Exploring distal regions of the A₃ adenosine receptor binding site: sterically constrained N⁶-(2-phenylethyl)adenosine derivatives as potent ligands

Abstract

We synthesized phenyl ring-substituted analogues of N⁶-(1S,2R)-(2-phenyl-1-cyclopropyl)adenosine, which is highly potent in binding to the human A₃AR with a K_i value of 0.63 nM. The effects of these structural changes on affinity at human and rat adenosine receptors and on intrinsic efficacy at the hA₃AR were measured. A 3-nitrophenyl analogue was resolved chromatographically into pure diastereomers, which displayed 10-fold stereoselectivity in A₃AR binding in favor of the 1S,2R isomer. A molecular model defined a hydrophobic region (Phe168) in the putative A₃AR binding site around the phenyl moiety. A heteroaromatic group (3-thienyl) could substitute for the phenyl moiety with retention of high affinity of A₃AR binding. Other related N⁶-substituted adenosine derivatives were included for comparison. Although the N⁶-(2-phenyl-1-cyclopropyl) derivatives were full A₃AR agonists, several other derivatives had greatly reduced efficacy. N⁶-Cyclopropyladenosine was an A₃AR antagonist, and adding either one or two phenyl rings at the 2-position of the cyclopropyl moiety restored efficacy. N⁶-(2,2-Diphenylethyl)adenosine was an A₃AR antagonist, and either adding a bond between the two phenyl rings (N⁶-9-fluorenylmethyl) or shortening the ethyl moiety (N⁶-diphenylmethyl) restored efficacy. A QSAR study of the N⁶ region provided a model that was complementary to the putative A₃AR binding site in a rhodopsin-based homology model. Thus, a new series of high-affinity A₃AR agonists and related nucleoside antagonists was explored through both empirical and theoretical approaches.

1. Introduction

The adenosine receptors (ARs) consist of four subtypes (A₁, A_2A, A_2B, and A₃) and represent a physiologically important family of G protein-coupled receptors.¹ AR agonists are current targets for the development of therapeutic agents for a variety of diseases, including agents with neuroprotective, antiseizure, anti-inflammatory, anti-ischemic, and cardioprotective effects.^2–4 A₃AR agonists are also potentially useful for the treatment of cancer. In certain tumor cells, a cytostatic effect of the A₃AR agonist appears to be related to its downstream activation of the Wnt pathway.⁵

We have studied the microscopic interactions of ligands with the A₃AR and other members of the AR family from the perspectives of both ligand modification and structure–function aspects of the receptors.^6–9 Extensive mutagenesis studies and molecular modeling based on a high-resolution template of rhodopsin have implicated TM (transmembrane domain) regions 3, 6, and 7 in the coordination of adenosine agonists and a putative rotation of TM6 in the activation of the A_2A and A₃ARs.

Adenosine is a nonselective AR agonist that is rapidly degraded in circulation. Its clinical use for treatment of supraventricular tachycardia is predicated on a short duration of action. However, other foreseeable applications of synthetic adenosine agonists would require greater stability in vivo.¹⁰ An early AR agonist showing improved stability toward adenosine deaminase was the nonselective agonist 2-chloroadenosine 1, which also served as the first stable adenosine receptor radioligand.¹¹ The 2-chloro modification has since been incorporated in many potent AR agonists, including the 5′-uronamide derivative Cl-IB-MECA 2, the first highly selective A₃AR agonist. The intrinsic efficacy of various adenosine derivatives as A₃AR agonists was studied, leading to the characterization of structure–activity relationships (SARs) for efficacy, which are distinct from those of affinity.^6,7 The ability of a tightly binding adenosine derivative to activate the A₃AR is highly dependent on ligand conformation. The flexibility of the ribose moiety has been established as a required feature for A₃AR agonists. Steric constraint of the ribose moiety, for example, the introduction of a bridged carbocyclic ring system, tends to reduce intrinsic efficacy of the adenosine derivatives at the human A₃AR (hA₃AR). These effects are overridden by the presence of a flexible 5′-uronamide group.

Highly selective A₃AR agonists thus far reported, for example, 2 (2-chloro-N⁶-(3-iodobenzyl)-5′-N-methylcarboxamidoadenosine), all contain multiple substitutions of the adenosine molecule.¹ A recent study by Gao et al.⁷ identified the singly substituted adenosine derivative compound 3 (N⁶-(1S,2R)-(2-phenyl-1-cyclopropyl) adenosine), with a K_i value in binding to the hA₃AR of 0.63nM (38-fold more potent than the 1R,2S isomer 4), as a lead for the development of N⁶ derivatives of adenosine with high hA₃AR affinity. The present study aimed to identify analogues of 3 and other N⁶ derivatives of adenosine that are selective, while maintaining the remainder of the molecule unchanged, and to categorize structural features of the N⁶ substituents that affect the intrinsic efficacy (Chart 1).

Chart 1.

The structure of the various adenosine agonists studied at the A₃AR.

2. Results

2.1. Chemical synthesis

Most of the analogues of 3 were prepared as diastereomeric mixtures (Schemes 1–3). The phenylcyclopropyl amines 41a–n were prepared from the corresponding trans-cinnamic acid derivatives 37.¹² With the exception of the 3-cyano-cinnamic acid (37k), which was obtained by a Knövenagel condensation between 3-cyano-benzaldehyde and malonic acid, the other acids were commercially available.

Scheme 1.

General synthetic route used to prepare N⁶-(2-phenyl-1-cyclopropyl)adenosine analogues.

Scheme 3.

Synthetic route used to prepare N⁶-(2,2-diphenyl-1-cyclopropyl) adenosine 24.

After esterification with methanol in the presence of H₂SO₄, compounds 38a–n were cyclopropanated with diazomethane in the presence of catalytic Pd(OAc)₂. This is a very high-yield reaction with the only inconvenience being that it must be monitored by NMR, because the starting material and product were undistinguishable by TLC. Hydrolysis of the esters 39a–n gave the desired trans-arylcyclopropanecarboxylic acids 40a–n. The primary amines 41a–n were obtained by a Curtius degradation, in a one-pot reaction with diphenyl phosphorazidate, with the exception of 44 and 45 (Scheme 2). For these two compounds a two-step reaction with sodium azide was preferred, because in this way the resulting acyl azides were transformed into the 2-(trimethylsilyl)ethyl carbamate, which could be hydrolyzed with tetrabutylammonium fluoride without affecting the acetamido moiety. A diphenylcyclopropyl intermediate, the carboxylic acid 47, was prepared by the method reported (Scheme 3).²¹ In this case the cyclopropanation of 3-phenyl-cinnamic acid was unsuccessful, probably because of the steric hindrance. Once obtained, the amines 41, 44a–n, 45, and 48 reacted with 6-chloropurine riboside to give the desired adenosine analogues 6–15 and 19–25. Compound 18 was obtained by simple hydrogenation on Pd/C of the nitro derivative 15.

Scheme 2.

Synthetic route used to prepare compounds 19 and 25 via a trimethylsilylethyl intermediate.

One of the analogues, containing a 3-nitro group (15), was resolved into pure diastereomers with HPLC, using the chiral column Chiralpak AD. The assignment of absolute configuration was done by analogy to the unsubstituted phenylcyclopropyl derivatives as standards.⁷ In the condition used for the chiral separation, the pure N⁶-[(1R,2S)-2-phenyl-1-cyclopropyl]adenosine (4) showed a retention time of 10.1 min and N⁶-[(1S,2R)-2-phenyl-1-cyclopropyl]adenosine (3) of 13.3 min. By analogy, in the resolution of the racemic 15, the first eluting peak at the retention time of 20.8 min was assigned as the N⁶-[(1R,2S)-2-(3-nitrophenyl)-1-cyclopropyl] adenosine (17) and the later one at 31.5min as N⁶-[(1S,2R)-2-(3-nitrophenyl)-1-cyclopropyl]adenosine (16).

3. Biological activity

The analogues 5–25 of N⁶-(trans-2-phenyl-1-cyclopropyl) adenosine (Table 1) generally bound to the hA₃AR in the low nanomolar range. Selectivity was high when compared with the A_2AARs, but only moderate, at most, compared with the A₁ARs. The diastereomeric mixture of N⁶-(trans-2-phenyl-1-cyclopropyl)adenosine 5 was prepared for comparison with the phenyl-substituted analogues. Compound 5 bound to the hA₃AR with a K_i value of 0.86nM. Its affinity at the rat A₃AR (rA₃AR) was 460-fold lower than at the human receptor. Substitution of the 2-phenyl ring at the 3-position was favorable for hA₃AR affinity. Thus, among the methyl-substituted analogues 6–8, the most potent in A₃AR binding was the 3-methyl analogue 7. Similarly, within a series of chloro-substituted analogues 9–11, the order of potency in binding was 3-Cl > 4-Cl > 2-Cl. The 3-chloro analogue 10 bound to the hA₃AR with a K_i value of 0.98nM, and the affinity at the rA₃AR was 660-fold lower. Thus, there were dramatic species differences in affinity, and A₃AR selectivity of 10 was present when comparing human but not rat ARs. At rARs, compound 10 was moderately selective for the A₁AR.

