A₃-adenosine receptors: design of selective ligands and therapeutic prospects

Introduction

Adenosine A₁ and A₂ receptors have been the subject of intense activity in the pharmaceutical industry. Adenosine agonists, which are almost exclusively derivatives of adenosine (1), have been sought as potential hypotensive (2), antipsychotic (3, 4), antiarrhythmic (5), antilipolytic (thus antidiabetic, 6) and cerebroprotective (7–10) agents. Adenosine antagonists, of which xanthines and a number of fused heterocyclic ring systems are representative (1), have been under development as antiasthmatic (11), antidepressant (12), antiarrhythmic (13), renal protective (14, 15), antiparkinson (16) and cognitive enhancing (17–19) drugs. In spite of the massive effort to develop selective ligands, a number of agents that initially looked promising did not survive clinical trials. One of the reasons for this failure has been the side effects, e.g., an adenosine agonist Cl-936 (N⁶-(2,2-diphenylethyl)adenosine), which showed efficacy in animal testing for antipsychotic-like activity (3), but caused arteriopathy in dogs, nausea, and other side effects (20).

Nevertheless, the interest in adenosine-based therapy has not waned. On the contrary, as our knowledge of the biological effects of this endogenous protective modulator (adenosine) advances, the envisioned therapeutic applications become more sophisticated and promising. Recently, the use of adenosine agonists in treating stroke has come into focus (9,10), since, for this acute application, the interference by some of the previously documented side effects would, in principle, be diminished. There may soon be clinical trials involving the use of A₁ agonists in treating stroke (21).

The discovery of a novel and distinct adenosine receptor subtype, the A₃ receptor, has opened new therapeutic vistas in the purine field. This receptor subtype has a unique pharmacological profile, distribution in the body, and effector coupling. Papers on selective A₃ agents are just beginning to appear (22), but clinical trials have not yet been attempted. It will take some time before the medicinal chemistry of A₃ receptors advances to the degree of selectivity already achieved for A₁ and A_2a receptors and before the physiological role of A₃ receptors is clarified.

Cloning of the A₃ receptor

Meyerhof et al. (23) cloned an orphan receptor from rat testes that resembled in sequence the known adenosine receptors. An identical clone from the rat brain was shown by the laboratories of Gary Stiles at Duke University and Olivier Civelli, then of University of Oregon, to function as an adenosine receptor (24) and was termed the A₃ receptor. This designation is not to be confused with an earlier tentative use of the same nomenclature to describe an unrelated phenomenon (25). This new receptor was unique in that adenosine had a very low affinity (initially estimated at 30 μM, see below) and its action was not antagonized by xanthines, such as theophylline, as are A₁ and A₂ receptors. Typical K_i values at A₃ receptors of roughly 10⁻⁴ M have been obtained (26, 27) for many xanthines that have nearly nanomolar potency at the other subtypes. A similar receptor, designated S17, was cloned from a sheep brain cDNA library by the laboratory of S. Reppert and identified by J. Linden and colleagues as an A₃-type receptor, i.e., it inhibited adenylyl cyclase, and this action was xanthine-insensitive (28). However, the sheep A₃ receptor was only 72% homologous in sequence with the rat A₃ receptor. This low degree of homology (species homologues are usually in the 90–95% range) raises the question whether these clones represent a single subtype. The cloning from a human brain cDNA library of the sequence homologous to the sheep A₃ receptor was also reported (29).

Distribution of the A₃ receptor

The cloning of the A₃ receptor from species other than rat, e.g., sheep (28) and human (29), has indicated that there are interspecies differences in its peripheral distribution. In the rat, the A₃ receptor has a very narrow distribution, being most highly expressed mainly in the testes, but also in lung, kidneys, heart, and brain. In the sheep, the A₃ receptor transcript is found in the lung, spleen, pars tuberalis, and pineal gland, with lower levels in the testes, kidneys, and brain (28). Curiously, the transcript was not detected in the sheep heart. The human A₃ receptor transcript (29) was most highly expressed in the lung and liver.

The distribution in the brain has been measured in absolute quantities through radioligand binding (2), and in relative levels in in situ hybridization experiments (24, 28, 30). There is a widespread, relatively low level of A₃ receptor binding sites throughout the mouse brain, with B_max values in the range of 120 fmol/mg protein (found in the striatum, cerebellum) to 220 fmol/mg protein (as seen in the hippocampus). In membranes from the forebrain, species differences in receptor density have also been found (31), with the B_max values of 29 (gerbil), 43 (rat), and 118 (rabbit) fmol/mg protein. In the rat brain (24), the transcript is weakly expressed in the cortex, striatum, and olfactory bulb. In the sheep brain (28), the transcript is modestly expressed in the cortex, striatum, hypothalamus, and cerebellum. Thus, preliminary indications are that irrespective of region, the densities are low, i.e., comparable to the levels of nicotinic receptors in the brain (32), and roughly 10–30 fold lower than levels of cortical A₁ adenosine receptors or striatal A_2a receptors (22).

