A Binding Site Model and Structure-Activity Relationships for the Rat A₃ Adenosine Receptor

SUMMARY

A novel adenosine receptor, the A₃ receptor, has recently been cloned. We have systematically investigated the hitherto largely unexplored structure-activity relationships (SARs) for binding at A₃ receptors, using ¹²⁵I-N⁶-2-(4-aminophenyl)ethyladenosine as a radioligand and membranes from Chinese hamster ovary cells stably transfected with the rat A₃-cDNA. As is the case for A₁ and A_2a, receptors, substitutions at the N⁶ and 5′ positions of adenosine, the prototypic agonist ligand, may yield fairly potent compounds. However, the highest affinity and A₃ selectivity is found for N⁶,5′-disubstituted compounds, in contrast to A₁ and A_2a receptors. Thus, N⁶-benzyladenosine-5′-N-ethylcarboxamide is highly potent (K_i, 6.8 nM) and moderately selective (13- and 14-fold versus A₁ and A_2a). The N⁶ region of the A₃ receptor also appears to tolerate hydrophilic substitutions, in sharp contrast to the other subtypes. Potencies of N⁶,5′-disubstituted compounds in inhibition of adenylate cyclase via A₃ receptors parallel their high affinity in the binding assay. None of the typical xanthine or nonxanthine (A₁/A₂) antagonists tested show any appreciable affinity for rat A₃ receptors. 1,3-Dialkylxanthines did not antagonize the A₃ agonist-induced inhibition of adenylate cyclase. A His residue in helix 6 that is absent in A₃ receptors but present in A₁/A₂ receptors may be causal in this respect. In a molecular model for the rat A₃ receptor, this mutation, together with an increased bulkiness of residues surrounding the ligand, make antagonist binding unfavorable when compared with a previously developed A₁ receptor model. Second, this A₃ receptor model predicted similarities with A₁ and A₂ receptors in the binding requirements for the ribose moiety and that xanthine-7-ribosides would bind to rat A₃ receptors. This hypothesis was supported experimentally by the moderate affinity (K_i 6 μM) of 7-riboside of 1,3-dibutylxanthine, which appears to be a partial agonist at rat A₃ receptors. The model presented here, which is consistent with the detailed SAR found in this study, may serve to suggest future chemical modification, site-directed mutagenesis, and SAR studies to further define essential characteristics of the ligand-receptor interaction and to develop even more potent and selective A₃ receptor ligands.

Adenosine receptors, belonging to the superfamily of the G protein-coupled receptors, are generally divided into two major subclasses, A₁ and A₂, on the basis of the following: (i) the differential affinities of a number of adenosine receptor agonists and antagonists; (ii) their primary structures; (iii) the second messenger systems to which they couple. Thus, A₂ receptors (which can be further subdivided into A_2a and A_2b) stimulate adenylate cyclase, whereas A₁ receptors may couple to a variety of second messenger systems, including inhibition of adenylate cyclase, inhibition or stimulation of phosphoinositol turnover, activation of guanylate cyclase, activation of potassium influx, and inhibition of calcium influx (1, 2). A recent addition to the adenosine receptor family has been the A₃ receptor, which was cloned from rat brain (3) and rat testis (4) and which was first recognized as an adenosine receptor on the basis of its primary structure. In the putative transmembrane domains, it shows 58% identity with the canine A₁ receptor and 57% with the canine A_2a receptor. Like the A₁ receptor, it is negatively coupled to adenylate cyclase (3).

The physiological role of the A₃ receptor is mostly unexplored. Its distribution is fairly limited, and it is found primarily in the central nervous system (3), testes (4), heart (3), and the immune system, where it appears to be involved in the modulation of release from mast cells or other cells of mediators of the immediate hypersensitivity reaction (5). Activation of A₃ receptors also appears to cause xanthine-insensitive hypotensive response in pithed rats (44). In terms of therapeutic potential, a principal deficiency of A₁- and A_2a-selective agents has been their propensity for side effects, due to the ubiquitous nature of these receptors. However, the limited distribution of A₃ receptors raises hopes that A₃-selective compounds may be more useful as potential therapeutic agents.

Few ligands for this novel receptor have been reported.¹ Some nonselective N⁶-substituted adenosine derivatives have been described as agonists for the A₃ receptor, including APNEA (N⁶-2-(4-aminophenyl)ethyladenosine), which has been used successfully as a radioligand in its iodinated form (3). Curiously, xanthines (classical A₁ and A₂ antagonists) do not appear to bind to this receptor (3). Because the SAR at A₃ receptors is practically unexplored, we have systematically investigated a wide range of purine and nonpurine agents for affinity in binding to arrive at leads for achieving selectivity. We have integrated these pharmacological findings with insights derived from molecular modeling of A₁ receptors to present a binding site model unique for A₃ receptors.

Materials and Methods

Chemicals

F-12 (Ham’s) medium, fetal bovine serum, and penicillin/streptomycin were from Life Technologies, Inc. (Gaithersburg, MD) [¹²⁵I] APNEA was prepared as described previously (6). Adenosine deaminase (ADA) was from Boehringer Mannheim (Indianapolis, IN). Composition of lysis buffer was as follows: 10 mM Tris, 5 mM EDTA, pH 7.4, at 5 °. 50/10/1 buffer was as follows: 50 mM Tris, 10 mM MgCl₂, 1 mM EDTA, pH 8.26, at 5 °. Displaces were from RBI (Natick, MA) except xanthine, inosine, 8-chlorotheophylline (Aldrich, Milwaukee, WI); 1,9-dimethylxanthine, 3,9-dimethylxanthine (Fluka, Ronkonkoma, NY); 3,7-dimethylxanthine, 8-bromoadenosine, adenosine-N-oxide, α-D-adenosine, 2′-deoxyadenosine, 3′-deoxyadenosine, 5′-deoxyadenosine, 5′-deoxy-5′-methylthioadenosine, 5′-deoxy-5′-isobutylthioadenosine, S-adenosylmethionine, guanosine, 6-thioguanosine, 6-thiopurine riboside (Sigma); AMP (Boehringer Mannheim); xanthosine, uridine, thymidine, cytidine (Janssen/Spectrum, Gardens, CA). The following compounds were gifts, which are gratefully acknowledged: 1,3-dibutylxanthine, 1,3-dihexylxanthine, 1,3-dibenzylxanthine, 8-cyclohexycaffeine, 7-benzyltheophylline, N⁶-dimethyladenosine, 3-deazaadenosine, 7-deazaadenosine, β-L-adenosine, 2′-0-methyladenosine, adenine-β-D-arabinofuranoside, xylofuranosyladenosine, β-D-psicofuranosyladenine, 5′-deoxy-5′-aminoadenosine, 5′-carboxamidoadenosine, 2-thio-3-propylxanthine, 1-propyl-8-cyclopentylxanthine (Dr. J. Daly, NIH, Bethesda, MD); N⁶-cyclohexylNECA, 9-ethyl-N⁶-cyclopentyladenine, N⁶-dimethylNECA, N⁶-benzyl-N⁶-methyladenosine (Dr. R. Olsson, University of South Florida, Tampa, FL); CP 66713 (Dr. R. Sarges, Pfizer, Groton, CT); CGS 15943, (Dr. J. Francis, Ciba-Geigy, Summit, NJ). The syntheses of the following compounds have been described previously: theophylline-7-riboside, 1,3-dipropylxanthine-7-riboside, 1,3-dibutylxanthine-7 riboside (7); imidazo[4,5-c]quinolin-4-amine (8); N⁶-2-sulfoethyladenosine, N⁶-4-sulfophenyladenosine, N⁶-3-(4-sulfophenyl)-propyladenosine, N⁶-4-(4-sulfophenyl)butyladenosine (9). All other materials were from standard local sources and of the highest grade commercially available.