Table 1.

Binding affinities of adenosine derivatives at human and rat A₁, A_2AARs and A₃ARs and maximal agonist effects at the hA₃AR expressed in CHO cells.^a The adenosine derivatives are substituted at the N⁶ position as indicated. Compounds 1 and 2 are additionally substituted at the 2-position with chloro. All compounds except 2 are simple 9-β-D-riboside derivatives

#	N6 substitution	Ki (nM)					Maximum effect (%)d
		rA1ARa	rA2AARb	hA1ARa	hA2AARb	hA3AR, h or (r)c
trans-N6-(2-Phenyl-1-cyclopropyl) analogues
3	(1S,2R)-2-Phenyl-1-cPr	11.8± 2.4e	560± 232e	30.1± 6.1	2250 ± 430	0.63 ± 0.17e (358 ± 33)e	117 ± 9
4	(1R,2S)-2-Phenyl-1-cPr	15.2± 3.2e	3040± 490e	15.6± 1.7	2340 ± 330	24.1 ± 10.9e (694 ± 157)e	87 ± 4
5	2-Phenyl-1-cPr	10.4± 1.9	2980± 310	124± 30	2530 ± 720	0.86 ± 0.09 (399 ± 28)e	101 ± 5
6	2-(2-Methylphenyl)-1-cPr	18.3± 3.3	1080± 250	116± 19	4480 ± 460	12.9 ± 3.9	98 ± 4
7	2-(3-Methylphenyl)-1-cPr	17.3± 2.9	2500± 550	12.1± 2.7	970 ± 340	1.59 ± 0.50	107 ± 11
8	2-(4-Methylphenyl)-1-cPr	20.0± 1.9	2760± 970	28.0± 3.4	3130 ± 710	1.96 ± 0.14	99 ± 7
9	2-(2-Chlorophenyl)-1-cPr	11.8± 1.5	1430± 330	28.4± 13.7	1820 ± 170	6.0 ± 1.4	104 ± 6
10	2-(3-Chlorophenyl)-1-cPr	10.3± 1.1	349± 56	24.9± 2.1	1520 ± 540	0.98 ± 0.27 (642 ± 153)	101 ± 3
11	2-(4-Chlorophenyl)-1-cPr	9.70± 2.73	1430± 540	79.5± 41.0	4120 ± 890	1.63 ± 0.26	103 ± 7
12	2-(3-Fluorophenyl)-1-cPr	13.6± 1.7	1410± 750	35.1± 8.3	2930 ± 330	7.6 ± 1.4	110 ± 12
13	2-(3,5-Difluorophenyl)-1-cPr	18.8± 1.8	4060± 1330	39.7± 4.0	3270 ± 700	34.7 ± 3.1	99 ± 8
14	2-(3-Trifluoromethyloxyphenyl)-1-cPr	36.3± 9.9	1360± 230	146± 63	7110 ± 930	77.3 ± 21.3	97 ± 6
15	2-(3-Nitrophenyl)-1-cPr	15.4± 2.8	1060± 260	47.5± 0.5	2260 ± 270	18.8 ± 5.2	100 ± 5
16	2-(3-Nitrophenyl)-1-cPr, B (1S,2R)	13.6± 3.2		83.9± 8.2	>10,000	11.2 ± 2.7	108 ± 4
17	2-(3-Nitrophenyl)-1-cPr, A (1R,2S)	13.2± 4.6		55.4± 7.8	1320 ± 370	116 ± 13	64 ± 7
18	2-(3-Aminophenyl)-1-cPr	46.9± 7.8	8040± 1380	131± 52	>10,000	9.0 ± 3.3	103 ± 7
19	2-(3-Acetamidophenyl)-1-cPr	21.5± 9.4		32.6± 7.5	2130 ± 700	6.02 ± 0.74	100 ± 9
20	2-(3-Cyanophenyl)-1-cPr	57.4± 5.8	2450± 610	45.2± 3.3	2480 ± 780	22.5 ± 7.5	99 ± 7
21	2-(3-Methoxyphenyl)-1-cPr	15.8± 1.4	795± 193	29.1± 5.9	2830 ± 340	2.8 ± 0.3	103 ± 2
22	2-[3-(Trifluoromethyl)phenyl]-1-cPr	15.6± 1.6	459± 97	104± 30	2370 ± 780	1.9 ± 0.2	101 ± 5
23	2-[3,5-Di(trifluoromethyl)phenyl]-1-cPr	46.1± 6.7	869± 351	40.7± 2.4	7660 ± 1040	387 ± 37	98 ± 8
24	2,2-Diphenyl-1-cPr	22.6± 8.7	54.6± 17.6	71± 15	579 ± 186	91 ± 14	100 ± 4
25	2-(3-Thienyl)-1-cPr	5.65± 1.97		31.6± 8.2	2120 ± 670	2.88 ± 0.90	99 ± 7
Other analogues
1	CADO	6.7± 1.0f	63e	7.5± 1.4	630 ± 220	87 ± 24 (1890 ± 900)g	100 ± 7e
2	3-Iodobenzyl, Cl-IB-MECA	820e	470e	1240± 320h	5360 ± 2470h	1.4 ± 0.3h (0.33)e	100
26	Benzyl	175 ± 20e	285e	77.8± 6.5	2180 ± 670	41.3 ± 5.3e (120 ± 20)g	55 ± 3
27	2-Phenylethyl	24.0± 8.8f	521± 90	12.9± 2.1	676 ± 39	2.1 ± 0.4 (240 ± 58)g	84 ± 5
28	(R)-1-Phenyl-2-propyl, (R)-PIA	1.2± 0.1f	124f	2.04i	859i	8.7± 0.9e (158)g	102 ± 6e
29	(S)-1-Phenyl-2-propyl, (S)-PIA	49.3f	1820f	75i	7780i	68 ± 12e (920)g	97 ± 3e
30	(R)-2-Phenyl-1-propyl	1.4± 0.1e	319± 114	4.0± 1.3	325 ± 85	9.1± 0.3 (202 ± 20)	99 ± 4
31	(S)-2-Phenyl-1-propyl	3.5± 0.3e	459± 162	26.6± 6.8	1120 ± 260	38.8 ± 4.1 (276 ± 20)	98 ± 6
32	Cyclopropyl	14.7± 1.7	3490± 260	6.9± 2.4	7860 ± 550	100 ± 33 (1950 ± 120)	0j
33	2,2-Diphenylethyl	44.1± 1.7	75.4± 14.9	49.9± 16.2	510 ± 49	3.9 ± 0.7 (538 ± 202)	0j
34	2-(3,5-Dimethoxyphenyl)-2-(2-methylphenyl)ethyl, DPMA	112 ± 49	4.4g	168± 29e	153 ± 26e	106 ± 22e (3570 ± 1700)g	0e
35	Diphenylmethyl	208 ± 36	2490± 420	490± 242	>10, 000	3340 ± 360	87 ± 6
36	9-Fluorenylmethyl	9.41± 3.11	33.4± 24.4	14.0± 4.0	145 ± 26	0.91 ± 0.38	99 ± 6

^aBinding experiments at rat brain and recombinant human A₁ARs used [³H]R-PIA (2.0 nM) as radioligand, unless noted.

^bBinding experiments at rat brain and recombinant human A_2AARs used [³H]CGS21680 (15 nM) as radioligand, unless noted.

^cAll A₃AR binding experiments were performed using adherent CHO cells stably transfected with cDNA encoding the human or rat A₃ receptor. [¹²⁵I]I-AB-MECA was used as radioligand.

^dData of hA₃AR at 10 μM. Functional assay consisting of inhibition of forskolin-stimulated adenylyl cyclase. The value for compound 2 is the standard for 100% efficacy.

^eData from Gao et al.⁷

^fData from Daly et al.¹⁴

^gData from van Galen et al.²³

^hData from Jacobson et al.²⁴

ⁱData from Klotz et al.²⁵ Radioligands used were [³H]CPX and [³H]NECA at human A₁ and A_2A ARs, respectively.

^jWhen the stock DMSO solution was subjected to repeated freeze-thaw cycles, the observed efficacy was affected. Values given are for freshly prepared solutions.