Second messenger systems

The A₃ receptor is coupled to at least two second messenger systems, inhibition of adenylyl cyclase (24) and stimulation of phospholipase C (33). When expressed in Chinese hamster ovary (CHO) cells, A₃ receptors from rat, sheep, or human are coupled to inhibition of adenylyl cyclase (24, 28, 29). Recently, it was shown using novel selective agonists that the A₃ receptor in rat brain slices activates phospholipase C in a GTP-dependent manner (34). The latter finding provides a possible mechanistic explanation for the effects of A₃ activation in stroke (see below), since this second messenger system has been implicated in the development of neuronal damage in stroke.

Ligand development

Selective A₃ agonists

Scores of adenosine derivatives have been synthesized as adenosine agonists (1). In general, modification of the N⁶-position with hydrophobic moieties has provided selectivity for A₁ receptors, and substitution at the C2-position with amino-, oxo-ether, or alkynyl chains has resulted in A_2a selectivity. Replacement of the ring oxygen atom of the ribose moiety by a carbon or sulfur atom also provided a degree of A_2a selectivity (35).

We have found that many of the selective A₁ and A_2a agonists also have considerable affinity in binding at A₃ receptors (26). For example, the A₁-selective CPA (12, N⁶-cyclo-pentyladenosine) is 2-fold more potent at A₃ vs. A_2a receptors (Table I). The A_2a-selective CGS21680 (13,2-[4-[(2 - carboxyethyl)phenyl]ethylamino] - 5′ - N - ethylcarboxy-amidoadenosine) is 4.5-fold more potent at A₃ vs. A₁ receptors.

Table I.

Affinities of adenosine derivatives at rat brain A₁, A_2a, and A₃ receptors, arranged in order of decreasing affinity at rat A₃ receptors.^a

Compound	R1	R2	R3	Ki (μM)
				Ki(A1)	Ki(A2a)	Ki(A3)
1b	CH3NHCO	Cl	3-I-Bz	0.82	0.47	0.00033
2c	CH3NHCO	H	3-I-Bz	0.054	0.056	0.0011
3d	CH3NHCO	H	3-I-4-NH2-Bz	0.018	0.197	0.0013
4	CH3NHCO	CH3S	3-I-Bz	2.140	3.210	0.0023
5	CH3NHCO	CH3NH	3-I-Bz	4.89	4.12	0.00312
6e	EtNHCO	H	Bz	0.087	0.095	0.0068l
7	HOCH2	H	3-I-Bz	0.02	0.0175	0.0095
8f	CH3NHCO	H	H	0.0836	0.0668	0.072l
9g	EtNHCO	H	H	0.0063	0.0103	0.113l
10h	HOCH2	H	4-NH2-(CH2)2	0.014	0.172	0.116
11i	HOCH2	Cl	cyclopentyl	0.0006	0.95	0.237l
12j	HOCH2	H	cyclopentyl	0.00059	0.462	0.24l
13k	EtNHCO	NH(CH2)2-φ-p-(CH2)2-COOH	H	2.6	0.015	0.584l
14	HOCH2	Cl	H	0.0093	0.063	1.89l

^aK_i ± SEM determined in radioligand binding assays expressed in μM (n = 3–6), using the following radioligands: A₁, [³H]PIA in rat cortical membranes; A_2a, [³H]CGS 21680 binding in rat striatal membranes; A₃, [¹²⁵I]AB-MECA binding, unless noted, in membranes of CHO cells stably transfected with the rat A₃-cDNA. A percent value indicates the percent displacement of radioligand at the concentration (M) given in parentheses.

^bCl-IB-MECA;

^cIB-MECA;

^dI-AB-MECA;

^eBz-NECA;

^fMECA;

^gNECA;

^hAPNEA;

ⁱCCPA;

^jCPA;

^kCGS 21680.

^lA₃ affinity determined versus [¹²⁵I]APNEA binding.