Synthesis

N⁶-Benzyladenosine-N¹-oxide

N⁶-Benzyladenosine (25 mg, 70 μmol) and m-chloroperbenzoic acid (38 mg, 220 μmol) were dissolved in 1 ml of acetic acid. The solution was stirred at room temperature for 2 days. The solvent was evaporated under a stream of nitrogen, and the residue was dissolved in a minimum of methanol and chromatographed on a silica plate (250 μ) using acetonitrile:water, 4:1 (v/v). The UV-absorbing band at R_F = 0.53 was extracted with methanol to provide 7.3 mg (28% yield). Mass spectrum and ¹H-NMR spectrum were consistent with the assigned structure.

Adenosine-5′-N-ethyluronamide-N¹-oxide

Adenosine-5′-N-ethyluronamide-N¹-oxide was synthesized by a method similar to N⁶-benzyladenosine-N¹-oxide. Following recrystallization from hot methanol/ether, the pure product was obtained in 28% yield. Mass spectrum (electron impact, peaks at 324 (m), 308) and ¹H-NMR spectrum were consistent with the assigned structure.

N⁶-Benzyladenosine-5′-N-ethyluronamide

To a solution of NECA (50 mg, 0.162 mmol) in DMF (1 ml) was added benzyl bromide (56 ml, 0.47 mmol), and the solution was stirred for 2 days at 40 ° while protected from moisture. DMF was removed in vacuo giving a syrup that crystallized when acetone and ether were added. The solvent was removed by decantation, and the solid was dried and dissolved in methanol (2 ml). K₂CO₃ (10 mg) was added and warmed under reflux overnight. The reaction mixture was cooled, filtered, and evaporated. The product was purified by preparative TLC (CHCl₃:MeOH, 13:2) in 42% yield. Melting point, 170–173 °. ¹H-NMR (in Me₂SO-d₆) was as follows: δ 1.06 (t, J = 7 Hz, 3H, CH₃), 3.20 (m, 2H, CH₂), 4.13 (t, J = 4Hz, 1H, H-3′), 4.30 (s, 1H, H-4′), 4.62 (m, 1H, H-2′), 4.71 (broad s, 2H, N⁶-CH₂Ph), 5.53 (d, J = 7 Hz, 1H, OH-2′), 5.73 (d, J = 4 Hz, 1H, OH-3′), 5.96 (d, J = 8 Hz, 1H, H-1′), 7.30 (m, 5H, Phenyl), 8.25 (s, 1H, H-2), 8.42 (s, 1H, H-8), 8.55 (broad s, 1H, N⁶H-CH₂Ph), 8.86 (t, J = 5 Hz, 1H, NH-Et). Mass Spectrum (CI-NH₃): m/z 399 (MH⁺, base).

Inosine-5′-N-ethyluronamide (NECI)

2′,3′-O-Isopropylideneinosine-5′-uronic acid (20 mg, 62 μmol) (10), 1-[3-(dimethylamino)propyl]-3-ethylcarbodiimide hydrochloride (25 mg, 130 μmol) and N-hydroxysuccinimide (13 mg, 112 μmol) were dissolved in a minimum volume of dimethylformamide. Ethylamine (70% in water, 7 μl) was added and after 1 hr of stirring, was cooled to 0 °, and was precipitated with water to give 14 mg (65% yield). The product (10 mg, 29 μmol) was heated in 1 N HCl for 2 hr at 60 °. After cooling and neutralizing with NaHCO₃, it was purified twice using reverse phase SepPak cartridges with water as eluant. Lyophilization of the fraction afforded 6.95 mg (78% yield) of an amorphous solid Melting point, 168°C (d). ¹H-NMR (in Me₂SO-d₆) was as follows: δ 1.03 (t, J = 7 Hz, 3H, CH₃), 3.17 (m, 2H, CH₂), 4.15 (broad s, 1H, H-3′), 4.30 (s, 1H, H-4′), 4.54 (m, 1H, H-2′), 5.61 (broad s, 1H, OH), 5.68 (broad s, 1H, OH), 5.96 (d, J = 7 Hz, 1H, H-1′), 8.08 (s, 1H, H-2), 8.39 (s, 1H, H-8). Mass spectrum (CI-NH₃): m/z 310 (MH⁺, base).

Cell culture and membrane preparation

CHO cells stably expressing the rat A₃ receptor (3) were grown in F-12 medium containing 10% fetal bovine serum and penicillin/streptomycin (100 units/ml and 100 μg/ml, respectively) at 37 ° in a 5% CO₂ atmosphere. When cells had reached confluency, they were washed twice with 10 ml of ice-cold lysis buffer. After addition of 5 ml of lysis buffer, cells were mechanically scraped and homogenized in an ice-cold Dounce homogenizer (20 strokes by hand). The suspension was centrifuged at 43,000 × g for 10 min. The pellet was resuspended in the minimum volume of ice-cold 50 mM Tris/10 mM GCl₂/1 mM EDTA (pH 8.26 at 5°) buffer required for the binding assay and homogenized in a Dounce homogenizer. Typically, six to eight 175-cm² flasks were used for a 48-tube assay. ADA was added to a final concentration of 3 units/ml, and the suspension was incubated at 37 ° for 15 min; the membrane suspension was subsequently kept on ice until use. When large batches (~100 flasks) were processed homogenization was performed with a Polytron (Brinkman, Luzern, Switzerland), and further work-up was as described above. The preparation was stored at −70 ° and retained its [¹²⁵I] APNEA binding properties for at least 1 month.

Radioligand binding assay

Binding of [¹²⁵I]APNEA to CHO cells stably transfected with the rat A₃ receptor clone was performed essentially as described (6). Assays were performed in 50/10/1 buffer in glass tubes and contained 100 μl of the membrane suspension, 50 μl of [¹²⁵I] APNEA (final concentration 0.5 nM), and 50 μl of inhibitor. Inhibitors were routinely dissolved in Me₂SO and were then diluted with buffer; final Me₂SO concentrations never exceeded 1%. This concentration did not influence [¹²⁸I]APNEA binding. Incubations were carried out in duplicate for 1 hr at 37 ° and were terminated by rapid filtration over Whatman GF/B filters, using a Brandell cell harvester (Brandell, Gaithersburg, MD). Tubes were washed three times with 3 ml of buffer. Radioactivity was determined in a Beckman γ 5500B γ-counter. Nonspecific binding was determined in the presence of 40 μM R-PIA. K_i values were calculated according to Cheng and Prusoff (11), assuming a K_d for [^l25I]APNEA of 17 nM (3).