The 3-fluoro analogue 12 was 8-fold less potent than 10 at the hA₃AR. 3,5-Difluoro or 3-trifluoromethyl oxy substitution, in 13 and 14, respectively, further lowered the affinity. The 3-nitro analogue 15 displayed A₃AR selectivity of 120-fold compared with the hA_2AAR. The two isomers of this diastereomeric mixture, 16 and 17, differed 10-fold in A₃AR affinity. The more potent 1S,2R analogue 16 was >900-fold selective for the hA₃ compared with hA_2AAR. The corresponding 3-amino analogue 18 was 9-fold weaker than 15 at the hA₃AR. This 3-amino derivative also displayed decreased affinity at the human and rat A_2A ARs. The 3-acetamido derivative 19 was more potent than the corresponding 3-amino derivative at all subtypes. Another electron-withdrawing group, 3-cyano in compound 20, did not provide high affinity, but an electron-donating group, 3-methoxy in 21, resulted in high A₃AR affinity. The analogue 22, which contains an electron-withdrawing 3-trifluoromethyl group, bound to the hA₃AR with high affinity. However, two 3-trifluoromethyl groups in 23 greatly reduced the affinity at the A₃AR. The 2,2-diphenylethyl substituted analogue 24 was not highly potent at the A₃AR and was generally nonselective. However, a heteroaromatic group could be substituted in the place of the phenyl moiety of compound 5; thus, a 2-(3-thienyl) analogue 25 displayed high A₃AR affinity.

For comparison, a number of previously reported analogues were studied in the same assays. The N⁶-benzyl group has been widely explored in the design of A₃AR agonists, such as Cl-IB-MECA 2. The combination of N⁶-(3-iodobenzyl) and 5′-uronamido groups in 2 greatly enhanced both the A₃AR potency and the selectivity in comparison to 1. However, the simple 2-H benzyl derivative 26 was only 2-fold more potent at the hA₃AR than 1. The next higher homologue, the 2-phenylethyl derivative 27, was of greatly increased affinity and selectivity at the human (but not rat) A₃AR. Branching of the alkyl groups in N⁶-substituted adenosines has been well studied at the A₁AR and A_2AAR.^13–16 At the hA₃AR, a moderate degree of stereoselectivity of binding was observed for introduction of a methyl group in the R configuration at either the 1-(28 compared with 29) or 2-(30 compared with 31) position.

Although the cyclopropyl group of the N⁶-(2-phenyl-1-cyclopropyl) analogues 3–25 was introduced for steric constraint, intended to freeze the biologically preferred conformation, these analogues could also be considered hybrids of the N⁶-phenylethyl analogue 27 and the N⁶-cyclopropyl analogue 32. Compound 32, however, was not highly potent in binding to the hA₃AR.

A functional assay of A₃AR-mediated inhibition of forskolin-stimulated adenylyl cyclase showed marked effects of certain N⁶ substitution on intrinsic efficacy. The prototypical A₃AR agonist Cl-IB-MECA 2 is considered a full agonist in most studies,^1,6 although in calcium responses in monocyte-derived dendritic cells it appeared to be a partial agonist.¹⁸ The efficacy of the analogues was reported as a single percentage value (relative to the full agonist 2) at a nucleoside concentration of 10 μM. The efficacy at the hA₃AR of the N⁶-(2-phenyl-1-cyclopropyl) analogues 3–25 was generally nearly full (i.e., 100%). In a full concentration-response experiment (Fig. 1), compound 10 was found to be a full agonist at the hA₃AR in the inhibition of adenylate cyclase. The functional potency of 10 was comparable to that of 2. The functional EC₅₀ values for both were ~3 nM, which is in agreement with the K_i values in binding of ~1 nM. In contrast, N⁶-cyclopropyladenosine 32 appeared to be an antagonist, with lack of efficacy at 10 μM, at a concentration 100 times greater than its K_i value in binding at the hA₃AR. A N⁶-(2,2-diphenylethyl) analogue 33 also proved to be an antagonist. A Schild analysis of inhibition by 33 of the effects of an A₃AR agonist at the hA₃AR was carried out (Fig. 2); it demonstrated that the K_B value for this competitive antagonist was 5.0 nM. This was consistent with the previously observed antagonist properties of the closely related DPMA 34 at the hA₃AR.⁷ Shortening the ethyl moiety by one carbon, that is to obtain 35, or constraining the two rings with a biaryl bond, that is 36, largely restored efficacy, although the A₃AR affinity of these two compounds varied greatly. The 9-fluorenylmethyl analogue 36 was as potent at the hA₃AR as the most potent N⁶-(2-phenyl-1-cyclopropyl) analogues in this study. Nevertheless, 36 displayed only minimal selectivity in comparison to the A₁AR. The effect of selected compounds, at a single concentration of 10 μM, on stimulation of adenylyl cyclase via the hA_2BAR in stably transfected CHO (Chinese hamster ovary) cells was examined. 5′-N-Ethyluronamidoadenosine (NECA) stimulated cyclic AMP accumulation in a concentration-dependent manner corresponding to an EC₅₀ of 140 ± 19 nM (n = 4). The maximum stimulation level of NECA at 10 μM was expressed as 100%. Most compounds tested at 10 μM had only a small effect (<50%) on cyclic AMP accumulation. The ranges of percentage stimulation were 30–50% (3, 6–10, and 15), 15–30% (5, 11, 12, 17, 19, 25, and 35), <15% (13, 14, 16, 18, 20–23, and 31). Compounds that stimulated to >50% of the NECA response were 4 (60 ± 2%), 24 (78 ± 2%), 30 (56 ± 3%), 34 (6 3± 8%), and 36 (96 ± 2%).

Figure 1.

Effects of Cl-IB-MECA and 10 on cAMP production in CHO cells stably expressing the hA₃AR. The cAMP level corresponding to 100% (10 μM forskolin) was 220±30 pmolmL⁻¹. The EC₅₀ value for compound 10 was 3.7 ±0.6 nM. Results are from a single representative experiment, which was carried out in triplicate.

Figure 2.

Antagonism by compound 33 of the inhibition of cyclic AMP production elicited by Cl-IB-MECA in CHO cells stably transfected with the hA₃AR. The experiment was performed in the presence of 10 μM rolipram and 3 units/mL adenosine deaminase. Forskolin (10 μM) was used to stimulate cyclic AMP levels. The level of cAMP corresponding to 100% was 220±30 pmolmL⁻¹. The K_B value for compound 33 was calculated to be 5.0 nM.

4. Molecular modeling

As the first step in modeling the environment within the hA₃AR surrounding the distal portion of the N⁶-binding region, the agonist template molecule N⁶-(1S,2R)-(2-phenyl-1-cyclopropyl) adenosine 3 was docked in a rhodopsin-based homology model of the receptor.⁸ We used the previously reported Cl-IB-MECA/hA₃AR complex as the starting geometry for the position of the ribose moiety of 3, followed by a systematic conformational search varying the rotatable bonds of the N⁶ substitutent, that is t₀ (C₅–C₆–N₆–C_cp) and t₁ (C₆–N₆–C_cp–C_cp) angles. The constraint of the cyclopropyl ring fixed the t₂ (N₆–C_cp–C_cp–C_ar) torsion angle. The energetically optimized result of docking 3 in the hA₃AR is shown in Figure 3. Mutagenesis results were consistent with molecular modeling that featured direct interaction of nucleosides with TMs 3, 6, 7, and EL2 (the second extracellular loop).^6,8,30 Residues that were within 5 Å proximity to 3 in this putative binding site were L91 (3.33), T94 (3.36), H95 (3.37), N150 (EL2), Q167 (EL2), F168 (EL2), S181 (5.42), M177 (5.38), V178 (5.39), F182 (5.43), F239 (6,44), W243 (6.48), L246 (6.51), S247 (6.52), N250 (6.55), I268 (7.39), S271 (7.42), H272 (7.43), and N274 (7.45). For comparison, the less potent isomer 4 was docked in a similar manner.

Figure 3.

The complex of the A₃AR with two regioisomers in the putative binding site of the hA₃AR model. (A) (1S,2R)-2-phenyl-1-cyclopropyl-adenosine 3 and (B) (1R,2S)-2-phenyl-1-cyclopropyladenosine 4. All ligands are displayed as ball and stick models, and the side chains of hA₃AR are shown as stick models. The H-bonding between ligand and hA₃AR is displayed in yellow. The A₃AR is represented by a tube model with a different color for each TM domain (TM3 in yellow, TM4 in green, TM5 in cyano, TM6 in blue, TM7 in purple).

