N₅O₂Cl₂ (M + H)⁺: 424.1307; found 424.1314.

9-((3aR,3bR,4aS,5R,5aS)-3b-(((tert-Butyldiphenylsilyl)oxy)methyl)-2,2-dimethylhexahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-5-yl)-N-(dicyclobutylmethyl)-9H-purin-6-amine (57)

Dicyclobutylmethylamine (302 mg, 2.17 mmol) and DIPEA (0.75 mL, 4.37 mmol) were added to a solution of compound 55 (250 mg, 0.43 mmol) in 2-propanol and the solution heated at 75 °C overnight. Solvent was evaporated under vacuum and the residue was purified on flash silica gel column chromatography (hexane:ethyl acetate = 3:1) to give compound 57 (267 mg, 91%) as a colorless syrup. ¹H NMR (CD₃OD, 400 MHz) δ 8.36 (s, 1H), 8.19 (s, 1H), 7.67–7.644 (m, 4H), 7.44–7.33 (m, 6H), 5.40 (d, J = 7.2 Hz, 1H), 5.04 (s, 1H), 4.72 (d, J = 6.8 Hz, 1H), 4.40 (t, J = 7.2 Hz, 1H), 4.23 (d, J = 10.8 Hz, 1H), (d, J = 10.8 Hz, 1H), 2.54–2.50 (m, 2H), 1.99–1.88 (m, 10H), 1.76–1.75 (m 2H), 1.66–1.65 (m, 1H), 1.51 (s, 3H), 1.25 (s, 3H), 1.16 (t, J = 4.8 Hz, 1H), 1.10 (s, 9H), 0.92–0.88 (m, 1H). HRMS calc. C₄₀H₅₂N₅O₃Si (M + H)⁺: 678.3839; found 678.3832.

9-((3aR,3bR,4aS,5R,5aS)-3b-(((tert-Butyldiphenylsilyl)oxy)methyl)-2,2-dimethylhexahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-5-yl)-2-chloro-N-(dicyclobutylmethyl)-9H-purin-6-amine (58)

Compound 58 (87%) was prepared from compound 56 following the same method as for compound 57.¹H NMR (CD₃OD, 400 MHz) δ 8.27 (s, 1H), 7.66 (d, J = 6.8 Hz, 4H), 7.44–7.31 (m, 6H), 5.35 (d, J = 7.2 Hz, 1H), 4.95 (s, 1H), 4.71 (d, J = 7.2 Hz, 1H), 4.39 (t, J = 8.0 Hz, 1H), 4.23 (d, J = 6.4 Hz, 1H), 3.76 (d, J = 10.8 Hz, 1H), 2.54–2.50 (m, 2H), 2.01–1.89 (m, 10H), 1.83–1.76 (m, 2H), 1.63–1.60 (m, 1H), 1.52 (s, 3H), 1.26 (s, 3H), 1.14–1.11 (m, 1H), 1.08 (s, 9H), 0.98–0.93 (m, 1H). HRMS calc. C₄₀H₅₁N₅O₃SiCl (M + H)⁺: 712.3450; found 712.3445.

N-((1R,2S,4S)-Bicyclo[2.2.1]heptan-2-yl)-9-((3aR,3bR,4aS,5R,5aS)-3b-(((tert-butyldiphenylsilyl)oxy)methyl)-2,2-dimethylhexahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-5-yl)-2-chloro-9H-purin-6-amine (59)