The level of nonspecific binding with [¹²⁵I]APNEA in transfected CHO cells was 20–30%. There was some variability in the Hill coefficients (range from 0.8 to 1.2). Untransfected CHO cells displayed a low level of binding displacable by 100 μM R-PIA (at 5 nM [¹²⁵I]APNEA, only 5–10% of the level of specific A₃ binding in transfected cells), but this binding did not have pharmacological characteristics of adenosine receptors.

Binding of [³H]PIA to A₁ receptors from rat brain membranes and of [³H]CGS 21680 to A_2a receptors from rat striatal membranes was performed as described previously (9).

Adenylate cyclase assay

Adenylate cyclase was assayed in membranes from CHO cells stably expressing the rat A₃ receptor, prepared as above, using a modification of previously reported methods (18). Maximal inhibition of adenylate cyclase activity corresponded to ~40% of total activity under conditions of stimulation (typically by 6–8-fold) in the presence of 1 μM forskolin.

Membranes were resuspended in 75 mM Tris, 200 mM NaCl, 1.25 mM MgCl2, pH 8.12, at 4°C (TNM buffer) to give a final concentration of 0.1 mg/ml, and 2 units/ml adenosine deaminase was added. For the cyclase assay, the TMN buffer was supplemented with 140 μM dATP, 5 μM GTP, 30 units/ml creatine kinase, 5 mM creatine phosphate, 2.2 mM dithiothreitol, 100 μM papaverine, and 1.5 μCi of [α-³²P] ATP. Then 40 μl of the membrane suspension, 40 μl of the supplemented buffer, and 20 μl of a solution of forskolin and the test compound (initially dissolved in Me₂SO then diluted in TMN buffer) were combined and incubated at 30 ° for 15 min, followed by termination by addition of a stop solution containing 20,000 cpm/ml [³H]cyclic AMP. The final concentration of Me₂SO did not exceed 1%, which had no effect on adenylate cyclase. Labeled cyclic AMP was isolated by chromatography on Dowex 50 and alumina columns, and ³²P was measured using scintillation counting.

Molecular modeling

Structures were built using Quanta (version 3.3; Polygen, Waltham, MA), and molecular mechanics calculations were performed with CHARMm (version 21.2; Harvard College, Boston, MA), running on a Silicon Graphics Indigo XZ 4000 workstation. An A₃ receptor model was built with the model of the A₁ receptor we recently proposed as a starting point (12). The latter model is based on the well defined structure of the seven-transmembrane domain protein bacteriorhodopsin and assumes that the spatial orientation of the α-helical domains of G protein-coupled receptors is similar to that of bacteriorhodopsin. An initial structure for the A₃ receptor was generated by copying the backbone coordinates (and side chain coordinates where applicable) of the relevant residues in the A₁ model to their A₃ counterparts. Initial bad contacts within the individual helices were relieved by energy minimization using 200 steps of steepest descents, followed by 200 steps of an adopted basis Newton Raphson. Subsequently, the model was minimized using 500 steps of steepest descents, followed by an adopted basis Newton Raphson minimization until the rms energy gradient was less than 0.01 k_cal/mol Å, while keeping the backbone positions fixed. As a last step, the model was energy-minimized under the same conditions with backbone constraints of 5 k_cal/mol. Adenosine and other ligands were docked into the presumed binding site starting with the orientation that CPA assumes in the A₁ model (12), followed by a two-step energy minimization as described above. Calculations were performed using an extended atom approach and without explicit water, however, a distance-dependent dielectric constant was used to account for the screening effect of solvent.

Results and Discussion

SAR for adenosine derivatives

Binding data for a variety of adenosine derivatives, as well as a number of nucleosides having bases other than adenine are given in Table 1. Representative binding curves for three purine-substituted adenosine derivatives are shown in Fig. 1. The affinity of adenosine itself cannot be accurately determined in this binding assay, due to the presence of adenosine deaminase, which is required to degrade endogenously generated adenosine. Hence, it is not possible to directly compare the affinities of adenosine with the derivatives tested here. The affinity of adenosine has previously been estimated at 30 μM (3), but this value should be taken as only a rough approximation.

TABLE 1. Affinities of selected compounds at A₁ A_2a and A₃ receptors, indicated as either K_i, (nM) or percent displacement at a concentration of 100 μM, unless otherwise indicated.

Values expressed as means ± S.E. were all measured in this study (three to five experiments). K_i, values at A₁ and A_2a receptors provided without S.E. are taken from the literature as indicated.