The purine ring of 3 was located in a hydrophobic pocket defined by L91 (3.33) and L246 (6.51). The amine of the N⁶ substituent in proximity to N250 (6.55) was H-bonded with the hydroxyl group of S247 (6.52), and the purine N³ atom formed an H-bond with the side chain of Q167 (EL2). The 2′-OH group of the ribose ring was involved in H-bonding with the backbone carbonyl group of I268 (7.39), and the 3′-OH group formed H-bonds with the carbonyl group of the backbone of S271 (7.42) and the side chain of H272 (7.43), consistent with our A₃ neoceptor model.³¹ The 5′-hydroxyl group also formed an H-bond with T94 (3.36). The phenyl moiety showed an additional hydrophobic interaction with F168 (EL2). Compared with the Cl-IB-MECA complex, in the case of N⁶-phenylcyclo propyladenosine, the phenyl ring showed a better π–π stacking interaction with the aromatic ring of F168 (EL2). The distances between the centers of the N⁶-phenyl ring and the F168 aromatic ring were 3.69, 4.55, and 4.21 Å, for compounds 3, 4, and Cl-IBMECA, respectively. Thus, the 1S,2R isomer 3 showed a better π–π stacking interaction than 4 in the distal region of the binding site of the N⁶ substituent, consistent with its higher measured binding affinity.

The docked conformation of 3 displayed a preference for t₀, t₁, and t₂ angles of approximately −110°, −80°, and 140°, respectively, in the docking complex of 4 the angles of t₀, t₁, and t₂ were −130°, 80°, and −140°. Thus, the preferred t₁ and t₂ angles differed between the two diastereomers. The 3-hA₃AR complex showed ~5 kcal lower energy than that of the agonist 4, correlating with the binding affinity. The binding affinity of 3 was ~40-fold higher than that of 4. The relative stability of the hA₃AR complex with the nitrophenyl diastereomer N⁶-(1S,2R)-(2-phenyl-1-cyclopropyl) adenosine 16 and its less potent 1R,2S diastereomer 17 was also checked. The receptor complex of 16 showed ~1.4 kcal lower energy than the complex of 17, consistent with the stereoselectivity of binding at the hA₃AR. Thus, the docking study of the hA₃AR suggested that the 1S,2R diastereomer might show more favorable enthalpy of binding to the hA₃AR compared with the 1R,2S isomer, thus increasing the binding affinity.

The adenosine analogues with substitution of the phenyl ring of the N⁶ substituent displayed a more limited rotational freedom of this ring when docked in the hA₃AR.³² The 1S,2R diastereoisomers were assumed to be generally more potent and therefore were used in the docking. The phenyl substituents tended to be directed toward EL2, especially residue N150. However, the docked complexes of 15 and 18 showed different locations of the nitro and amino groups of the phenyl ring, that is directed toward TM4 and TM5 rather than EL2, because of more favorable nonbonding van der Waals interactions, while keeping the same preference of t₀, t₁, and t₂ angles as for other phenyl substitutions.

All ligands in Table 1 were subjected to similar receptor docking, and binding energies were calculated. The relative binding energies for analogues substituted on the phenyl moiety at the ortho, meta, and para positions were 4.0, 0, and 1.4 kcal for a Cl atom, and 1.3, 0, and 0.3 kcal for a methyl group. For both substituents, receptor docking confirmed that the meta position was the most and the ortho position was the least energetically favorable for binding to the hA₃AR, and this would be consistent with the experimental results.

Following the initial receptor docking of the set of N⁶-modified adenosine derivatives, 3D-QSAR (quantitative structure activity relationship) studies were performed with CoMFA (comparative molecular field analysis)^34,35 and CoMSIA (comparative molecular similarity indices analysis).³⁶ Two distinct models for the training set in 3D-QSAR, 1 and 2 as defined in the experimental section, ³³ were derived by superimposition of the complexes in which the ligands were docked in the putative binding site of the hA₃AR. The results were compared according to the two models (Table 2). PLS (partial least squares) analysis of model 1 generated in the CoMFA 3D-QSAR model displayed a modest q² value of 0.44 and an r² value of 0.98 for 27 compounds only after removing five compounds (1, 10, 15, 18, and 29). These compounds were detected as outliers from the residual plot of the leave-one-out cross-validation. Model 2, with a similar 3D orientation to the docking study, generated a better 3D-CoMFA result than did model 1. This model displayed a q² value of 0.57 and an r² value of 0.97, for 29 out of 32 compounds. Fewer compounds (13, 21, and 32) were outliers in model 2 than in model 1.

Table 2.

The result of CoMFA and CoMSIA 3D-QSAR for two models

Statistics	CoMFA		CoMSIA
	Model 1	Model 2	Model 1	Model 2
q2	0.44	0.57	0.53	0.47
Number of compounds	27	29	30	30
Number of components	5	6	6	6
r2	0.98	0.97	0.96	0.96
SEEa	0.13	0.16	0.18	0.19
F	198	124	99	87
Contribution (%)
Steric	59.70	60.80	10.10	10.70
Electrostatic	40.30	39.20	30.60	38.90
Hydrophobic			28.30	31.20
H-bond donor			24.80	15.10
H-bond acceptor			6.20	4.20

^aStandard error of estimate.

CoMSIA methods with additional hydrophobic and H-bonding fields³⁷ have been shown to be of comparable statistical significance to traditional CoMFA models, but with somewhat more easily interpreted isocontour surface maps. Here, CoMSIA performed better than CoMFA. The better statistical result of CoMSIA was attributed to the large contributions of hydrophobic and H-bonding interactions (Table 2). In addition, the conformational diversity, using model 1 for 3D-QSAR, did not affect any statistical parameters of CoMSIA, whereas CoMFA was very sensitive to the conformation used for the model. Both CoMSIA models required two different outliers to reach the predictable model with a q² value of >0.4. The result of the CoMSIA model was a q² value of 0.53 for model 1 and a q² value of 0.47 for model 2 with the same r² value of 0.96 for both models.

The superimposition of the CoMFA/CoMSIA map onto the binding site of the receptor was interpretable with respect to the SAR, because all conformations of the two models used for 3D-QSAR analyses were obtained from the docking complex without any modifications, such as RMS fitting and database alignment with a template. The CoMFA and CoMSIA maps of the two models showed similar contour maps for steric, electrostatic, hydrophobic, and H-bonding fields, except for the contours representing regions of hydrophilicity and disfavored steric bulk (data not shown). The best contour maps for each property through the correlation with the binding site environment were the steric and the electrostatic maps from the CoMFA model 2 and the hydrophobic and H-bonding maps from the CoMSIA model 1 (Fig. 4). The projection of the CoMFA/CoMSIA contour maps onto the binding site, which was validated by the experimental results, displayed a good complementarity. As shown in Figure 4B, blue regions, indicating that an electropositive group would increase the affinity, were located near the meta position of the phenyl group (V141 (4.56), N150 (EL2)) and surrounding the adenine ring (N250 (6.55), Q167 (EL2), T94 (3.36)). The interpretation of the existence of two blue contour maps at the meta position of the phenyl ring was that most of the electron-withdrawing groups with a negative charge did not improve the binding affinity compared with the unsubstituted compound 3. A red CoMFA contour favoring a negatively charged group at the para position of the N⁶-phenyl ring binding site was in proximity to S181 (5.42). For the steric CoMFA map in Figure 4A, a large green contour indicating tolerance of steric bulk around the N⁶-phenyl group matched well with the hydrophobic binding site surrounded by F168 (EL2), M177 (5.38), and F182 (5.43). The green contour of CoMFA also coincided closely with the contour region of CoMSIA favoring hydrophobicity. Model 1 with the conformational diversity showed more regions of disfavored steric bulk in CoMFA and favored hydrophilic groups in CoMSIA than did model 2 (Fig. 4C). One of the hydrophilic contours from CoMSIA overlapped well with the less bulky contours from CoMFA. The H-bonding field map (Fig. 4D) was very useful because of the importance of H-bonding of adenosine analogues in binding to the hA₃AR, especially at the ribose-binding position and at the exocyclic amine of the adenine moiety. Because the ribose ring was the core moiety, H-bonding fields were detected only near the N⁶ and 5′-position. An H-bonding donor contour around the N⁶ position was located at the side chain of S247 (6.52), and an H-bond acceptor contour around the N⁶ position was directed toward the side chain of N250 (6.55). Thus, the steric, electrostatic, hydrophobic, and H-bonding contour maps were supported by the docking complex, with each contour map matched well with its surrounding amino acids with hydrophobic, hydrophilic, and H-bonding properties in the putative binding site of hA₃AR.

Figure 4.