(1R, 2S, 4S)-Bicyclo[2.2.1]nonane (306 mg, 2.75 mmol) and triethylamine (0.76 mL, 5.51 mmol) were added to a solution of compound 56 (336 mg, 0.55 mmol) in methanol and the solution stirred at room temperature overnight. Solvent was evaporated under vacuum and the residue was purified on flash silica gel column chromatography (hexane:ethyl acetate = 4:1) to give compound 59 (319 mg, 82%) as a colorless syrup. ¹H NMR (CD₃OD, 400 MHz) δ 8.26 (s, 1H), 7.66–7.64 (m, 4H), 7.43–7.32 (m, 6H), 5.34 (d, J = 6.8 Hz, 1H), 4.95 (s, 1H), 4.69 (d, J = 7.2 Hz, 1H), 4.22 (d, J = 6.8 Hz, 1H), 4.03 (br s, 1H), 3.76 (d, J = 6.8 Hz, 1H), 2.36–2.34 (m, 2H), 1.94–1.87 (m, 1H), 1.62–1.59 (m, 4H), 1.52 (s, 3H), 1.48–1.36 (m, 2H), 1.29–1.24 (m, 5H), 1.12 (t, J = 5.2 Hz, 1H), 1.09 (s, 9H), 0.98–0.94 (m, 1H). HRMS calc. C₃₈H₄₇N₅O₃SiCl (M + H)⁺: 684.3137; found 684.3143.

((3aR,3bR,4aS,5R,5aS)-5-(6-((Dicyclobutylmethyl)amino)-9H-purin-9-yl)-2,2-dimethyltetrahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-3b(3aH)-yl)methanol (60)

TBAF (0.6 mL, 1M solution in THF) was added to a solution of compound 57 (267 mg, 0.39 mmol) in THF and the solution stirred for 1 h at room temperature. Solvent was evaporated and the residue was purified on silica gel column chromatography (CH₂Cl₂:MeOH = 30:1) to give compound 60 (161 mg, 93%) as a colorless syrup. ¹H NMR (CD₃OD, 400 MHz) δ 8.29 (s, 1H), 8.21 (s, 1H), 5.43 (d, J = 7.2 Hz, 1H), 5.03 (s, 1H), 4.69 (d, J = 7.2 Hz, 1H), 4.40 (t, J = 6.4 Hz, 1H), 4.11 (d, J = 10.8 Hz, 1H), 3.46 (d, J = 10.8 Hz, 1H), 2.55–2.51 (m, 2H), 1.98–1.82 (m, 10H), 1.79–1.76 (m, 3H), 1.53 (s, 3H), 1.25 (s, 3H), 1.19 (t, J = 5.2 Hz, 1H), 1.00–0.96 (m, 1H). HRMS calc. C₂₄H₃₄N₅O₃ (M + H)⁺: 440.2662; found 440.2662.

((3aR,3bR,4aS,5R,5aS)-5-(2-Chloro-6-((dicyclobutylmethyl)amino)-9H-purin-9-yl)-2,2-dimethyltetrahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-3b(3aH)-yl)methanol (61)

Compound 61 (89%) was prepared from compound 58 following the same method as for compound 60. ¹H NMR (CD₃OD, 400 MHz) δ 8.22 (s, 1H), 5.39 (d, J = 7.2 Hz, 1H), 4.95 (s, 1H), 4.71 (d, J = 6.8 Hz, 1H), 4.38 (t, J = 8.0 Hz, 1H), 4.01 (d, J = 11.6 Hz, 1H), 3.63 (d, J = 11.6 Hz, 1H), 2.53–2.47 (m, 2H), 1.99–1.84 (m, 10H), 1.78–1.69 (m, 3H), 1.52 (s, 3H), 1.26 (s, 3H), 1.15 (t, J = 4.8 Hz, 1H), 1.00–0.97 (m, 1H). HRMS calc. C₂₄H₃₃N₅O₃Cl (M + H)⁺: 474.2272; found 474.2274.

((3aR,3bR,4aS,5R,5aS)-5-(6-(((1R,2S,4S)-Bicyclo[2.2.1]heptan-2-yl)amino)-2-chloro-9H-purin-9-yl)-2,2-dimethyltetrahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-3b(3aH)-yl)methanol (62)

Compound 62 (91%) was prepared from compound 59 following the same method as for compound 60. ¹H NMR (CD₃OD, 400 MHz) δ 8.21 (s, 1H), 5.38 (d, J = 6.4 Hz, 1H), 4.95 (s, 1H), 4.69 (d, J = 7.2 Hz, 1H), 4.00–3.97 (m, 2H), 3.62 (d, J = 12.0 Hz, 1H), 2.35–2.33 (m, 2H), 1.99–1.87 (m, 1H), 1.71–1.67 (m, 1H), 1.61–1.58 (m, 3H), 1.52 (s, 3H), 1.48–1.33 (m, 2H), 1.31–1.22 (m, 5H), 1.14 (t, J = 4.8 Hz, 1H), 0.99–0.95 (m, 1H). HRMS calc. C₂₂H₂₉N₅O₃Cl (M + H)⁺: 446.1959; found 446.1953.