Compound	A1 affinitya	A2a affinityb	A3 affinityc	Reference
Purine and 5′-modified adenosines
1. ADACd	0.85 nM	210 nM	281 ± 51 nM	(22)
2. R-PIA	1.2 nM	124 nM	158 ± 52 nM	(23)
3. S-PIA	49.3 nM	1,820 nM	920 ± 311 nM	(23)
4. CPA	0.59 nM	462 nM	240 ± 36 nM	(23)
5. CHA	1.3 nM	514 nM	167 ± 26 nM	(23)
6. N6-Phenyladenosine	4.62 nM	663 nM	802 ± 279 nM	(23)
7. N6-Benzyladenosine	120 nM	285 nM	120 ± 20 nM	(23)
8. N6-Phenethyladenosine	12.7 nM	161 nM	240 ± 58 nM	(24)
9. N6-Dimethyladenosine	10,000 nM	28,900 ± 8,500 nM	32,500 ± 5,100 nM	(10)
10. DPMA	142 nM	4.4 nM	3,570 ± 1,700 nM	(25)
11. N6-(2-Sulfoethyl)adenosine	41%	0%	32,400 ± 7,600 nM	(9)
12. N6-(p-Sulfophenyl)adenosine	74 nM	8,900 nM	526 ± 142 nM	(9)
13. N6-3-(p-Sulfophenyl)propyladenosine	610 nM	3,840 nM	844 ± 67 nM	(9)
14. N6-4-(p-Sulfophenyl)butyladenosine	432 nM	11,300 nM	808 ± 116 nM	(9)
15. 1-Deaza-2-chloro-N6-CPA	1.6 nM	13,200 nM	770 ± 234 nM	(26)
16. 2-Chloroadenosine	9.3 nM	63 nM	1,890 ± 900 nM	(23)
17. 2-Chloro-N6-CPA	0.6 nM	950 nM	237 ± 71 nM	(27)
18. 2-(Phenylamino)adenosine	560 nM	119 nM	4,390 ± 1,170 nM	(23)
19. CGS 21680	2,600 nM	15 nM	584 ± 32 nM	(28)
20. NECA	6.3 nM	10.3 nM	113 ± 34 nM	(23)
21. N6-CyclohexylNECA	0.43 nM	170 nM	16.0 ± 5.4 nM	(29)
22. N6-BenzylNECA	87.3 ± 13.9 nM	95.3 ± 24.6 nM	6.8 ± 2.5 nM
23. N6-DimethylNECA	9,600 nM	13,500 ± 3,600 nM	2,260 ± 490 nM	(10)
24. N6-Benzyl-N6-methyladenosine	7,600 ± 1,900 nM	40,100 ± 6,200 nM	78.4 ± 4.6%
25. 8-Bromoadenosine	41.5 ± 3.2%	22,700 ± 5,100 nM	31.3 ± 6.0%
26. 3-Deazaadenosine	21,500 nM	59,800 ± 4,600 nM	61,700 ± 34,500 nM	(30)
27. 7-Deazaadenosine (tubercidine)	>100,000 nM	48.3 ± 0.4%	38.9 ± 17.7%	(30)
28. Adenosine-N1-oxide	246 ± 31 nM	328 ± 60 nM	3,090 ± 1,910 nM
29. NECA-N1-oxide	154 ± 20 nM	101 ± 19 nM	468 ± 58 nM
30. N6-Benzyladenosine-N1-oxide	864 ± 88 nM	8,530 ± 1,250 nM	7,250 ± 1,680 nM
ribose-modified adenosines
31. β-L-Adenosine	29,000 ± 4,700 nM	25.4 ± 1.1%	9.5 ± 4.2%
32. α-D-Adenosine	350,000 nM	128,000 ± 25,000 nM	14.2 ± 6.5%	(31)
33. 2′-Deoxyadenosine	30.9 ± 8.0%	38.9 ± 2.9%	28.3 ± 2.3%
34. 2′-O-Methyladenosine	29.4 ± 7.5%	49.0 ± 5.0%	42.9 ± 9.4%
35. 3′-Deoxyadenosine (cordycepin)	5.8 ± 2.8%	26.3 ± 3.4%	32.7 ± 2.0%
36. 5′-Deoxyadenosine	269 ± 135 nM	596 ± 54 nM	2,830 ± 460 nM
37. 5′-Deoxy-5′-aminoadenosine	42,700 ± 6,000 nM	38,500 ± 3,800 nM	20.6 ± 2.2%
38. 5′-Deoxy-5′-methylthioadenosine	281 nM	1,100 nM	1,420 ± 530 nM	(23)
39. 5′-Deoxy-5′-isobutylthioadenosine	1,140 ± 130 nM	6,890 ± 1,750 nM	3,630 ± 360 nM
40. S-Adenosylmethionine	675 ± 87 nM	2,780 ± 250 nM	2,470 ± 450 nM
41. AMP	–*	57.5 ± 4.0%	17.2 ± 6.3%
42. Adenine-β-D-arabinofuranoside	20.2 ± 8.4%	26.0 ± 8.4%	23.7 ± 3.8%
43. β-D-Psicofuranosyladenine	36.1 ± 4.9%	51.5 ± 7.4%	21.1 ± 0.9%
non-adenosine nucleosides
44. Xanthosine	9.1 ± 2.4%	8.5 ± 2.5%	23.4 ± 8.8%
45. Uridine	14.3 ± 6.9%	2.8 ± 5.2%	18.9 ± 2.8%
46. Thymidine	23.4 ± 2.5%	1.7 ± 3.4%	21.3 ± 4.9%
47. Cytidine	18.0 ± 1.2%	16.0 ± 1.5%	24.5 ± 10.2%
48. Inosine	16,700 ± 2,900 nM	50,000 ± 12,700 nM	45,100 ± 38,800 nM
49. Guanosine	27,800 ± 9,600 nM	85,100 ± 15,700 nM	98,500 ± 28,700 nM
50. (4-Nitrobenzyl)-6-thioguanosine	15,000 ± 3,500 nM	48,500 ± 11,300 nM	40,700 ± 26,300 nM
51. 6-Thioguanosine	44.2 ± 2.3%	27.7 ± 5.8%	44.8 ± 18.1%
52. 6-Thiopurine riboside	61.2 ± 3.9%	33.6 ± 3.3%	41.9 ± 5.0%
53. NECI	43.7 ± 10.3%	30.6 ± 2.3%	5,000 ± 1,150 nM
non-xanthine adenosine antagonists
54. CP 66713	270 nM	21 nM	29.7 ± 7.8%	(32)
55. CGS 15943	21 nM	3.3 nM	38.0 ± 14.5%	(33)
56. IQA	1,600 nM	1,400 nM	32.6 ± 10.8%	(8)
57. 9-Ethyl-N6-cyclopentyladenine	440 nM	17,000 nM	30.4 ± 9.1%	(34)
58. EHNA	455 ± 10 nM	59.6 ± 2.8%	57.5 ± 14.3%
59. Amiloride	11,000 nM	17,000 nM	22.0 ± 3.5%	(35)
60. Xanthine	298,000 nM	16.2 ± 2.6%	14.0 ± 7.9%	(36)
61. 1-MethylX	11,400 nM	36,200 nM	11.1 ± 1.6%	(23)
62. 3-MethylX	35,000 nM	38.0 ± 0.9%	18.1 ± 6.7%	(37)
63. 7-MethylX	52.3 ± 7.9%	37.7 ± 4.9%	16.4 ± 9.6%
64. 9-MethylX	26.6 ± 3.2%	16.1 ± 1.9%	22.8 ± 9.5%
65. 1,3-DimethylX (theophylline)	8,500 nM	25,000 nM	23.1 ± 9.5%	(23)
66. 1,7-DimethylX (paraxanthine)	30,000 nM	19,400 ± 3500 nM	15.5 ± 12.1%	(38)
67. 1,9-DimethylX	29.4 ± 1.6%	6.0 ± 6.3%	17.0 ± 7.9%
68. 3,7-DimethylX (theobromine)	83,400 nM	187,000 nM	19.9 ± 7.1%	(23)
69. 3,9-DimethylX	19.7 ± 7.9%	4.2 ± 5.9%	19.0 ± 6.8%
70. 1-Methyl-3-ButylX	7,000 nM	16,000 nM	30.1 ± 12.4%	(23)
71. 1,3-DibutylX	500 nM	2,930 ± 700 nM	143,000 ± 29,000 nM	(39)
72. 1,3-DihexylX	1,260 ± 90 nM	14.3 ± 3.0% (10 μM)	9.2 ± 6.5% (10 μM)
73. 1,3-DibenzylX	2,000 nM	3.61 ± 0.94% (10 μM)	20.3 ± 8.5% (10 μM)	(39)
74. 1,3,7-TrimethylX (caffeine)	29,000 nM	48,000 nM	30.1 ± 12.4%	(23)
75. 1,3,9-TrimethylX (isoC)	>1,000,000 nM	14.4 ± 5.7%	13.2 ± 12.4%	(40)
76. 2-Thio-3-propylX	26,100 ± 1,500 nM	32,500 ± 4,800 nM	27.7 ± 11.3%
7-substituted alkylxanthines
77. 7-BenzylT	6,000 nM	46,000	29.7 ± 0.2%	(41)
78. 7-β-HydroxyethylT	105,000 nM	17,400 ± 900	21.1 ± 13.3%	(39)
79. T-7-Riboside	27,000 ± 3,200 nM	n.t.	89,400 ± 13,400 nM
80. 1,3-DipropylX-7-riboside	15,900 ± 1,800 nM	32.0 ± 1.1%	81,200 ± 10,700 nM
81. 1,3-DibutylX-7-riboside	4,190 ± 1030 nM	19,500 ± 4,200 nM	6,030 ± 2,320 nM
8-substituted alkylxanthines
82. 8-PhenylT	86 nM	850 nM	12.0 ± 6.0%	(23)
83. 8-CyclopentylT	11 nM	1,400 nM	38.7 ± 2.5%	(23)
84. 8-Cyclopentyl-1-propylX	226 ± 37 nM	48,700 ± 5,000 nM	22.6 ± 7.7%
85. 8-Cyctopentyl-1,3-dipropylX	0.46 nM	340 nM	18.7 ± 2.9% (10 μM)	(42)
86. 8-CyclohexylC	28,000 nM	10,400 ± 2,600 nM	35.2 ± 1.8%	(37)
87. 8-ChloroT	30.2 ± 6.7%	24.7 ± 3.9%	16.8 ± 9.5%
88. XAC	1.2 nM	63 nM	7.1 ± 0.9%	(43,22)
89. 8-(3-Chlorostryryl)C (CSC)	28,200 nM	54 nM	4.2 ± 5.1% (10 μM)	(46)
90. 8-Sulfophenyl-1,3-dipropylX	140 nM	790 nM	21.9 ± 6.2%	(47)