CoMFA stddev * coeff contour plots for model 2 (Top) and CoMSIA stddev * coeff contour plots for model 1 (bottom). (A) Predicted effects of structural modification on the binding affinity of 3 as docked in the hA₃AR: green contours indicate sterically favored regions; yellow contours indicate sterically disfavored regions. (B) Blue contours define a region where increased positive charge will result in increased affinity, and red contours define a region where increased negative charge will be favorable. (C) To enhance the binding affinity of hA₃AR, yellow and white areas define hydrophobic and hydrophilic preferences, respectively. (D) The H-bond donor field contour display regions where H-bond acceptors on the ligand are predicted to enhance (cyan) and disfavor (purple) binding. For the H-bond acceptor field, H-bond donors on the ligand are predicted to enhance (magenta) and disfavor (red) binding.

5. Discussion

CoMFA/CoMSIA 3D-QSAR and docking studies were conducted on a series of potent, conformationally constrained A₃AR agonists. The CoMFA/CoMSIA maps from the 3D-QSAR study and the putative binding site, based on both experimental results and a docking study, were integrated to propose a binding mode for hA₃AR agonists. The superimposition of the contour map from the CoMFA/CoMSIA study and the hA₃AR docking complex validated each other for the predictability of the ligand-based method as a 3D-QSAR and receptor-based approach through a homology modeling and docking study.

The molecular modeling studies of the A₃AR⁶ indicate that required flexibility of the ribose moiety and a movement of TM6 were correlated with receptor activation. Similarly, at the A_2AAR a rotation of a conserved Trp of TM6 has been proposed to be involved in activation.⁸ We have assembled this series of adenosine derivatives to apply a similar analysis of SAR to the N⁶ region.

The present study focused on derivatives of N⁶-(2-phenylethyl)adenosine, in which various degrees of rigidity (a cyclopropane ring) and steric bulk (e.g., methyl groups of 28–31) were included. The strikingly high affinity of the N⁶-(1S,2R)-2-phenyl-1-cyclopropyl analogue 3 was further explored through phenyl ring substitution. None of the diastereomeric analogues proved to be more potent and selective than 3, although moderate selectivity (22) and high affinity at the hA₃AR (7, 8, 10, 11, 21, 22, and 25) were achieved. The consistently greater affinity of these analogues at human compared with rat A₃ARs was also characteristic of the unsubstituted compounds 3 and 4 and the parent N⁶-(2-phenylethyl) adenosine 27.

Curiously, although the 2-phenyl-1-cyclopropyl analogues tended to be full agonists, several related derivatives had greatly reduced efficacy. The N⁶-cyclopropyl derivative 32 was an A₃AR antagonist; adding one or two phenyl rings at the 2-position restored efficacy. The N⁶-(2,2-diphenylethyl) derivative was a somewhat selective A₃AR antagonist, but either adding a bond between the two phenyl rings (N⁶-9-fluorenylmethyl) or shortening the ethyl moiety (N⁶-diphenylmethyl) restored efficacy. Thus, extending earlier findings,^6,7 the ability of an adenosine derivative to activate the A₃AR is highly dependent on the nature of the N⁶ substituent. A new series of high-affinity A₃AR agonists and related nucleoside antagonists was explored, and ligand docking in a molecular model of the hA₃AR defines a hydrophobic region for interaction with the N⁶-(2-phenylethyl) moiety. These findings for the N⁶ region may be combined synthetically with other structural modifications to enhance the pharmacological profile for either A₃AR selectivity or mixed A₁/A₃AR selectivity. Agonists of mixed A₁/A₃AR selectivity may be useful for treating cardiac ischemia.²² To investigate the structural basis for the differences in affinity between species and the striking variation in intrinsic efficacy will require the use of receptor mutagenesis.^6,8

6. Conclusion

A new series of high-affinity A₃AR agonists was explored. The adenosine derivatives are mainly sterically constrained analogues of N⁶-(2-phenylethyl)adenosine, found previously to display high affinity at the hA₃AR. The affinity and selectivity of these nucleosides is highly dependent on the species examined and on the substitution of a distal aryl substitution. A molecular model defines a hydrophobic region, which includes Phe168 of EL2, in the putative A₃AR binding site around the distal phenyl moiety. Upon probing of the SAR in this series, several novel nucleoside antagonists of the A₃AR were identified.

7. Experimental

7.1. Chemistry

7.1.1. Materials and instrumentation

Reagents and solvents were purchased from Sigma–Aldrich (St. Louis, MO). ¹H NMR spectra were obtained with a Varian Gemini 300 spectrometer using CDCl₃ as a solvent. The chemical shifts are expressed as ppm downfield from TMS. High-resolution FAB mass spectrometry was performed with a JEOL SX102 spectrometer using 6 kV Xe atoms. The chiral separation was done with a Hewlet Packard 1090 HPLC system using a Chiralpak AD column at an isocratic method with methanol as a mobile phase, with flow rate 1mL/min. Peaks were detected by UV absorption with a diode array detector. All final compounds were analyzed by LC/MS showing more than 96% purity. TLC analysis was carried out on aluminum sheets precoated with silica gel F₂₅₄ (0.2 mm) from Aldrich.

7.2. General procedure for the synthesis of compounds 6–18 and 20–23

7.2.1. Methyl trans-3-(3-Chlorophenyl)propenoate (38e)

Concentrated H₂SO₄ (0.1mL) was added to a solution of trans-3-(3-chlorophenyl) propenoic acid (0.2 g, 1.09 mmol) in MeOH (5 mL). The solution was heated at reflux overnight. After cooling the solution, the acid was neutralized with saturated aqueous Na₂CO₃. The aqueous solution was extracted with ether (3 × 50 mL). The organic phase was washed with brine, dried over Na₂SO₄, filtered, and concentrated. The residue was purified by flash chromatography on silica gel (ether/petroleum ether 1:9), to give 38e as a white solid. ¹H NMR (CDCl₃, 300,MHz) δ 7.63 (d, J = 16.0Hz 1H), 7.51 (br s, 1H), 7.34–7.26 (m, 2H), 6.43 (d, J = 16.0Hz 1H), 3.81 (s, 3H).

7.2.2. trans-2-(3-Chlorophenyl)-cyclopropanecarboxylic acid methyl ester (39e)

Diazomethane was generated with a diazomethane-generating glassware kit (Aldrich). A solution of N-methyl-N-nitroso-4-toluenesulfonamide (Diazald, 2.23 g, 10.4 mmol) in ether (24 mL) was added dropwise to a mixture of KOH (1.75 g, 31.2 mmol) in H₂O (18 mL), ether (4mL), and 2-(2-ethoxyethoxy)ethanol (18mL) kept at 70 °C. The ethereal solution of diazomethane was continuously distilled into a stirred solution of 38a (206 mg, 1.05 mmol) and Pd(OAc)₂ (1.16 mg, 0.00052) in CH₂Cl₂/ether (28/10 mL) kept at 0 °C. The rate of distillation was controlled to match the rate of addition. After the addition of Diazald was complete, the solution of 38a was stirred at rt for 30 min. The excess of diazomethane was destroyed with acetic acid. The resulting mixture was washed with a saturated solution of NaHCO₃, brine, dried over Na₂SO₄, filtered, and concentrated. NMR confirmed the resulting product to be pure 39e, to be used for the next reaction without further purification. ¹H NMR (CDCl₃, 300 MHz) δ 7.21–7.18 (m, 2H), 7.08–7.07 (m, 1H), 7.00–6.97 (m, 1H), 3.72 (s, 3H), 2.53–2.46 (m, 1H), 1.93–1.87 (m, 1H), 1.64–1.59 (m, 1H), 1.34–1.29 (m, 1H).

7.2.3. trans-2-(3-Chlorophenyl)-cyclopropanecarboxylic acid (40e)

Aqueous NaOH (2M, 3 mL) was added to a solution of 39e (200 mg, 1.13 mmol) in MeOH (2 mL), and the mixture was stirred for 2 h at rt. The mixture was concentrated and H₂O (20 mL) was added. The aqueous phase was washed with ether, acidified with 3M aqueous HCl, and extracted with ether (20mL×3). The combined organic phases were dried over Na₂SO₄, filtered, and concentrated to give pure 40e. ¹H NMR (CDCl₃, 300 MHz) δ 7.23–7.20 (m, 2H), 7.10–7.08 (m, 1H), 7.02–6.98 (m, 1H), 2.61–2.54 (m, 1H), 1.93–1.89 (m, 1H), 1.71–1.65 (m, 1H), 1.44–1.37 (m, 1H).