Is there now a general approach for designing adenosine derivatives with A₃ selectivity? We have reported that one principle of achieving A₃ selectivity in adenosine agonists is the combination of the optimal N⁶- and 5′-substitutions (26). Specifically, among alkyl, cycloalkyl, and arylalkyl N⁶-substituents, a benzyl group is favored, due to its diminished potency at A₁ and A_2a receptors. The A₃ selectivity-enhancing effects of N⁶-benzyl modification are additive with the A₃ affinity-enhancing effects of the 5′-uronamido function, as in NECA (9, adenosine-5′-N-ethyluronamide). The first such hybrid molecule to show A₃ selectivity (26) was N⁶-benzyl-NECA (6, Table I). In a comparison of various 5′-uronamido groups in mono-substituted adenosine derivatives, the 5′-N-methylamide (8, MECA) had particularly favorable A₃ receptor vs. A₁/A_2a affinity (36).

A study of substituent effects on the N⁶-benzyl group has shown that substitution at the 3-position with sterically bulky groups, such as the iodo group, is optimal (36), leading to the development of the highly potent A₃ agonist N⁶-(3-iodo-benzyl)-adenosine-5′-N-methyluronamide (2, IB-MECA, Table I) which is 50-fold selective for A₃ vs. either A₁ or A₂ receptors. A closely related, but less selective ligand containing radioactive iodine, [¹²⁵I]AB-MECA (3, N⁶-(4-amino-3-iodobenzyl)-adenosine-5′-N-methyluronamide), was developed (37) for characterization of A₃ receptors and was found to have a K_d value of 3.6 nM in binding to rat A₃ receptors in the RBL-2H3 mast cell line. This radioligand has supplanted the use of the lower affinity [¹²⁵I]APNEA (radioiodinated N⁶-[2-(p-aminophenyl)ethyl]adenosine) (24, 26) in our efforts to elucidate structure-activity relationships (SAR). It is to be noted that APNEA (10), the noniodinated precursor which has been used in in vivo studies of A₃ receptors (see below), is actually 8-fold selective for A₁ receptors (38) (Table I). K_i values at the cloned rat A₃ receptors for the same compounds using as radioligand either [¹²⁵I]AB-MECA (3) or [¹²⁵I]APNEA, are quite comparable. [¹²⁵I]AB-MECA (3) is a suitable radioligand when used with transfected cell lines that only express the A₃ subtype. In studies of the brain, since this radioligand it is not highly selective for A₃ receptors it must be used in conjunction with an antagonist of A₁ and A₂ receptors. For this purpose, we have employed XAC, 20, the xanthine amine congener (31), which in the rat effectively blocks only A₁/A_2a/A_2b receptors at a concentration of 1 μM.

The influence of 2-substitution of adenosine on A₃ affinity was also studied (31). Initially it was found that 2-substituted analogs such as 2-chloroadenosine (14, CADO) and 2-chloro-N⁶-cyclopentyladenosine (11, CCPA) were relatively well tolerated at A₃ receptors (26). For example, CPA (12) and CCPA (11) are equipotent at rat A₃ receptors (Table I). The C2-modification was also compatible or additive with A₃ potency-enhancing modifications at other sites on the adenosine molecule. In fact, upon combination with the previously elucidated N⁶- and 5′-position modifications of IB-MECA (2), the 2-chloro group produced an even greater margin of A₃ selectivity. Thus, 2-chloro-IB-MECA (1, Cl-IB-MECA) displayed an A₃ selectivity in binding assays of 2500-fold vs. A₁ and 1600-fold vs. A_2a receptors. In the same study, the 2-methylthioether (4) and 2-methyl-amino (5) derivatives of IB-MECA were also shown to have both high selectivity and affinity for A₃ receptors. Thus, 2-substitution has indications of being a generally favored modification of A₃ selective agonists.

The affinity of adenosine at adenosine receptors is not readily measured directly in radioligand binding assays due to the need to add adenosine deaminase to destroy endogenous adenosine that lingers in most tissue preparations, regardless of extent of washing. In the study of Zhou et al. (24), the inhibition constant for adenosine was estimated as ~30 μM. However, we have estimated the affinity to be greater than originally proposed. Using a comparative method based on differences in affinity with various substitutions, we have estimated the K_i value to be close to 1 μM (Fig. 1). This has a highly significant bearing on the physiological role of A₃ receptors, since, under severe stress conditions, it is possible to exceed an endogenous concentration of adenosine of 1 μM, and thus (presumably) activate a large fraction of the A₃ receptors. A similar calculation for the same set of analogs binding to A₁ and A_2a receptors has estimated the K_i values of adenosine to be 10 and 30 nM, respectively.

Fig. 1.