9-((3aR,3bS,4aS,5R,5aS)-3b-(Chloromethyl)-2,2-dimethylhexahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-5-yl)-N-(dicyclobutylmethyl)-9H-purin-6-amine (63)

Thionyl chloride (0.048 mL, 0.39 mL) was added dropwise to a solution of compound 60 (58 mg, 0.13 mmol) in dry CH₃CN at −5 °C followed by pyridine (32 L, 0.39 mmol) and the solution stirred for 30 min in the same condition. Then the reaction mixture was brought to room temperature and stirred overnight. Water was added into the reaction and the reaction mixture was neutralized with 1M NaHCO₃ solution. The aqueous layer was extracted with CH₂Cl₂ (3 times), dried over Na₂SO₄, filtered and evaporated. The residue was purified by flash silica gel column chromatography (hexane:ethyl acetate = 2:1) to afford the compound 63 (37 mg, 62%) as a colorless syrup. ¹H NMR (CD₃OD, 400 MHz) δ 8.39 (s, 1H), 8.37 (s, 1H), 5.41 (d, J = 7.2 Hz, 1H), 5.09 (s, 1H), 4.83 (d, J = 6.8 Hz, 1H), 4.32 (t, J = 6.8 Hz, 1H), 3.93 (d, J = 10.4 Hz, 1H), 2.64–2.60 (m, 2H), 2.03–1.85 (m, 10H), 1.80–1.78 (m, 3H), 1.54 (s, 3H), 1.36 (t, J = 5.2 Hz, 1H), 1.27 (s, 3H), 1.19–1.15 (m, 1H). HRMS calc. C₂₄H₃₃N₅O₂Cl (M + H)⁺: 458.2323; found 458.2322.

2-Chloro-9-((3aR,3bS,4aS,5R,5aS)-3b-(chloromethyl)-2,2-dimethylhexahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-5-yl)-N-(dicyclobutylmethyl)-9H-purin-6-amine (64)

Compound 64 (65%) was prepared from compound 61 following the same method as for compound 63. ¹H NMR (CD₃OD, 400 MHz) δ 8.15 (s, 1H), 5.38 (d, J = 7.2 Hz, 1H), 4.94 (s, 1H), 4.78 (d, J = 6.8 Hz, 1H), 4.37 (t, J = 8.0 Hz, 1H), 4.15 (d, J = 11.6 Hz, 1H), 3.79 (d, J = 11.6 Hz, 1H), 2.55–2.49 (m, 2H), 1.99–1.86 (m, 10H), 1.84–1.74 (m, 3H), 1.53 (s, 3H), 1.30–1.27 (m, 4H), 1.14–1.10 (m, 1H). HRMS calc. C₂₄H₃₂N₅O₂Cl (M + H)⁺: 492.1933; found 492.1935.

N-((1R,2S,4S)-Bicyclo[2.2.1]heptan-2-yl)-2-chloro-9-((3aR,3bS,4aS,5R,5aS)-3b-(chloromethyl)-2,2-dimethylhexahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-5-yl)-9H-purin-6-amine (65)

Compound 65 (63%) was prepared from compound 62 following the same method as for compound 63. ¹H NMR (CD₃OD, 400 MHz) δ 8.14 (s, 1H), 5.36 (d, J = 7.2 Hz, 1H), 4.94 (s, 1H), 4.77 (d, J = 7.2 Hz, 1H), 4.13–4.02 (m, 2H), 3.80 (d, J = 11.6 Hz, 1H), 2.35–2.33 (m, 2H), 1.92–1.86 (m, 1H), 1.81–1.77 (m, 1H), 1.64–1.61 (m, 2H), 1.53 (s, 3H), 1.47–1.33 (m, 2H), 1.30–1.23 (m, 7H), 1.14–1.10 (m, 1H). HRMS calc. C₂₂H₂₈N₅O₂Cl₂ (M + H)⁺: 464.1620; found 464.1614.