7.2.4. trans-2-(3-Chlorophenyl)-cyclopropylamine (41e)

A mixture of 40e (105 mg, 0.72 mmol) in dry t-BuOH (1.3 mL), diphenylphosporazidate (170 μL, 0.8 mmol), and triethylamine (105 μL, 1 mmol) was stirred at 90 °C under nitrogen atmosphere for 48 h. The solution was concentrated and poured into 10% aqueous Na₂CO₃ (20mL) and extracted with ether (10mL×3). The combined organic phases were dried over Na₂SO₄, filtered, and concentrated. The resulting t-butyl carbamate was dissolved in MeOH (3 mL), and 1M aqueous HCl (5 mL) was added. The solution was maintained at reflux overnight. Then the mixture was cooled and concentrated. The solution was washed with ether (10mL×3), and the aqueous phase was alkalinized with 10% aqueous K₂CO₃ until pH 10. The mixture was extracted with EtOAc (10mL×3). The combined organic phases were dried over K₂CO₃, filtered, and concentrated. The residue was purified by PTLC (chloroform/methanol 9:1) to give 41e. ¹H NMR (CDCl₃, 300MHz) δ 7.19–7.08 (m, 2H), 6.98–6.96 (m, 1H), 6.91–6.88 (m, 1H), 2.56–2.52 (m, 1H), 1.86–1.81 (m, 1H), 1.80–1.48 (br s, 2H), 1.11–1.04 (m, 1H), 1.00–0.94 (m, 1H).

7.2.5. trans-2-(3-Chlorophenyl)-cyclopropyladenosine (10)

Compound 41e (12 mg, 0.07 mmol) and 6-chloropurine riboside (25.8 mg, 0.09) were placed in a sealed tube with absolute ethanol (1 mL) and triethylamine (10 μL) and heated at 80 °C for 8 h. The solution was evaporated and the residue dissolved in chloroform (10 mL) and washed with H₂O (10mL×3). The organic phase was dried over Na₂SO₄, filtered, and concentrated. The residue was purified by PTLC (chloroform/methanol 9:1) to give pure 10. ¹H NMR (CDCl₃, 300 MHz) δ 8.23–8.21 (m, 1H), 7.77–7.76 (m, 1H), 7.22–7.08 (m, 4H), 6.48 (br s, 1H), 6.16 (br s, 1H), 5.79 (d, J = 7.2Hz 1H), 5.05 (t, J = 5.7Hz 1H), 4.46 (d, J = 4.5Hz 1H), 4.34 (s, 1H), 3.94 (d, J = 13.2Hz 1H), 3.78–3.75 (m, 1H), 3.18 (br s, 1H), 2.25–2.06 (m, 1H), 1.73–1.51 (m, 1H), 1.41–1.25 (m, 2H). HR-MS: calculated for C₁₉H₂₁O₄N₅Cl 418.1282 found 418.1282.

7.2.6. N⁶-[trans-2-(2-Methylphenyl)cyclopropyl]adenosine (6)

Yield 45%. ¹H NMR (CDCl₃, 300 MHz) δ 8.18–8.15 (m, 1H), 7.76 (s, 1H), 7.18–7.14 (m, 4H), 6.72–6.65 (m, 1H), 6.36–6.31 (m, 1H), 5.79 (d, J = 7.2Hz 1H), 5.02 (br s, 1H), 4.43 (d, J = 4.5Hz 1H), 4.32 (s, 1H), 3.94 (d, J = 13.2Hz 1H), 3.75–3.70 (m, 1H), 3.22 (br s, 1H), 2.40 (d, J = 3 Hz, 3H), 2.18–2.13 (m, 1H), 1.36–1.23 (m, 4H). MS (APCI) m/z = 398 (M+1).

7.2.7. N⁶-[trans-2-(3-Methylphenyl)cyclopropyl]adenosine (7)

Yield 49%. ¹H NMR (CDCl₃, 300MHz) δ 8.19 (s, 1H), 7.74 (m, 1H), 7.22–7.15 (m, 1H), 7.05–6.95 (m, 3H), 6.29 (br s, 1H), 5.78–5.72 (m, 1H), 5.04 (m, 1H), 4.45 (m, 1H), 4.32 (s, 1H), 3.92 (d, J = 13.2Hz 1H), 3.73 (d, J = 13.2Hz 1H), 3.17 (s, 1H), 2.33 (s, 3H), 2.09 (m, 1H), 1.37–1.24 (m, 2H). MS (APCI) m/z = 398 (M+1).

7.2.8. N⁶-[trans-2-(4-Methylphenyl)cyclopropyl]adenosine (8)

Yield 47%. ¹H NMR (CDCl₃, 300MHz) δ 8.32 (s, 1H), 7.78 (s, 1H), 7.11 (m, 2H), 6.17 (br s, 1H), 5.79 (d, J = 7.8Hz 1H), 5.05 (m, 1H), 4.50 (d, J = 3.8Hz 1H), 4.35 (s, 1H), 3.96 (d, J = 13.2Hz 1H), 3.75 (d, J = 13.2Hz 1H), 3.49 (s, 1H), 3.20 (m, 1H), 2.84 (br s, 1H), 2.33 (s, 3H), 2.16 (m, 1H), 1.29–1.25 (m, 2H). MS (APCI) m/z = 398 (M+1).

7.2.9. N⁶-[trans-2-(2-Chlorophenyl)cyclopropyl]adenosine (9)

Yield 40%. ¹H NMR (CDCl₃, 300 MHz) δ 8.17–8.16 (m, 1H) 7.76 (s, 1H), 7.78–7.77 (m, 1H), 7.38–7.35 (m, 1H), 7.21–7.11 (m, 4H), 6.72–6.65 (br m, 1H), 6.48–6.31 (br m, 1H), 5.79–5.78 (m, 1H), 5.01 (br s, 1H), 4.45 (br s, 1H), 4.32 (s, 1H), 3.94 (d, J = 13.2Hz 1H), 3.75–3.70 (m, 1H), 3.52 (br s, 1H), 3.15(br s, 1H), 2.48 (m, 1H), 1.77 (br s, 1H), 1.43–1.23 (m, 4H). MS (APCI) m/z = 418 (M+1).

7.2.10. N⁶-[trans-2-(4-Chlorophenyl)cyclopropyl]adenosine (11)

Yield 42%. ¹H NMR (CDCl₃, 300 MHz) δ 8.35 (s, 1H), 7.80 (s, 1H), 7.32–7.15 (m, 4H), 6.27 (br m, 1H), 6.09 (br s, 1H), 5.80 (d, J = 7.8Hz 1H), 5.06 (br s, 1H), 4.51 (d, J = 3.8Hz 1H), 4.35 (s, 1H), 3.96 (d, J = 13.2Hz 1H), 3.78–3.73 (m, 1H), 3.25 (br s, 1H), 2.74 (br s, 1H), 2.14 (m, 1H), 1.33–1.25 (m, 2H). MS (APCI) m/z = 418 (M+1).

7.2.11. N⁶-[trans-2-(3-Fluorophenyl)cyclopropyl]adenosine (12)

Yield 35%. ¹H NMR (CDCl₃, 300 MHz) δ 8.25 (s, 1H), 8.15 (s, 1H) 7.86–7.24 (m, 4H), 7.07 (br s, 1H), 7.02 (br s, 1H), 5.79 (d, J = 7.8Hz 1H), 5.15 (m, 1H), 4.52–4.30 (m, 2H), 3.95–3.72 (m, 2H), 2.93–2.85 (m, 1H), 2.35 (br s, 1H), 1.43–1.20 (m, 2H). MS (APCI) m/z = 402 (M+1).

7.2.12. N⁶-[trans-2-(3,5-Difluorophenyl)cyclopropyl]adenosine (13)

Yield 40%. ¹H NMR (CDCl₃, 300 MHz) δ 8.35 (s, 1H) 7.81 (s, 1H), 6.78–6.66 (m, 3H), 6.09 (br s, 1H), 5.80 (d, J = 7.8Hz 1H), 5.07 (br s, 1H), 4.50 (br s, 1H), 4.35 (s, 1H), 5.96 (d, J = 12 Hz, 1H), 3.76 (d, J = 12 Hz, 1H), 3.21 (br s, 1H), 2.77 (br s, 1H), 2.17 (m, 1H), 1.35–1.15 (m, 2H). MS (APCI) m/z = 420 (M+1).

7.2.13. N⁶-[trans-2-(3-Trifuoromethyloxyphenyl)cyclopropyl] adenosine (14)

Yield 30%. ¹H NMR (CDCl₃, 300MHz) δ 8.35 (s, 1H), 8.18 (s, 1H) 7.96–7.34 (m, 4H), 7.07 (br s, 1H), 7.02 (br s, 1H), 5.79 (d, J = 7.8Hz 1H), 5.05 (m, 1H), 4.51–4.34 (m, 2H), 3.97–3.70 (m, 2H), 2.92–2.85 (m, 1H), 2.32 (br s, 1H), 1.41–1.25 (m, 2H). MS (APCI) m/z = 468 (M+1).