Estimation of the affinity of adenosine at rat A₃ receptors using an approach based on comparing K_i values for various adenosine analogs in radioligand binding assays. Estimation of affinity of adenosine at rat A₃ receptors was done by extrapolation from ratios of measured affinities of mono- and di-substituted analogs. The estimated value or range for the K_i of adenosine in nM is shown in a rectangular box for each group of compounds. K_i values (26,36, 38) in nM are shown for CHA (N⁶-cyclohexyl), CPA (N⁶-cyclopentyl), BzAdo (N⁶-benzyl), DiMeAdo (N⁶-dimethyl), IB-Ado (N⁶-3-iodobenzyl), and NECA (5′-N-ethyluronamide) with or without the modification shown in italics. For example, the transformation of IB-Ado to the corresponding N-methyluronamide, i.e., IB-MECA (second row of arrows, third entry, see Table I for structure), changes the K_i value from 9.5 to 1.1 nM. For the representative compounds selected, 5′-N-Et and 5′-N-Me uronamide modifications resulted in enhancement of affinity by 10- to 18-fold or 8-fold, respectively. 5′-N-Cyclopropyl uronamide and 2-chloro modifications did not markedly change affinities, and the N ¹-oxide modification diminished affinity by 4-fold. Thus, the estimated affinity of adenosine is ~ 1 μM.

The agonist structure-activity relationships, with emphasis on the N⁶-benzyl ring substituents, were studied in a quantitative model using the Comparative Molecular Field Analysis (CoMFA) program in Sybyl (39). CoMFA studies suggested that the NH-CH₂ group at the N⁶-position of IB-MECA could be replaced by O-NH or NH-NH without significant loss of A₃ affinity. Also in the same study, it was found that bulky chains located at the 3-position of the benzyl ring, which can adjust conformationally within the binding site of the receptor, were well tolerated in binding at A₃ receptors. The N⁶-iodobenzyl group is so favorable towards binding at A₃ receptors that the mono-substituted N⁶-iodobenzyladenosine, 7, is slightly selective for A₃ receptors (38). This is the only example reported so far of a mono-substituted adenosine analog with A₃ selectivity.

In order to define the structural requirements for molecular recognition by A₃ receptors, we have proposed a molecular model for ligand binding at this subtype (26) that is consistent with known structure-activity relationships. This model features anchoring of the ribose moiety of adenosine to a histidine residue in the seventh transmembrane helix, that is conserved among all adenosine receptor subtypes. Another histidine residue in the sixth transmembrane helix, also proposed to be involved in ligand recognition and common to A₁ and A₂ adenosine receptors, is absent in A₃ receptors. Thus, we verified our surmise that the affinity of xanthines at A₃ receptors could be enhanced by the presence of a ribose moiety.

In an effort to synthesize A₃ antagonists, we attempted to maximize the affinity of xanthine derivatives at the binding site. Molecular modeling (see below) followed by chemical synthesis suggested that one means of accomplishing this was to anchor the xanthines by adding a ribose group at the 7-position. Some members of this class of compounds, the 1,3-dialkyl-xanthine-7-ribosides, have been synthesized and were previously found by Uzerman and colleagues to bind to A₁ receptors (40). At rat brain A₃ receptors, 1,3-dibutylxanthine-7-riboside (17, DBXR) was found to bind with a K_i value of 6.03 μM (26), whereas the parent xanthine, 1,3-dibutylxanthine (25, Table II), displayed a K_i value of 143 μM. Thus, the presence of the ribose moiety enhances affinity of xanthines at rat A₃ receptors, presumably by interacting with the adenosine recognition region of the A₃ receptor, which is assumed to represent only a partial subset of the analogous region in A₁ receptors. At A₁ receptors there is a decrease in affinity in going from xanthines to xanthine-7-ribosides, since the xanthines themselves are of relatively high affinity, through recognition by a portion of the receptor that is missing from the A₃ receptor. As a hybrid between adenosine agonist and antagonist structures, DBXR (17) proved to be a partial agonist (26) in the A₃-mediated inhibition of adenylyl cyclase, as was reported previously for xanthine-7-ribosides acting at A₁ receptors.

Table II.

Affinities of xanthine derivatives at rat brain A₁, A_2a, and A₃ receptors, arranged in order of decreasing affinity at rat A₃ receptors.^a