^aDisplacement of [³H]PIA (or [³H]CHA) binding from rat brain membranes.

^bDisplacement of [³H]CGS 21680 (or [³H]NECA in the presence of 50 nM CPA) from rat striatal membranes.

^cDisplacement of [¹²⁵I]APNEA binding from membranes of CHO cells stably transfected with the rat A₃-cDNA.

^dThe abbreviations used are: ADAC, N⁶-[4-[[[4-[[[(2-aminoethyl)amino]carbonyl]methyl]anilino]carbonyl]methyl]phenyl]adenosine; DPMA, N⁶-[2-(3,5-dimethyoxyphenyl)-2-(2-methylphenyl)ethyl]adenosine; IQA, imidazo[4,5-c]quinolin-4-amine; EHNA, erythro-9-(2-hydroxy-3-nonyl)adenine; X, xanthine; T, theophylline; C, caffeine; n.t., not tested.

^*AMP displayed an extremely high slope factor (3.62 ± 0.39 μM) in A₁ displacement. The apparent K_i, was 47.5 ± 6.5 μM.

Fig. 1.

Displacement by three purine-modified adenosine analogues of specific binding of [¹²⁵I]APNEA (0.5 nM) from membranes from CHO cells transfected with rat A₃ receptors. The assay was carried out as described under Materials and Methods. Competitors were as follows (Hill coefficients in parentheses): ◆,2-chloroadenosine (1.10); △, 1-deaza-2-chloro-N⁶-cyclotopentyladenosine (1.07); and ●, inosine (1.03). The curves are representative of single experiments in which each point is determined in triplicate.

As is the case for A₁ receptors, the most potent compounds at the A₃ receptor are N⁶-substituted and/or 5′-N-ethylcarboxamide-substituted adenosine derivatives. There are, however, profound differences between the N⁶ topology of A₁ and A₃ receptors. In general, the affinities of N⁶-substituted adenosines are much higher at A₁ than at A₃ receptors, with a few notable exceptions. It appears that the selectivity ratio is proportional to A₁ affinity; the higher the affinity at A₁, the higher the selectivity for A₁ versus A₃. This indicates that at the A₁ receptor, N⁶ substituents can interact with a receptor region that is not present in the A₃ receptor. The stereoselectivity characteristic of the N⁶ region of A₁ and A₂ receptors is maintained at A₃, albeit that R-PIA is only 6-fold more potent than S-PIA.

In a series of N⁶-aryl(alkyl)-substituted compounds, N⁶-benzyladenosine is more potent (K_i, 120 nM) than N⁶-phenyladenosine (K_i, 802 nM) or N⁶-phenethyladenosine (K_i, 240 nM). This is surprising given the poor affinity of the former compound at A₁ (K_i, 120 nM) and at A_2a (K_i, 285 nM) receptors, whereas the phenyl and phenethyl derivatives have affinities in the lower nanomolar range at A₁ receptors (Table 1). Thus, N⁶-benzyladenosine is essentially nonselective.

Another significant difference with A₁ and A_2a receptors is that introduction of a p-sulfo group in N⁶-phenyladenosine slightly enhances affinity (N⁶-(p-sulfophenyl)adenosine, K_i, 526 nM), in sharp contrast to A₁ and A_2a receptors, where a sulfo group drastically reduces affinity (Table 1). Two other N⁶-sulfo derivatives, N⁶-3-(4-sulfophenyl)propyladenosine and N⁶-4-(4-sulfophenyl)butyladenosine have affinities in the same range (Table 1), but the 2-sulfoethyl derivative, which has a shorter N⁶ substituent, is considerably less potent (K_i, 32.4 μM). Because the polar sulfo group is apparently better tolerated at A₃ than at A₁ and A_2a receptors, sulfo substitution shifts affinity in the direction of A₃ selectivity.

The A₁-selective N⁶-cycloalkyl derivatives, CHA and CPA, are also among the more potent compounds at A₃ (K_i, 167 and 240 nM, respectively), as is the N⁶-functionalized congener N⁶-[4-[[[4-[ [ [(2-aminoethyl)amino]carbonyl]methyl]anilino]carbonyl]methyl]phenyl]adenosine (K_i, 281 nM). The A₂-selective N⁶-substituted compound, N⁶-[2-(3,5-dimethoxyphenyl)-2-(2-methylphenyl)ethyl]adenosine, has moderate potency at A₃ (K_i, 3.57 μM).

The affinity of N⁶-dimethyladenosine is quite low (K_i, 32.5 μM), similar to the poor affinity of this compound at A₁ and A_2a receptors. The affinity of N⁶-benzyl-N⁶-methyladenosine and N⁶-dimethylNECA is also considerably lower than the parent compounds N⁶-benzyladenosine and NECA but not as drastically reduced as at A₁ and A₂ receptors. Thus, although disubstitution at N⁶ reduces affinity, it enhances selectivity for A₃ versus A₁ and A_2a receptors (e.g., N⁶-dimethylNECA is 4-fold selective versus A₁ and 6-fold versus A_2a).