2-Chloro-N-(dicyclobutylmethyl)-9-((3aR,3bS,4aS,5R,5aS)-3b-(((2-fluorophenyl)thio)methyl)-2,2-dimethylhexahydrocyclopropa[3,4]cyclopenta[1,2-d][1,3]dioxol-5-yl)-9H-purin-6-amine (66)

NaH (5.6 mg, 0.23 mmol) was added portion wise to an ice-cold solution of 2-fluoro-thiophenol (38 mg, 0.07 mmol) in dry DMF (1 mL). After hydrogen evolution completed, compound 64 (29 mg, 0.05 mmol) was added in dry DMF (0.5 mL) and the solution stirred overnight at room temperature. The reaction mixture was quenched with water, and the aqueous layer was extracted with ethyl acetate (3 times), dried over Na₂SO₄, filtered and evaporated. The residue was purified on flash silica gel column chromatography (hexane:ethyl acetate = 1:1) to give compound 66 (29 mg, 85%) as a syrup. ¹H NMR (CD₃OD, 400 MHz) δ 8.06 (s, 1H), 7.52–7.48 (m, 1H), 7.21–7.16 (m, 1H), 7.06–7.00 (m, 2H), 5.36 (d, J = 6.8 Hz, 1H), 4.81 (s, 1H), 4.79 (d, J = 6.8 Hz, 1H), 4.40 (t, J = 6.8 Hz, 1H), 3.66 (d, J = 12.0 Hz, 1H), 3.29 (d, J = 12.0 Hz, 1H), 2.53–2.50 (m, 2H), 2.02–1.84 (m, 10H), 1.78–1.76 (m, 2H), 1.58–1.52 (m, 1H), 1.52 (s, 3H), 1.26 (m, 3H), 1.11 (t, J = 5.2 Hz, 1H), 0.97–0.93 (m, 1H). HRMS calc. C₃₀H₃₆N₅O₂ClSF (M + H)⁺: 584.2262; found 584.2271.

(2R,3R,4R,5R)-2-((Benzoyloxy)methyl)-5-(6-((dicyclobutylmethyl)amino)-9H-purin-9-yl)tetrahydrofuran-3,4-diyl dibenzoate (69)

Compound 69 (91%) was prepared from compound 67 following the same method as for compound 57. ¹H NMR (CD₃OD, 400 MHz) δ 8.26 (s, 1H), 8.08 (s, 1H), 8.06–7.93 (m, 6H), 7.64–7.57 (m, 3H), 7.48–7.37 (m, 6H), 6.54–6.49 (m, 2H), 6.36 (t, J = 5.6 Hz, 1H), 4.92–4.91 (m, 2H), 4.73 (dd, J₁ = 3.2 Hz, J₂ = 8.0 Hz, 1H), 4.37 (br s, 1H), 2.55–2.49 (m, 2H), 1.97–1.85 (m, 10H), 1.83–1.76 (m, 2H). HRMS calc. C₄₀H₄₀N₅O₇ (M + H)⁺: 702.2928; found 702.2919.

(2R,3R,4R,5R)-2-((Benzoyloxy)methyl)-5-(2-chloro-6-((dicyclobutylmethyl)amino)-9H-purin-9-yl)tetrahydrofuran-3,4-diyl dibenzoate (70)

Compound 70 (89%) was prepared from compound 68 following the same method as for compound 57. ¹H NMR (CD₃OD, 400 MHz) δ 8.21 (s, 1H), 8.04–7.94 (m, 6H), 7.63–7.57 (m, 3H), 7.46–7.37 (m, 6H), 6.50 (d, J = 4.0 Hz, 1H), 6.35–6.28 (m, 2H), 4.75–4.71 (m, 1H), 4.36 (t, J = 8.0 Hz, 1H), 2.51–2.46 (m, 2H), 1.98–1.87 (m, 10H), 1.82–1.75 (m, 2H). HRMS calc. C₄₀H₃₉N₅O₇Cl (M + H)⁺: 736.2538; found 736.2526.