Compound	R1, R3	R8	R7	Ki (μM) or % inhibition (conc.)
				Ki (A1)	Ki (A2a)	Ki (A3)
15b	CH3(CH2)3	H	Rib, R5=MeNHCO	37.300	19% (10−4)	0.229
16c	1-CH3(CH2)2 3-CH2(3-I-4-NH2)	φ-p-OCH2COOH	H	0.037	0.7	1.17
17d	CH3(CH2)3	H	Rib, R5-HOCH2	4.190	19.500	6.03
18	CH3 X = S	(CH2)3COOH	H	5.70	67.8	9.36
19e	CH3(CH2)2	φ-p-CH=CHCOOH	H	0.015	0.80	15.0
20f	CH3(CH2)2	φ-p-OCH2CONH-(CH2)2NH2	H	0.0112	0.063	29.0
21g	CH3(CH2)2	φ-p-SO3H	H	0.14	0.79	90.1
22	CH3	(CH2)3COOH	H	0% (10−4)	25% (10−4)	93.4
23	CH3(CH2)2	φ-p-OCH2CONH-(CH2)2NHCOCH2N-(CH2COOH)COCH2CH2-COOH	H	2.86	9.22	99.9
24	CH3	CH=CHCOOH	CH3	3% (10−4)	42	130
25	CH3(CH2)3	H	H	0.50	29.3	143j
26	CH3	H	CH3	29	48	30.1% (10−4)j
27	CH3	H	H	8.5	25	23.1% (10−4)j
28h	CH3(CH2)2	cyclopentyl	H	0.46	340	18.7% (10−5)j
29i	CH3	CH=CH-m-φCl	CH3	28.2	0.054	4.2% (10−5)j

^aK_i ± SEM determined in radioligand binding assays expressed in (μM (n = 3–6), using the following radioligands: A₁, [³H]PIA in rat cortical membranes; A_2a, [³H]CGS 21680 binding in rat striatal membranes; A₃, [¹²⁵I]AB-MECA binding, unless noted, in membranes of CHO cells stably transfected with the rat A₃-cDNA. A percent value indicates the percent displacement of radioligand at the concentration (M) given in parentheses.

^bDBXRM;

^cBWA522;

^dDBXR;

^eBWA1433;

^fXAC;

^gSPX;

^hCPX;

ⁱCSC;

^jA₃ affinity determined versus [¹²⁵I]APNEA binding.

Since 1,3-dibutylxanthine-7-riboside was nonselective, we explored the structure-activity relationships in this series of unnatural nucleosides in an effort to identify A₃ selective agents (41). Adding the same 5′-uronamide group (N-methylamide) that favored A₃ selectivity in adenosine derivatives had a similar effect in the xanthine-7-riboside series. 1,3 - Dibutylxanthine - 7 - riboside - 5′ - N- methylcarboxamide (15, DBXRM), with a K_i value of 229 nM at A₃ receptors, was 160-fold selective for rat A₃ vs. A₁ receptors and >400-fold selective vs. A_2a receptors (41). Although the intention was to identify A₃ antagonists, this derivative acted as a full agonist in the A₃ receptor-mediated inhibition of adenylyl cyclase, providing the first example of any non-adenosine derivative acting as a selective agonist at any subtype of adenosine receptors.

The parallel in SAR between adenosine derivatives and xanthine-7-ribosides is supportive of our A₃ receptor model which features the ribose moiety of the ligand, either adenosine or xanthine ribosides, coordinated by hydrogen bonding to the same amino acid residue, hypothetical, the histidine of the seventh transmembrane helix (26). There are also common features between the structure-activity relationships (27) for dialkylxanthines binding to A₃ receptors (see below) and xanthine-7-ribosides. The 1,3-dibutyl analogs (17 and 25, Table II) in both cases contain the optimal chain length (for neutral molecules). A major difference is that for the xanthines, selectivity at the rat A₃ receptor was not achieved (27). At A₁ receptors, the xanthines are generally more potent than the corresponding xanthine-7-ribosides (26), while at rat A₃ receptors the converse is true.

Site-directed mutagenesis (42) has identified features of the A₁ receptor which when incorporated into A₃ receptor chimers provide high affinity binding of xanthines. Surprisingly, a region of the second extracellular loop (the C-terminal half) of the A₁ receptor had this property. Replacement of both this segment and the sixth and seventh transmembrane helices of the rat A₃ receptor with the bovine A₁ sequences resulted in a 50,000-fold increase in the affinity of CPX (28).