Similar to other adenosine receptor subtypes, NECA (5′-N-ethylcarboxamide adenosine) is relatively potent (K_i, 113 nM). However, unlike A₁ and A_2a receptors (10), the effects of N⁶ and C5′ substitutions appear to reinforce each other. Thus, the 5′-N-ethylcarboxamide of CHA is considerably more potent than either CHA (K_i, 167 nM, 10-fold) or NECA (K_i, 113 nM, 7-fold), and, with a K_i of 16 nM, it is a highly potent compound at A₃ receptors. Likewise, N⁶-dimethylNECA is 14-fold more potent than N⁶-dimethyladenosine. These findings prompted us to synthesize the NECA analogue of N⁶-benzyladenosine. Because N⁶-benzyladenosine is more or less equipotent at all three subtypes, and the 5′-N-ethyl substituent apparently boosts affinity at A₃ receptors more than at A₁ and A_2a, it was expected that N⁶-benzylNECA would be a potent and somewhat A₃-selective agonist. Indeed, N⁶-benzylNECA has the highest A₃ affinity of all compounds tested in this study, with a K_i value of 6.8 nM, 18-fold more potent than the parent compound N⁶-benzyladenosine. This is also the most A₃-selective compound found in the present study (13-fold versus A₁ and 14-fold versus A_2a). This compound may prove useful in the pharmacological characterization of A₃ receptors, e.g., as a radioligand and as a lead for the further development of more selective A₃ agonists.

Certain C2 modifications may result in A_2a-selective agonists (1). At A₃ receptors, 2-chloroadenosine and 2-phenylaminoadenosine are of intermediate potency (K_i, 1.9 and 4.4 μM, respectively). An N⁶-substituted derivative (2-chloro-N⁶-CPA, K_i, 237 nM) and one bearing a 5′-N-ethylcarboxamide group (CGS 21680, K_i, 584 nM) are more potent C2-substituted derivatives. Thus, 2 substitution is tolerated to a degree at the A₃ receptor.

With regard to modifications of the ribose moiety, both the L-enantiomer and the α-anomer of adenosine are virtually inactive (IC₅₀ ≫ 100 μM), similar to other adenosine receptors (13). Psicofuranosyladenine, which contains an extra CH₂OH group at Cl′, is also very weak. 2′-Deoxy-, 2′-O-methyl, and 3′-deoxyadenosine all have low affinity (IC₅₀ > 100 μM), and inversion of the stereochemistry of the 2′-OH group (adenine-β-D-arabinofuranoside) similarly results in a low affinity compound. Thus, the presence of the 2′-OH in the S-configuration and the 3′-OH appears to be essential for high affinity. This has also been shown for other adenosine receptor subtypes (reviewed in Ref. 1). The 5′ position is more amenable to modifications than the 2′ or the 3′ position. The 5′-deoxy derivative of adenosine is moderately potent (K_i, 2.83 μM), and, as stated, the 5′-N-ethylcarboxamide derivative (NECA) is one of the more potent compounds tested (K_i, 113 nM). Some 5′-deoxyadenosine derivatives, including those with methylthio-, isobutylthio-, and methionine substituents have affinities in the lower micromolar range, whereas 5′-deoxy-5′-aminoadenosine and AMP (bearing a 5′-phosphate group) are virtually inactive IC₅₀ > 100 μM. This parallels the affinities of these compounds at A₁ and A_2a receptors (Table 1). In all, ribose SAR for the A₃ receptor is comparable with A₁ and A_2a receptors, suggesting that the ribose domain may be quite similar in all three receptor subtypes.

Some adenosine derivatives not commonly used in adenosine receptor studies were tested. 6-Thiopurine riboside and 8-bromoadenosine both have low affinity (IC₅₀ > 100 μM) at A₃, similar to A₁ and A_2a receptors. A bulky 8 substituent forces the ribose moiety in a predominantly syn conformation, which is a likely explanation for the inactivity of 8-bromoadenosine. The same has been shown for A₁ receptors (7). 7-Deazaadenosine has a IC₅₀ ≫ 100 μM, which indicates the importance of N⁷. 3-Deazaadenosine is slightly more potent, with a K_i of 62 μM. l-Deazaadenosine was not available for testing, but 1-deaza-2-chloro-CPA, (K_i, 770 nM) is only 3-fold less potent than 2-chloro-CPA (K_i, 237 nM). This suggests that the presence of N¹ is not crucial, and it is consistent with the profile at other adenosine receptor subtypes, where N¹-deazaadenosine > N³-deazaadenosine > N⁷-deazaadenosine (reviewed in Ref. 1). The N¹-oxides of adenosine, NECA, and N⁶-benzyladenosine are moderately potent compounds (K_i, 3.09, 0.47, and 7.25 μM, respectively) but less potent than at A₁ and A_2a receptors (Table 1).

Of the unsubstituted nonadenine nucleosides tested, only inosine (K_i, 45 μM) and guanosine (K_i, 99 μM) show some affinity for the A₃ receptor. The 5′-N-ethylcarboxamide derivative of inosine (NECI) is more potent (K_i, 5 μM), which is consistent with the affinity-enhancing effect of the 5′-carboxamido substituent of NECA. NECI is also selective for A₃ receptors, having an IC₅₀ larger than 100 μM at A₁ and A_2a receptors. An adenosine transport inhibitor, (4-nitrobenzyl)-6-thioguanosine (K_i, 41 μM), is slightly more potent than the parent compound, guanosine.

SAR for nonadenosine derivatives

According to Zhou et al. (3), xanthines do not appear to displace [¹²⁵I]APNEA binding to A₃ receptors. We first tested a variety of nonxanthines known to act as antagonists at A₁ and/or A_2a receptors, including CGS 15943, CP 66713, 1H-imidazo[4,5-c]quinolinamine, 9-ethylcyclopentyladenine, and amiloride (Table 1). None of these appeared to be particularly potent, with K_i values in the range of 100 μM or larger. The adenosine deaminase inhibitor, erythro-9-(2-hydroxy-3-nonyl)adenine, was somewhat more potent with an IC₅₀ slightly better than 100 μM (57.5% displacement at 100 μM), erythro-9-(2-Hydroxy-3-nonyl)adenine was of moderate affinity at A₁ receptors.

We therefore turned to a more detailed study of xanthine SAR (Table 1) than in the original paper (3). Xanthine was found to be a very weak displacer of [¹²⁵I]APNEA binding (14% at 100 μM). Substitutions at the 1 and the 3 positions enhance affinity. Compared with theophylline (1,3-dimethylxanthine, 23.1% at 100 μM), 1,3-dibutylxanthine is more potent (K_i, 143 μM). 1,3-Dihexylxanthine (9.2% at 10 μM) and 1,3-dibenzylxanthine (20.3% at 10 μM) also seem more potent than theophylline, but their limited solubility precludes direct comparison. This profile is similar, but not identical, to other adenosine receptor subtypes, where propyl and butyl are optimal, and benzyl is slightly less potent (1). Still, even the most potent of these 1,3-substituted xanthines is quite weak at rat A₃ receptors.

Due to the limited solubility of even the most potent xanthine, it was not feasible to compare precisely the degree of maximal displacement of the xanthines and the adenosine derivatives. Unfortunately, the more water-soluble xanthines, such as XAC (positively charged at physiological pH) and 8-(p-sulfophenyl)-1,3-dipropylxanthine (negatively charged at physiological pH) did not bind appreciably to rat A₃ receptors.