((3aR,4R,6R,6aR)-6-(6-((Dicyclobutylmethyl)amino)-9H-purin-9-yl)-2,2-dimethyltetrahydrofuro[3,4-d][1,3]dioxol-4-yl)methanol (71)

2,2-Dimethoxypropane (0.31 mL, 2.57 mmol) and p-toluenesulfonic acid (97 mg, 0.51 mmol) were added to a solution of compound 9 (200 mg, 0.51 mmol) in acetone (10 mL) and stirred for 5 h at room temperature. Reaction mixture was neutralized with NaHCO₃, filtered and evaporated. The residue was purified on flash silica gel column chromatography (hexane:ethyl acetate = 1:2) to give compound 71 (211 mg, 96%) as a colorless syrup. ¹H NMR (CD₃OD, 400 MHz) δ 8.13 (s, 2H), 5.38 (d, J= 6.8 Hz, 1H), 4.98 (s, 1H), 4.73 (d, J = 6.8 Hz, 1H), 4.43–4.33 (m, 2H), 3.84–3.67 (m, 2H), 2.08–1.98 (m, 2H), 1.96– 1.72 (m,10H), 1.73–1.69 (m, 2H), 1.52 (s, 3H), 1.26 (s, 3H). HRMS calc. C₂₂H₃₂N₅O₄ (M + H)⁺: 430.2454; found 430.2450.

9-((3aR,4R,6S,6aS)-6-(Chloromethyl)-2,2-dimethyltetrahydrofuro[3,4-d][1,3]dioxol-4-yl)-N-(dicyclobutylmethyl)-9H-purin-6-amine (72)

Compound 72 (67%) was prepared from compound 71 following the same method as for compound 63. ¹H NMR (CD₃OD, 400 MHz) δ 8.25 (s, 2H), 6.24 (d, J = 2.4 Hz, 1H), 5.50 (d, J₁ = 2.4 Hz, J₂ = 4.4 Hz, 1H), 5.16 (d, J₁ = 2.4 Hz, J₂ = 6.0 Hz, 1H), 4.45–4.37 (m, 2H), 3.84 (d, J₁ = 4.4 Hz, J₂ = 7.2 Hz, 1H), 3.70 (d, J₁ = 4.4 Hz, J₂ = 7.2 Hz, 1H), 2.54–2.50 (m, 2H), 2.02–1.83 (m, 10H), 1.76–1.75 (m, 2H), 1.61 (s, 3H), 1.40 (s, 3H). HRMS calc. C₂₂H₃₁N₅O₃Cl (M + H)⁺: 448.2115; found 448.2113.

N-(Dicyclobutylmethyl)-9-((3aR,4R,6S,6aS)-6-((ethylthio)methyl)-2,2-dimethyltetrahydrofuro[3,4-d][1,3]dioxol-4-yl)-9H-purin-6-amine (73)

Sodium thioethoxide (27.8 mg, 0.33 mmol) was added to a solution of compound 72 (37 mg, 0.08 mmol) in dry DMF (1 mL) at 0 °C and the solution stirred at room temperature for 1 h. The reaction mixture was quenched with water, and the aqueous layer was extracted with ethyl acetate (3 times), dried over Na₂SO₄, filtered and evaporated. The residue was purified on flash silica gel column chromatography (hexane:ethyl acetate = 2:1) to give compound 73 (34 mg, 87%) as a syrup. ¹H NMR (CD₃OD, 400 MHz) δ 8.26 (s, 1H), 8.25 (s, 1H), 6.19 (d, J = 2.4 Hz, 1H), 5.56 (d, J₁ = 1.6 Hz, J₂ = 4.8 Hz, 1H), 5.08 (m, 1H), 4.39–4.26 (m, 2H), 2.82–2.80 (m, 2H), 2.55–2.46 (m, 4H), 2.01–1.83 (m, 10H), 1.77–1.75 (m, 2H), 1.60 (s, 3H), 1.40 (s, 3H), 1.14 (t, J = 7.2 Hz, 3H). HRMS calc. C₂₄H₃₆N₅O₃S (M + H)⁺: 474.2539; found 474.2541.