7.2.14. N⁶-[trans-2-(3-Nitrophenyl)cyclopropyl]adenosine (15)

Yield 42%. ¹H NMR (CDCl₃, 300MHz) δ 8.26–8.1 6 (m, 1H), 8.15 (br s, 1H), 8.07 (d, J = 8.4 Hz, 1H), 7.82 (s, 1H), 7.60–7.57 (m, 1H), 7.50–7.39 (m, 4H), 5.85 (d, J = 6.3 1H), 5.02 (t, J = 6.5 Hz, 1H), 4.47 (br s, 1H), 4.35 (s, 1H), 3.97 (d, J = 13.2Hz 1H), 3.78–3.73 (m, 1H), 3.17 (br s, 1H), 2.25 (m, 1H), 1.50–1.41 (m, 2H). MS (APCI) m/z = 429 (M+1).

7.2.15. N⁶-[trans-2-(3-Aminophenyl)cyclopropyl]adenosine (18)

Yield 35%. ¹H NMR (CD₃OD, 300 MHz) δ 8.26 (s, 2H), 7.02 (t, J = 7.8 Hz, 1H), 6.56 (t, J = 7.8 Hz, 3H), 5.96 (d, J = 6.6, 1H), 4.76–4.72 (m, 1H), 4.33–4.31 (m, 1H), 4.17–4.16 (m, 1H), 3.91–3.71 (m, 2H), 2.15–2.07 (m, 1H), 1.35–1.22 (m, 2H). MS (APCI) m/z = 399 (M+1).

7.2.16. N⁶-[trans-2-(3-Acetamidophenyl)cyclopropyl]-adenosine (19)

Yield 39%. ¹H NMR (CD₃OD, 300 MHz) δ 8.26 (br s, 2H), 7.43–7.36 (m, 2H), 7.25–7.19 (m, 1H), 6.97–6.95 (m, 1H), 5.97 (d, J = 6.6, 1H), 4.36–4.31 (m, 1H), 4.17–4.14 (m, 1H), 3.92–3.87 (m, 2H), 3.25–3.07 (m, 1H), 2.17–2.04 (m, 4H), 1.38–1.23 (m, 2H). MS (APCI) m/z = 441 (M+1).

7.2.17. N⁶-[trans-2-(2-Cyanophenyl)cyclopropyl]adenosine (20)

Yield 45%. (CDCl₃, 300 MHz) δ 8.27 (s, 1H) 7.83 (s, 1H), 7.33–7.30 (m, 1H), 7.29–7.20 (m, 3H), 6.19–6.03 (m, 2H), 5.79 (d, J = 7.2Hz 1H), 5.05 (m, 1H), 4.48 (m, 1H), 4.34 (s, 1H), 3.94 (d, J = 13.2Hz 1H), 3.81–3.75 (m, 1H), 3.22 (m, 1H), 2.95–2.86 (m, 1H), 2.12–2.08 (m, 1H), 1.48–1.25 (m, 2H). MS (APCI) m/z = 408 (M+1).

7.2.18. N⁶-[trans-2-(3-Methoxyphenyl)cyclopropyl]adenosine (21)

Yield 38%. ¹H NMR (CDCl₃, 300MHz) δ 8.24 (s, 1H), 7.77 (br s, 1H), 6.85–6.72 (m, 3H), 6.25 (br s, 1H), 5.79 (br s, 1H), 5.05 (m, 1H), 4.48–4.34 (m, 2H), 3.96–3.72 (m, 6H), 3.20 (br s, 1H), 2.17 (m, 1H), 1.50–1.38 (m, 2H). MS (APCI) m/z = 414 (M+1).

7.2.19. N⁶-[trans-2-(3-Trifluoromethylphenyl)cyclopropyl] adenosine (22)

Yield 30%. ¹H NMR (CDCl₃, 300 MHz) δ 8.34 (s, 1H), 7.8 1 (br s, 1H), 7.52–7.40 (m, 4H), 6.30 (br s, 1H), 6.12 (br s, 1H), 5.81 (d, J = 6.9, 1H), 5.07 (m, 1H), 4.48 (m, 1H), 4.35 (s, 1H), 3.97 (m, 1H), 3.81–3.75 (m, 1H), 3.25 (br s, 1H), 2.25 (m, 1H), 1.51–1.38 (m, 2H). MS (APCI) m/z = 452 (M+1).

7.2.20. N⁶-[trans-2-[3,5-Di(trifluoromethyl)phenyl]cyclopropyl] adenosine (23)

Yield 32%. ¹H NMR (CDCl₃, 300 MHz) δ 8.37 (s, 1H), 8.21 (s, 1H) 7.86–7.34 (m, 3H), 7.07 (br s, 1H), 6.99 (br s, 1H), 5.79 (d, J = 7.8Hz 1H), 5.07 (m, 1H), 4.51–4.34 (m, 2H), 3.97–3.70 (m, 2H), 2.92–2.89 (m, 1H), 2.32 (br s, 1H), 1.31–1.25 (m, 2H). MS (APCI) m/z = 520 (M+1).

7.2.21. N⁶-(2,2-Diphenylcyclopropyl)adenosine (24)

Yield 52%. ¹H NMR (CDCl₃, 300MHz) δ 8.31–8.24 (m, 2H), 7.51–7.13 (m, 10H), 6.48 (br s, 1H), 5.89 (br s, 1H), 5.61 (d, J = 7.2Hz 1H), 4.91–4.87 (m 1H), 4.34 (d, J = 4.5 Hz 1H), 4.24 (m, 1H), 3.90–3.83 (m, 2H), 3.70–3.63 (m, 1H), 1.78–1.45 (m, 2H). MS (APCI) m/z = 460 (M+1).

7.2.22. N⁶-[2-(3-Thienyl)cyclopropyl]adenosine (25)

Yield 45%. ¹H NMR (CDCl₃, 300 MHz) δ 8.32 (s, 1H), 7.80 (s, 1H) 7.02–7.01 (m, 2H), 6.13 (br s, 1H), 5.81 (d, J = 7.5Hz 1H), 5.06 (m, 1H), 4.50 (d, J = 4.5Hz 1H), 4.35 (s, 1H), 3.96 (d, J = 12.8 Hz, 1H), 3.76 (d, J = 12.8 Hz, 1H), 3.22–3.13 (m, 1H), 2.82 (br s, 1H), 2.22–1.18 (m.1H), 1.31–1.20 (m, 2H). MS (APCI) m/z = 390 (M+1).

7.2.23. N⁶-(Cyclopropyl)adenosine (32)

Yield 65%. ¹H NMR (CDCl₃, 300 MHz) δ 8.25 (s, 1H), 7.77 (s, 1H) 6.46 (d, J = 11.4 Hz, 1H), 6.03 (br s, 1H), 5.79 (d, J = 7.5Hz 1H), 5.06 (m, 1H), 4.48 (d, J = 4.5Hz 1H), 4.34 (s, 1H), 3.94 (d, J = 10.8 Hz, 1H), 3.12–2.93 (m, 2H), 0.93 (d, J = 6.9, 2H), 0.66 (s, 1H). MS (APCI) m/z = 308 (M+1).

7.2.24. N⁶-[2,2-Di-(phenylethyl)cyclopropyl]adenosine (33)

Yield 60%. ¹H NMR (CDCl₃, 300MHz) δ 8.13 (s, 1H), 7.59 (s, 1H) 7.32–7.17 (m, 10H), 6.01 (br s, 1H), 5.68 (d, J = 7.2Hz 1H), 4.92 (t, J = 6.5Hz 1H), 4.36–4.26 (m, 4H), 4.09 (br s, 1H), 3.90–3.86 (d, J = 12.9Hz 1H), 3.73–3.66 (m, 2H). MS (APCI) m/z = 448 (M+1).