Species differences and A₃ antagonists

In the study of Linden et al. (28) it was noted that there were substantial species differences in the affinity of xanthines. As opposed to the rat A₃ receptor, which was described as xanthine-insensitive, the sheep and human receptors (28, 29) bound certain xanthines, especially those 8-aryl xanthines bearing a negative charge on the 8-position substituents such as BWA522 (16) and BW1433 (19), with considerably higher affinities (Fig. 2). Thus, the amine derivative XAC bound with K_i values of 181 and 71 nM at sheep and human A₃ receptors, respectively, and the carboxylic acid-containing xanthine BWA522 (16,3-(4-aminobenzyl) - 8 - [4 - [[[carboxy]methyl]oxy]phenyl] -1 - propyl-xanthine), bound with K_i values of 3 and 18 nM. At rat A₃ receptors, BWA522 was later found to bind with a K_i value of 1.17 mM (31). In general, the affinities of 8-arylxanthines at rat, rabbit, and gerbil brain A₃ receptors were considerably less (typically by nearly two orders of magnitude) than the previously reported affinities at cloned sheep and human A₃ receptors. This enhancement of affinity was present for a cationic (XAC, 20) as well as anionic xanthines. 8-Cyclopentyl-1,3-dipropylxanthine (28, CPX) distinguished clearly the specificity of human (relatively high affinity, K_i 0.76 μM) vs. sheep (very low affinity, K_i 49 μM) A₃ receptors, with the affinity in the rat, rabbit and gerbil being intermediate.

Fig. 2.

Species differences in the affinity of xanthine derivatives (for structures refer to Table II). The K_i values were determined in radioligand binding assays (vs. ¹²⁵I-labeled adenosine derivatives) in membranes of CHO expressing cloned A₃ receptors (28, 29, 31).

The development of selective antagonists as pharmacological probes and radioligands for A₃ receptors remains a challenge. The cautious use of xanthines to define the physiological actions of A₃ receptor activation is essential. It is still undetermined whether these species differences justify proposing distinct (A_3a and A_3b) receptor subtypes (43).

The SAR of xanthines at rat A₃ receptors has been explored as well (27). The presence of a sulfonate (21), carboxylate (16 and 18) or multiple carboxylate (23) groups did not result in a significant enhancement of affinity at rat A₃ receptors (Table II), although as previously observed, an anionic group tended to diminish potency at A₁ and A_2a receptors. The rat A₃ receptor affinity was not highly dependent on the distance of a carboxylate group from the xanthine pharmacophore. 2-Thio (18) vs. 2-oxo (22) substitution favored A₃ potency, and 8-alkyl (22) vs. 8-aryl (16, 20, and 21) substitution favored A₃ selectivity, although few derivatives were truly selective for rat A₃ receptors. 1,3-Dimethyl-8-(3-carboxypropyl)-2-thioxanthine, 18, was 7-fold selective for A₃ vs. A_2a receptors. 1,3,7-Trime-thyl-8-(trans-2-carboxyvinyl)xanthine, 24, was somewhat selective for A₃ vs. A₁ receptors. For 8-aryl xanthines, affinity at A₃ receptors was enhanced by 1,3-dialkyl substituents in the order dibutyl > dipropyl > diallyl. Curiously, xanthines having K_i values in the 1–20 μM range at rat A₃ receptors failed to antagonize the A₃ agonist-induced inhibition of adenylyl cyclase in transfected CHO cell membranes (27). At sheep A₃ receptors, however, at least one xanthine (100 μM BW1433, 19) did antagonize the effects of NECA on adenylyl cyclase (22).

Interspecies differences in agonist affinity at A₃ receptors were less pronounced than those for xanthines (31). Among adenosine agonists of varied structure (5′-, 2-, and N⁶-derivatives), the relative binding affinities at rat A₃ receptors are similar to those at human, but not sheep, A₃ receptors; however, in all species examined, the stereoselectivity for R- vs. S-PIA (N⁶-(2-phenylisopropyl)adenosine), a well-characterized pattern at A₁ receptors, was preserved. The selectivity for the R-diastereomer was 6-fold at rat (26), 11-fold at sheep (28), and 10-fold at human A₃ receptors (29).

In vivo actions of A₃ receptor activation and therapeutic prospects

The first in vivo study of a selective A₃ agonist (22) indicated that it was a very potent locomotor depressant. The IC₅₀ value for IB-MECA in open field behavior in mice was 16 μg/kg, i.p., but it was not as fully efficacious as either selective A₁ or A_2a agonists. Specifically, the A₁ and A_2a selective agonists caused nearly complete immobility (although not sedation), whereas the A₃ agonist could cause only a maximal 60% reduction in locomotor activity. The depression elicited by the A₃ agonist was not reversed, statistically significantly, by either an A₁ antagonist, CPX (28), or by a selective A_2a antagonist, 8-chlorostyrylcaffeine (29, CSC). In view of the subsequent studies of seizures and ischemia, the results were suggestive of a potent, central effect of A₃ activation.