Pharmacological characterization

Functional assay at rA₁AR was performed by GVK Biosciences, Hyderbad, India (Study No: 050–13-IVP). rA₁AR membranes (Perkin Elmer Rat A1: 6110511400UA) were pretreated with adenosine deaminase at 1U/mL. 20 μl of 1 nM [³⁵S]GTPγS was added to all wells of a 96-well plate. 20 μl of 10 μM GDP, 20 μl of membrane preparations (5 μg membranes/well) and 20 μl of PVT SPA (PolyVinyl Toluene Scintillation Proximity Assay) beads (1 μg/well) were added, and the total assay volume was made up to 100 μl by adding assay buffer (20 mM HEPES pH 7.4, 100 mM NaCl, 1 mM MgCl₂). The reaction mixture was incubated for 60 min at room temperature. Post-incubation, the plate was centrifuged at 2000 rpm for 10 min, and radioactivity was measured.

Functional assays at hA₁AR, mA₁AR and mA₃AR were performed as described.⁴¹ In the hA₁AR-mediated stimulation of [³⁵S]GTPγS binding and inhibition of forskolin-stimulated cAMP production, an antagonist of the A_2BAR (8-[4-[4-(4-chlorophenzyl)piperazide-1-sulfonyl)phenyl]]-1-propylxanthine, 1 μM) was added to prevent activation of the endogenous receptor in HEK cells.

Body temperature.

All animal experiments performed in the manuscript were conducted in compliance with institutional guidelines. Tb and activity were measured in male C57BL/6J mice and A₁AR, A₃AR, and A₁AR/A₃AR deficient mice continuously by telemetry as previously described.³⁰ The following compounds (vehicle) were administered ip.: compounds 4, 10, 29 and 31 (10% DMSO); compound 1a (saline); compound 7 (10% DMSO/27% PEG400); compounds 9, 12, 24, 25, 26, 27 and 40 (15% DMSO/15% Kolliphor EL/70% saline).

Central infusions:

Mice were anesthetized with ketamine/xylazine (80/10 mg/kg, ip.). Sterile guide cannulas (5.25 mm, 26 gauge, Plastics One, Roanoke, VA) were unilaterally implanted into the lateral ventricle (coordinates relative to bregma: −0.34 mm anterior, 1.0 mm lateral, and +1.7 mm ventral) and fixed with dental cement (Parkell, Edgewood, NY). Compounds in 5 μl were infused (0.5 μl/min) through a 33-gauge cannula protruding 0.5 mm past the tip of the guide cannula using PE-50 tubing fitted to a 5 μl syringe (Hamilton, Reno, NV).

Molecular Modeling

Ligand-Protein Complex Preparation

The cryo-EM structure of A₁AR was retrieved from the Protein Data Bank (PDB ID: 6D9H)⁵⁵ and prepared using the Protein Preparation Wizard tool⁵⁷ of the Schrödinger suite (Maestro 2018–1).⁵⁸ His78/251/278 were protonated on the N^δ nitrogen (named HSD according to the CHARMM nomenclature), while His264 was protonated on N^ε (HSE). The G_i protein was removed from the complex before proceeding with the simulations.

Residues Glu172 (EL2), Met177 (5.37), Thr257 (6.58), Lys265 (EL3) and Thr270 (7.34) were mutated into alanines, and the resulting receptor was used for docking simulations. A water molecule was copied from the A_2AAR 2YDO⁵⁹ structure (RESID 2017) and minimized with OPLS3⁶⁰ force field. Compound 9 was docked to the receptor using Glide^61,62 docking software of the Schrödinger suite,⁵⁸ with XP scoring function⁶³ and constraining the compound core to the adenosine scaffold. Glu172 (EL2), Met177 (5.37), Thr257 (6.58), Lys265 (EL3) and Thr270 (7.34) were restored from the cryo-EM structure and finally minimized in presence of the docked compound with OPLS3 force field.⁶⁰ The conformation of compound 24 was obtained by flexible alignment to compound 9 selected pose.