7.2.25. trans-2-(3-Acetamidophenyl)-cyclopropylamine (44)

Ethylchloroformate (0.035 mL, 0.35 mmol) was added to a solution of 42 (50 mg, 0.25 mmol) and triethylamine (0.04mL, 0.3 mmol) in dry acetone at −10 °C. The solution was stirred at the same temperature for 2 h, and a solution of NaN₃ (25 mg, 0.38 mmol) in H₂O (0.5 mL) was added. The stirring was interrupted after 1 h, and additional H₂O (5 mL) was added. The solution was concentrated and extracted with EtOAc (20mL×3). The combined organic layers were dried over Na₂SO₄, filtered, and concentrated. The residue was dissolved in toluene (50 mL), and the solution was concentrated by half to eliminate traces of H₂O. The solution was heated at 90 °C for 2 h while observing N₂ evolution. Then the toluene was evaporated, and the resulting isocyanate was dissolved in dry 2-(trimethylsilyl) ethanol (2mL) and the solution was heated to 60 °C overnight. The solution was cooled and concentrated. The residue was purified by flash chromatography on silica gel (ether/petroleum ether 1:9), to give the pure carbamate. The carbamate was treated with tetrabutylammonium fluoride (1M solution in THF 0.28 mL, 0.28 mmol) at 50 °C for 24 h. The solution was cooled, and H₂O (5 mL) was added, and the resulting mixture was stirred for 20 min. Then it was concentrated, and the aqueous solution was acidified with 1M HCl and washed with ether (20mL×3), and the aqueous phase was alkalinized with 10% aqueous Na₂CO₃. The mixture was extracted with EtOAc (10 mL×3). The combined organic phases were dried over K₂CO₃, filtered, and concentrated. The residue was purified by PTLC (chloroform/methanol 9:1) to give 44. ¹H NMR (CDCl₃, 300 MHz) δ 8.18–8.01 (m, 1H), 7.33–7.28 (m, 1H), 7.17–7.12 (m, 1H), 6.73–6.71 (m, 1H), 3.30–3.24 (m, 1H), 2.54 (br s, 1H), 2.18 (s, 3H), 1.87–1.83 (m, 1H), 1.61 (br s, 2H), 1.44–1.35 (m, 2H).

7.2.26. 2,2-Diphenylcyclopropanecarboxylic acid (47)

1,1-Diphenylethylene (1.02 g, 5.67 mmol), dry Cu₂SO₄ (58.3 mg, 0.36 mmol), and dry benzene (2 mL) were put in a three-necked flask equipped with two condensers and a dropping funnel. The solution was stirred and heated at 75 °C for 6 h. Ethyldiazoacetate (1.3 mL) was added dropwise over 1 h. The mixture was cooled to rt and stirred overnight. NaOH (1.45 g, 36 mmol) in ethanol (11.5 mL) was added, and the mixture was refluxed for 6 h. The solution was concentrated in vacuo, and H₂O (20mL) was added. The aqueous mixture was heated to 90 °C and filtered and allowed to cool overnight. The crystals were filtered and redissolved in hot water, and the solution was filtered. The filtrate was acidified with 10% aqueous HCl, and the resulting precipitate was separated and identified as pure 47. ¹H NMR (CDCl₃, 300 MHz) δ 7.38–7.35 (m, 2H), 7.27–7.12 (m, 8), 2.52–2.47 (m, 1H), 2.08–2.05 (m, 1H), 1.62–1.57 (m, 1H).

8. Pharmacology

8.1. Materials

[³H]R-PIA and [¹²⁵I]I-AB-MECA were from Amersham Pharmacia Biotech (Piscataway, NJ), and [³H]CGS21680 was from Perkin–Elmer. (Boston, MA). Adenosine deaminase was obtained from Sigma (St. Louis, MO). All other compounds were obtained from standard commercial sources and were of analytical grade.

8.2. Biological assays

The procedures for [³H]R-PIA and [³H]CGS21680 binding to A₁ and A_2A receptors, respectively, were as previously described.⁷ Briefly, membranes (10–20 μg of protein) were incubated with radioligand and the competing adenosine derivative in duplicate, together with increasing concentrations of the competing compounds, in a final volume of 0.4mL Tris·HCl buffer (50mM, pH 7.4) at 25 °C for 60 min. Binding reactions were terminated by filtration through Whatman GF/B glass–fiber filters under reduced pressure with a MT-24 cell harvester (Gaithersburg, MD). Filters were washed three times with ice-cold buffer and placed in scintillation vials with 5mL scintillation fluid, and bound radioactivity was determined by using a liquid scintillation counter. Functional assays of adenylyl cyclase (either stimulation via the hA_2B AR or inhibition via the hA₃AR) stably transfected CHO cells was carried out as previously described.²⁶

8.3. Statistical analysis

Binding and functional parameters were estimated with GraphPAD Prism software (GraphPAD, San Diego, CA). IC₅₀ values obtained from competition curves were converted to K_i values with the Cheng–Prusoff equation. ¹⁹ Data were expressed as mean ±standard error.

8.4. Molecular modeling

All calculations were performed on a Silicon Graphics Octane workstation (300 MHz, MIPS R12000 (IP30) processor, Mountain View, CA). All ligand structures were constructed using the Sketch Molecule of SYBYL 6.9.²⁰ A conformational search of compounds 3 and 4 to be docked was performed by random search for all rotatable bonds. The options of random search were 3,000 iteration, 3 kcal energy cutoffs, and chirality checking. In all cases, MMFF force field²⁷ and charge were applied using distance-dependent dielectric constants and conjugate gradient method until the gradient reached to 0.05 kcal/mol/Å. After clustering the low-energy conformers from the result of the conformational search, the representative conformers from all groups were reoptimized by semiempirical molecular orbital calculations with the PM3 method in the MOPAC 6.0 package.²⁸

A hA₃AR model (PDB code: 1o74) constructed by homology to the high-resolution X-ray structure of bovine rhodopsin¹⁷ was used for the docking study. The (1S,2R)-isomers of all ligands in Table 1 were docked within the hA₃AR model. The atom types were manually assigned with the Amber all atom force field²⁹ and their charges were calculated before docking. The starting geometry of ligand conformation was chosen from the hA₃AR complex model with Cl-IB-MECA, which was already validated by point mutation.⁶ The ribose-binding position of this series was fixed by an atom-by-atom fitting method for the carbon atoms of the ribose ring. Only N⁶-binding regions were variously positioned in the putative binding cavity, rotating the flexible bonds of N⁶ substituents, t₀ and t₁ angles. Several conformations without any steric bump were selected for further optimization. The initial structures of all complexes were optimized with the Amber force field with a fixed dielectric constant of 4.0 and terminating gradient of 0.1 kcal mol⁻¹Å⁻¹. Binding energy was calculated by the following equation: binding E=complex E−(receptor E+ligand E). These energies are not rigorous thermodynamic quantities but can only be used to compare the relative stabilities of the complexes. Consequently, these interaction energy values cannot be used to calculate binding affinities because changes in entropy and solvation effects are not taken into account. In addition, nonbonding van der Waals and electrostatic energies were calculated.

For the training set of 3D-QSAR, two models were generated. In model 1, a series of energetically favorable, bound conformations from the docking complex were selected and aligned in the 3D Cartesian space, and similar conformations for the t₀, t₁, and t₂ binding preference were used for model 2. The K_i values of hA₃AR for the training set were converted to pK_i (−log K_i) values as dependent variables in the CoMFA and CoMSIA. To derive the CoMFA and CoMSIA descriptors as independent variables, a 3D cubic lattice was automatically generated as a single grid with 2 Å space, overlapping all aligned molecules and extended by at least 4 Å along all axes. The steric fields were calculated with a Lennard–Jones potential and the electrostatic fields were calculated with a Coulombic potential at each lattice of a sp³ carbon probe atom with a van der Waals radius of 1.52 Å and a charge of +1.0. The default energy cutoff of 30 kcal/mol was used. CoMSIA descriptors were derived from the same lattice box used for the CoMFA calculations. For the calculation of CoMSIA similarity indices, five different similarity fields including steric, electrostatic, hydrophobic, H-bond donor, and H-bond acceptor were calculated at the regularly spaced grid points with a common probe atom with radius of 1 Å, charge, hydrophobicity, and H-bonding properties of +1. A Gaussian function with the default value (0.3) of the attenuation factor, α, was used for the distance dependence between the molecule and the probe atoms. The steric CoMSIA fields were from the internally coded parameters of the van der Waals table in SYBYL program. The electrostatic fields were calculated from the atomic partial charge of MMFF94. The hydrophobic fields were derived from atom-based values based on the research of Viswanadhan et al.³⁸ The H-bond donor and acceptor fields were obtained by a rule-based method from experimental values, creating dummy atoms at donor and acceptor sites like extension points of DISCO (distance comparisons).³⁷

The PLS regression analyses³⁹ were used to derive a linear relationship. The predictive value of the models was evaluated first by SAMPLS (sample distance partial least squares)⁴⁰ and then by leave-one-out cross-validation, with a 2 kcal/mol column filtering with a column scaling of CoMFA standard. The outlier points whose target values were badly predicted in the residual plot from the cross-validation analyses were omitted to get the predictable model with a sufficiently high q² value (>0.4). For the conventional r² value, final noncrossvalidation with the number of components to the optimum value from the cross-validation analysis was performed. CoMFA and CoMSIA stdev * coeff contour maps were generated by a default value of contribution.