Administration of IB-MECA in mice also caused rapid scratching behavior, of which the frequency of occurrence appeared to increase with the dose. Since activation of A₃ receptors facilitated release of histamine in a rat mast cell line (33), it was proposed that the scratching could be related to histamine. Coadministration of an H₁-histamine antagonist, cyproheptidine, eliminated this behavior. Thus, it appears that IB-MECA may release histamine in vivo. Compound 48/80, a mast cell releasing agent, administered in vivo mimics some of the effects of acutely administered A₃ agonists (44, 45).

Cardiovascular effects of A₃ receptor stimulation have been demonstrated indirectly (46), through coadministration of a nonselective agonist, APNEA (10, Table I), and an antagonist that blocks action only at A₁ and A_2a receptors. It appears that the hypotensive action of APNEA is related to a lowering of cardiac output, perhaps mediated by mast cell degranulation (45), and is not the result of vasodilatation, a well characterized effect of A_2a activation.

At our lab, we have studied the effects of IB-MECA (0.1 mg/kg, i.p.) on cardiovascular parameters (blood pressure, cerebral blood flow using a laser Doppler probe) in rats (Fig. 3) and gerbils (47). This dose corresponded to a relatively high dose in the locomotor study in mice (22), namely one that elicited a maximal depression in locomotor activity. The A₃ agonist administered alone caused a lowering of the blood pressure with little effect on heart rate. The pronounced hypotensive effect began within 2 min postinjection and was maintained throughout the entire 90 min monitoring period. The effects of this agonist on blood flow and respiratory rates were not significant. In contrast, a potent and selective A₁ agonist, N⁶-cyclopentyladenosine, 12, at a dose of 0.1 mg/kg, i.p., caused an intense drop in both blood pressure (by 50%) and heart rate (by 20%) from their initial values. The A₁/A₂ antagonist BWA1433 (1,3-dipropyl-8-[4-(carboxyethynyl)phenyl]xanthine, 4 mg/kg i.p., 19, Table II) did not reverse the hypotensive effects of IB-MECA. Blood flow and respiratory rate were increased by administration of the adenosine antagonist alone. IB-MECA alone or the coadministration of BWA1433 and IB-MECA had nearly no effect on body temperature.

Fig. 3.

Effects on systolic blood pressure in rats. Rats (Sprague Dawley, 300–340 g) were lightly anesthetized using 1.5–2% halothane. They were monitored for heart rate with a tail sensor and blood pressure using the tail cuff method. Measurements were made at regular intervals: every 2 min for the first 10-min period, and every 5 min thereafter, until 30 min, at which time a 30-min interval was used, until 90 min. Each drug group had 5–7 animals. Drugs were dissolved in a vehicle consisting of a 80:20 mixture of Alkamuls EL-620 and saline, pH 7.4. For a 0.1 mg/kg dose, a solution of 2.5 mg/ml in vehicle was prepared. When administered alone, IB-MECA (0.1 mg/kg) was given at the beginning of the monitoring period. When coadministered with the antagonist BWA1433 (4 mg/kg), the antagonist was given at the beginning of the monitoring period, and IB-MECA (0.1 mg/kg) was given after 10 min.

A₃ agonists for cerebral ischemia

We have studied the effects of chronic vs. acute administration of selective adenosine agents, both agonists and antagonists, at the various subtypes. A₁ agonists given acutely are cerebroprotective in models of stroke and seizures. For A₁ receptors, there is a paradoxical reversal of the effects depending on either chronic or acute dosing regimen, as determined using spatial memory and seizure models. Thus, a chronically administered A₁ antagonist (CPX) was antiischemic and anticonvulsant, and a chronically administered A₁ agonist (CPA) was pro-cognitive (9).

A similar paradoxical reversal in cerebroprotection in gerbils was also seen for an A₃ agonist (Fig. 4). Chronically administered IB-MECA (100 μg/kg, i.p.) in gerbils dramatically improved the histopathological and neurological outcome after both 10 and 20 min ischemia induced by bilateral occlusion, and the survival rate was 90% compared to 60% in the controls. Moreover, the treatment preserved short-term memory following 10 min cerebral ischemia (47, 48). The acute administration of the same dose followed by ischemia resulted, on the other hand, in a more extensive deterioration of hippocampal cells, behavioral indicators, and decreased survival in treated vs. control animals.

Fig. 4.

Effects of acute or chronic IB-MECA in gerbils. A) Survival in gerbils following 10 min bilateral carotid artery occlusion i.e., global ischemia (47). IB-MECA (0.1 mg/kg) was injected i.p. either 15 min before ischemia (acute) or daily for 10 days prior to ischemia, with a 24-h gap between the last injected dose and the ischemia (chronic). B) Seizure mortality as a result of NMDA alone or NMDA following IB-MECA treatment (44). NMDA was administered at a dose of 60 mg/kg, i.p. IB-MECA at the dose indicated was injected i.p. either 15 min before NMDA (acute) or daily for 6 weeks prior to NMDA, with a 24-h gap between the last injected dose and the NMDA (chronic). Mortality was determined in the first 5 h following seizures. There was no change in mortality between 5 and 24 h.