Molecular Dynamics

The HTMD (version 5.2.0)⁶⁴ module was used to prepare the systems for MD simulations. An 80×80 Å 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipid bilayer was obtained using the VMD Membrane Plugin tool.⁶⁵ After computing the orientation of the A₁AR structure in a membrane through the Positioning of Proteins in Membrane (PPM) web server,⁶⁶ the ligand-protein complexes were embedded into the membrane leaflet and the overlapping lipids (within 1.3 Å) were removed. The systems were solvated with TIP3P⁶⁷ water molecules, resulting in a simulation box of 80×80×120 Å, and Na⁺/Cl⁻ counter-ions were added to neutralize the system in a 0.154 M concentration. The simulations were run using the ACEMD⁶⁸ MD engine with CHARMM36^69,70 force field for protein, lipids, water and ions, and CGenFF (3.0.1)^71,72 for the ligands. Missing ligands parameters were assigned by analogy using the ParamChem⁷³ web service (version 1.0.0), with few modifications: adenosine parameters were used for the adenine moiety, and the dihedral angle C5-C6-N6-C1 (Figure S9) was optimized by means of the HTMD⁶⁴ Parametrize tool, using QM calculations as target data (Hartree-Fock (HF) theory and 6–31G* basis set) and Psi4 code (version 1.2.1).⁷⁴

The systems were subjected to an equilibration phase involving 5000 conjugate-gradient minimization steps and 40 ns MD simulation in the NPT ensemble, by applying initial positional harmonic restraints to ligand and protein atoms (0.8 kcal mol⁻¹ Å⁻² for ligand atoms, 0.85 kcal mol⁻¹ Å⁻² for Cα carbon atoms, and 0.4 kcal mol⁻¹ Å⁻² for the other protein atoms), linearly reduced in the last 20 ns. A Berendsen barostat (relaxation time 800 fs) was used to maintain the pressure at 1 atm. The equilibrated systems were subjected to three 30 ns MD simulations in the NVT ensemble each. In both equilibration and productive simulations, the temperature was maintained at 310K through a Langevin thermostat (with damping constant 1 ps⁻¹ and 0.1 ps⁻¹ in the case of equilibration and production, respectively). A timestep of 2 fs was employed and the M-SHAKE⁷⁵ algorithm was used to constrain bonds containing hydrogen atoms. A cutoff of 9 Å was used for non-bonded interactions, with a switching distance of 7.5 Å. The Particle Mesh Ewald (PME)⁷⁶ method was used to compute long-range electrostatic interactions beyond the non-bonded cutoff, with grid spacing of 1Å.

Trajectory Analysis

Trajectories were aligned by superposing protein Cα carbon atoms to the initial conformation and were analyzed in terms of ligand heavy atoms RMSD, ligand-protein electrostatic and van der Waals interactions, number of H-bonds and number of contacts between protein and ligand. Analysis were performed through an in-house Tcl script employing VMD⁶² text commands. Ligand-protein electrostatic and van der Waals interactions were computed with NAMD.⁷⁷ All the data were plotted using the Gnuplot (version 5.0) software⁷⁸ and figures were produced using Chimera.⁷⁹

Abbreviations:

AR

adenosine receptor

HEK 293

human embryonic kidney 293

DAT

the dopamine transporter

DMF

dimethylformamide

DIPEA

diisopropylethylamine

GPCR

G protein-coupled receptor

HRMS

high resolution mass spectrometry

MD

molecular dynamics

NET

the norepinephrine transporter

PDSP

NIMH Psychoactive Drug Screening Program

RMSD

root-mean-square deviation

SAR

structure activity relationship

SERT

the serotonin transporter

Tb

body temperature

TBAF

tetrabutylammonium fluoride

TBAP

tetrabutylammonium dihydrogenphosphate

TFA

trifluoroacetic acid

THF

tetrahydrofuran

TM

transmembrane

TSPO

the translocator protein