Unlike A₁ and A_2a receptors, 8 substituents do not appear to contribute much to affinity and none of the 8-substituted xanthine derivatives tested is particularly potent (Table 1). This is surprising, because at A₁ and A_2a receptors, affinities of N⁶-substituted adenosines and xanthines similarly substituted at the 8 position closely parallel each other, suggesting that N⁶ and C8 substituents interact with the same receptor domain (1, 14). Clearly, this is a subject for further investigation. Substitutions at the 7 position appear to be tolerated, in contrast to A₁ receptors, where 7 substituents tend to diminish affinity (7); e.g., both caffeine and 7-benzyltheophylline are slightly more potent at A₃ receptors than the 7-unsubstituted parent compound, theophylline. The A_2a-selective antagonist 8-(3-chlorostyryl)caffeine (47) did not inhibit binding at rat A₃ receptors.

Molecular modeling and prediction of affinity of xanthine-7-ribosides

Like other G protein-coupled receptors, the amino acid sequence of A₃ receptors contains seven hydrophobic stretches of approximately 25 residues that are believed to traverse the cell membrane as α-helices (3). No detailed three-dimensional structures (X-ray or NMR) for any of the G protein-coupled receptors are known, but it is now well accepted that the structure of bacteriorhodopsin, which has been solved by cryo-electron microscopy (15), is a suitable starting point for the modeling of G protein-coupled receptors (16). We have recently described a model for the A₁ adenosine receptor (12), and, here, we present a similar model for the A₃ receptor. Details of the building of the model are given under Materials and Methods.

A close-up of the proposed binding site of the A₃ receptor with N⁶-benzyladenosine as the ligand is shown in Fig. 2. This model is based on pharmacological observations and analogies with the A₁ receptor and is consistent with the SAR described above. N⁶-Benzyladenosine was chosen as a typical agonist for the A₃ receptor, because it is relatively potent (Table 1). In the A₁ model (12), CPA was chosen as a typical high affinity A₁- selective agent for the purpose of docking to the binding site. There, CPA is coordinated by two histidine residues in helices VI and VII, whose involvement is in full agreement with both chemical modification studies (17) and site-directed mutagenesis (18). It should be mentioned here that another quite different A₁ receptor model has been developed in which no specific interaction with histidine residues was proposed (48).

Fig. 2.

The proposed adenosine binding site of the A₃ receptor, with N⁶-benzyladenosine as the ligand. The ligand is shown as a ball-and-stick representation (thick bonds) and the receptor is shown in liquorice bond style (thinner bonds). The 2′- and 3′-OH may form a hydrogen bond with His²⁷⁴, and 5′-OH can form a hydrogen bond with Ser²⁷⁷.

There are a number of similarities as well as some substantial differences between our models for the A₁ and A₃ receptors. As in the A₁ model, the agonist ligand is present in the anti conformation (χ, the torsion angle of the glycosidic bond is 76°), consistent with earlier modeling and NMR studies (19). The model proposes several points of interaction between the receptor and the ribose moiety. Hydrogen bonds could be formed between the 2′- and 3′-OH groups and His²⁷⁴ in helix VII and between the 5′-OH group and Ser²⁷⁷ in helix VII. There are equivalent interactions with His²⁷⁸ and Ser²⁸¹ in the A₁ receptor model. Thus, the ribose binding domain seems to be quite similar for A₃ and A₁ receptors, in good agreement with the agonist SAR described here; the same appears to be true for A_2a receptors (45). In the present model, the side chain of Phe¹⁸⁴ (helix V) is located near the glycosidic bond of receptor-bound N⁶-benzyladenosine. The same Phe residue and Thr⁹⁶ (helix III) appear to be in proximity to the C5′ region.

A major structural difference is that the A₃ receptor does not contain the histidine residue in helix VI that is common to all A₁ and A₂ receptors cloned so far (20) and that has been shown to have an effect on both agonist and antagonist binding to A₁ receptors (18). In particular, the His of helix VI has been shown to be important for antagonist affinity, a finding that suggests a linkage between the absence of this His residue and the lack of high affinity binding of antagonists such as XAC at rat A₃ receptors. In the model of agonists binding to A₁ receptors (12), this residue forms a hydrogen bond with N⁶-H. In the A₃ receptor, a serine residue (Ser²⁴⁹) is found in the analogous position, and it could be argued that serine could serve a similar function as a hydrogen bond acceptor. However, due to a different orientation of the purine moiety in the A₃ binding site, this serine seems to be too far from N⁶-H (~7 Å) to be able to form this bond. The reason for the different orientation is as follows. The majority of the amino acid residues that are different between A₁ and A₃ receptors occur in the immediate vicinity (within 5 Å) of the agonist ligand, and many of the A₃ residues are considerably more bulky (e.g., Phe⁹⁵) than their A₁ counterparts. This results in a ligand binding environment (for both the purine and ribose domains) that is much more constricted than is the case for the A₁ receptor (12). In Fig. 3 the binding orientations of CPA to the A₁ receptor and of N⁶-benzyladenosine to the A₃ receptor are shown to illustrate these differences. They provide a tentative explanation for the apparent low affinity of xanthines and nonxanthine A₁/A₂ antagonists. In addition, the ribose moiety that likely serves to anchor the native ligand adenosine to the receptor is absent in these compounds.

Fig. 3.

Comparison of CPA as bound in the A₁ model (left) and N⁶-benzyladenosine in the A₃ model (right). Color coding is as follows: yellow, ligand; blue, H.I; orange, H.II; green, H.III; red, H.IV; purple, H.V; white, H.VI; pink, H.VII. The A₁ model is identical to that proposed in Ref. 12. See Results for a discussion of specific residues proposed to be in proximity to the bound N⁶-benzyladenosine.

As in the A₁ model, a hydrophobic pocket directed toward the extracellular space, capable of accommodating large N⁶ substituents, is present. Fig. 2 shows the binding environment of the benzyl substituent of N⁶-benzyladenosine. Indeed, a number of N⁶-substituted adenosine derivatives, including the long chain functionalized congener N⁶-[4-[[[4-[[[(2-aminoethyl)amino]carbonyl]methyl]anilino]carbonyl]methyl]-phenyl]adenosine, have considerable affinity for the A₃ receptor (Table 1). Amino acid residues in proximity to the N⁶-benzyl ring according this model are the side chains of Phe¹⁸¹ (hydrophobic) and Asp¹⁷⁷ (anionic) and the backbone atoms of Tyr¹⁷⁸ (all in helix V). Due to the rather constricted agonist binding domain, the N⁶ region is much closer to the membrane surface than in the A₁ model. It might be hypothesized that the paucity of hydrophobic residues for an N⁶ substituent to interact with, because of the proximity of the N⁶ region to the membrane surface, accounts for the rather moderate affinity of N⁶-substituted adenosines, in comparison with the much higher affinities of a number of similar agonists at A₁ and A_2a receptors. This would also agree well with a polar sulfo substituent being tolerated close to the N⁶ region of A₃ but not A₁ or A_2a receptors. A second (hydrophobic) pocket is present adjacent to C2, which can accommodate fairly large C2 substituents. Again, this agrees well with the considerable potency of some C2-substituted agonists, e.g., CGS 21680 (Table 1). The backbone of Tyr²⁵⁶ (helix VI) is situated near the purine C2 position of N⁶-benzyladenosine.