Chronically administered IB-MECA was protective in chemically induced (NMDA or pentamethylenetetrazole) seizures (44). Significant improvement in seizure latency, neurological impairment, and survival was observed. In electrically induced seizures, chronic but not acute IB-MECA reduced postepileptic mortality.

It is unknown whether the protective effects of chronically administered IB-MECA or the opposite effects of acute IB-MECA are related to its effect on blood flow, neuronal mechanism, or both. Following chronically administered IB-MECA and stroke, postischemic cerebral reperfusion was significantly improved (47). Postischemic cerebral blood flow was greatly reduced following acutely administered IB-MECA (Fig. 5). Compound 48/80, which releases histamine and constricts arterioles, apparently diminishes the neurotoxic effects of peripherally administered NMDA or pentylenetetrazole, as does acutely administered IB-MECA (44). This suggests that the protective effects of acutely, but not chronically administered IB-MECA against chemically induced seizures may be related to vascular changes reducing the availability of the toxin in the brain.

Fig. 5.

Cerebral blood flow rate (CBFR) in gerbils (n = 5/group) following either 10 min or 20 min bilateral carotid artery occlusion (47). Control gerbils are indicated by open circles. IB-MECA (0.1 mg/kg) was injected i.p. either 15 min before ischemia (acute, filled triangles) or daily for 10 days prior to ischemia, with a 24-h gap between the last injected dose and the ischemia (chronic, filled circles).

A₃ agonists for cardiac preconditioning

Downey and coworkers have shown that the cardioprotective effect of a brief exposure to an adenosine agonist in rabbits prior to an ischemic infarct (preconditioning) is not reversed by some xanthines, e.g., 200 nM CPX, that are known to block A₁ receptors, while another xanthine, SPT (21, 8-p-sulfophenyltheophylline), causes antagonism only at a high concentration of 100 μM (49). This has been interpreted as consistent with an A₃ component of preconditioning. Similar results were obtained in two other studies (50, 56). Thus, it is conceivable that exposure of the heart to a selective A₃ agonist might be highly protective.

The cell type on which A₃ receptors are located in the heart is not well established. The activation of A₃ receptors on mast cells has been proposed by Fozard and coworkers to lead to histamine release and subsequently hypertension (45). The presence on cardiomyocytes has not yet been demonstrated.

A₃ antagonists as antiinflammatory or antiasthmatic agents

Beaven et al. (52) have suggested that A₃ antagonists may have potential as antiinflammatory agents acting via mast cells. In human lung, A₃ receptors appear to be expressed mainly on eosinophils and possibly upregulated in pulmonary disease, suggesting a relevance to asthma (53). Adenosine was previously shown to be bronchoconstrictor in the asthmatic lung (54), although now, perhaps it appears to be acting through A₃ receptors. A selective A₃ antagonist might be therapeutically useful in treating asthma and other inflammatory disorders.

Other potential applications: reproduction, cancer, etc

Since A₃ receptors are expressed in testes (23) and appear to be involved in spermatogenesis, perhaps A₃ selective agents could be used in altering male fertility.

A number of other phenomena in which adenosine agonists produce a biological effect in a manner that is not antagonized by xanthines have been explained by invoking A₃ receptors. In these cases, the lack of a reliable antagonist, selective or otherwise, is a severe disadvantage. Adenosine was found to cause an increase in the serotonin uptake in RBL 2H3 cells (55), and this action was suggested to be of the A₃ type, although it was antagonized by XAC. Thus, there may be a connection between A₃ receptors and antidepressant drug therapy. There may be a connection with cancer, as well (56). It was proposed that the action of adenosine to inhibit the adhesion of killer lymphocytes to adenocarcinoma cells is through an A₃ receptor (56, 57). The implications of this finding for drug development are yet to be explored.

Summary

The A₃ receptor has been established as a distinct receptor subtype through cloning and the synthesis of selective agonists. The distribution of this subtype in the body is unique and more limited than for A₁ or A₂ receptors. Actions associated with this subtype include mast cell degranulation, cerebroprotection, cardioprotection, and possibly alteration of fertility. The effects on the inflammatory system and the high level of A₃ receptor expression in the lungs suggests the use of A₃ antagonists for asthma. Thus, there is a tremendous potential for development for therapeutic purposes of selective drugs, either agonists or antagonists, acting at this receptor.