It should be noted here that due to the relative scarcity of pharmacological and structural data for this novel receptor, of necessity, there are uncertainties in this model. The question about the involvement of Ser²⁴⁹ in binding of the purine moiety and whether the different agonist orientations in the A₁ and A₃ models will hold up should be addressed experimentally. Chemical modification studies, site-directed mutagenesis, and more SAR work will be needed to further define essential characteristics of the ligand-receptor interaction, which, in turn, may lead to a more refined model. At this stage, the A₃ model should mainly be seen as a starting point to generate ideas for experiments, and its usefulness is illustrated by the following.

The notion from the modeling studies that the primary point of interaction between receptor and ligand appears to be the ribose moiety of agonists, combined with the observation that 7 substitution apparently is tolerated, prompted us to test some xanthine-7 ribosides that were previously synthesized (7). It was reasoned that the ribose hydroxyl groups might serve to anchor the xanthine nucleus to the receptor. Indeed, of the compounds tested, 1,3-dibutylxanthine-7 riboside (DBXR) was quite potent (K_i, 6μM), almost 25-fold more potent than the parent 1,3-dibutylxanthine, which was one of the most potent xanthines tested (Table 1). Affinity at A₁ and A_2a receptors is 4.19 and 19.5 μM, respectively, so this compound has only very moderate selectivity.

Fig. 4 illustrates a proposed model for DBXR binding to the A₃ receptor. According to this model, the 1,3-dialkyl substituents of DBXR are located in hydrophobic regions near the exofacial surface of the A₃ receptor. Specifically, the N¹ -butyl chain is located near the side chain of Tyr²⁵⁶ of helix VI, and the N³-butyl chain is located near the side chain of Phe¹⁸¹ of helix VI.

Fig. 4.

Proposed mode of binding of 1,3-dibutylxanthine-7-riboside to the A₃ receptor. Details are as for Fig. 1. The N¹-butyl chain is located near the side chain of Tyr²⁵⁶ of helix VI, and the N³-butyl chain is located near the side chain of Phe¹⁸¹ of helix VI.

Effects on adenylate cyclase

The effects of key compounds on the inhibition of adenylate cyclase in CHO cells stably expressing the rat A₃ receptor was measured (Fig. 5). Indeed, adenosine derivatives N⁶-cyclohexylNECA and N⁶-benzylNECA proved to be agonists at A₃ receptors, with full efficacy, as observed with R-PIA and NECA (data not shown). The maximal inhibition of adenylate cyclase elicited by N⁶-cyclohexylNECA was 45.7 ± 2.1%. The IC₅₀ determined for this compound was 1.30 ± 0.31 μM. N⁶-BenzylNECA was nearly as potent (IC₅₀ 1.61 μM) with a maximal inhibition of 47.6 ± 6.3%.

Fig. 5.

Inhibition of adenylate cyclase in membranes from CHO cells transfected with rat A₃ receptors. The assay was carried out as described under Materials and Methods. Each data point is shown as mean ± S.E. for four to seven determinations. Agents were as follows (number of separate experiments in parentheses): triangles, 1,3-dibutylxanthine-7-riboside (5); circles, N⁶-benzylNECA (7); and squares, N⁶-cyclohexyl-NECA (4). In these cell membranes, a K_D value for binding of [¹²⁵I]APNEA to A₃ receptors was 5.67 ± 0.73 nM with a B_max of 1.51 ± 0.40 pmol/mg protein.

At A₁ receptors, xanthine-7 ribosides have been shown to act as antagonists or partial agonists (7, 21). At rat A₃ receptors, 1,3-dibutylxanthine-7 riboside did inhibit adenylate cyclase, but the dose-response curve was more shallow than for the adenosine derivatives (only 20.9 ± 4.0% inhibition at 100 μM); thus, it appears to be a partial agonist. The inosine derivative NECI was so weak in inhibiting adenylate cyclase that a full-dose response was not able to be measured; at a concentration of 10⁻⁴ M, it inhibited adenylate cyclase by 9.4 ± 3.8% (n = 7). Although the majority of the compounds were not assayed in this functional assay, the rank order of potency paralleled the order of potency in displacing the specific binding of radioligand at A₃ receptors.

1,3-Dibutylxanthine at 100 μM neither antagonized the action of an adenosine agonist (N⁶-cyclohexylNECA) acting at A₃ receptors nor itself inhibited adenylate cyclase in the transfected CHO cells. Theophylline was also unable to antagonize the inhibition of adenylate cyclase elicited by N⁶-benzylNECA. N⁶-BenzylNECA alone had an IC₅₀ value of 1.35 ± 0.65 μM, with maximal inhibition of 28.8 ± 0.9% (100 μM). In the presence of 1 mM theophylline, the IC₅₀ value was 0.91 ± 0.1 μM, with a maximal inhibition of 34.2 ± 1.5%. Theophylline had no effect on the basal level of adenylate cyclase or on how much forskolin was able to stimulate.

Conclusions

N⁶-BenzylNECA was identified as the first highly potent and moderately A₃-selective agonist. Combined with the other SAR differences found between A₃ and A₁/A₂ affinity, such as polar substituents being tolerated in the N⁶ region, it should provide a good lead toward the development of even more potent and selective A₃ agonists.

Mutation experiments have shown that the His residue in the sixth transmembrane helix of both A₁ and A₂ receptors is involved in the high affinity binding of antagonists such as XAC (18). The complete inactivity of xanthines at rat A₃ receptors, which lack that His, is consistent with this model. This A₃ receptor molecular model has also been in part validated with the moderate affinity of 1,3-dibutylxanthine-7 riboside, which appears to act as a partial agonist. This study did not identify any A₃ antagonists among a wide range of the known A₁/A₂ receptor antagonists.

The predictions of the computer-generated model for the binding site must be tested through further efforts in ligand synthesis and modification of the receptor structure through site-directed mutagenesis of A₃ receptors. More elaborate SAR studies to further define optimal substituents for interaction with the N⁶ and C5′ regions of the A₃ adenosine receptor are currently underway.

ABBREVIATIONS

APNEA

N⁶-2-(4-aminophenyl)ethyladenosine

CGS 15943

9-chloro-2-(2-furyl)[1,2,4]triazolo[1,5-c]quinazolin-5-amine

CGS 21680

2-[4-(2-carboxyethyl)phenyl]ethylamino-5′-N-ethlcarboxamidoadenosine

CHA

N⁶-cyclohexyladenosine

CPA

N⁶-cyclopentyladenosine

DBXR

1,3-dibutylxanthine-7 riboside

NECA

5′-N-ethylcarboxamidoadenosine

NECI

5′-N-ethylcarboxamidoinosine

R/S-PIA

N⁶-[(R/S)-1-methyl-2-phenylethyl]adenosine

SAR

structure-activity relationships

XAC

8-[-4-[[[[(2-aminoethyl)amino]carbonyl]methyl]oxy]phenyl]-1,3-dipropylxanthine

Me₂SO

dimethylsulfoxide

CHO

Chinese hamster ovary