Identification of an Adenosine Receptor Domain Specifically Involved in Binding of 5′-Substituted Adenosine Agonists

Abstract

The bovine A₁ adenosine receptor (A₁AR) and rat A₃ adenosine receptor (A₃AR) display distinct agonist and antagonist binding properties. To identify regions involved in ligand recognition, A₁AR/A₃AR chimeric receptors were created, expressed in COS-7 cells, and analyzed by radioligand binding. A chimeric receptor in which the third intracellular loop of the A₁AR was replaced with that of the A₃AR bound agonists and the antagonist, [³H]xanthine amine congener, with affinities identical to wild-type A₁AR. A chimeric receptor with the fifth transmembrane domain (TM5) and third intracellular loop of the A₁AR replaced with that of the A₃AR displayed antagonist affinity similar to wild-type A₁AR. However, relative to the A₁AR, this chimeric demonstrated much greater affinity for 5′-substituted adenosine analogs, whereas affinity for N⁶-substituted compounds was unaffected. Substitution of a 6-amino acid cassette of the exofacial half of TM5 of the A₃AR into the A₁AR produced enhanced binding of exclusively a 5′-substituted analog, indicating involvement of this specific region in ligand recognition. These findings suggest that the 5′- and N⁶-substituents of adenosine agonists bind to distinct regions of ARs and that TM5 of the A₃AR interacts more favorably with 5′-substituted compounds than does that of the A₁AR.

Cell surface adenosine receptors (ARs)¹ mediate several of the varied physiological effects of adenosine. Subtype classification of ARs is based on structural distinctions, biochemical effects elicited by receptor activation, and unique pharmacological profiles (Olah and Stiles, 1992; Tucker and Linden, 1993). Molecular cloning has provided evidence for the existence of at least four distinct AR subtypes (A₁, A_2a, A_2b, A₃) in various species (Tucker and Linden, 1993). All cloned ARs display the structural motif, including the putative seven transmembrane helices typical of the G protein-coupled receptor family (Dohlman et al., 1991). At the amino acid level, the AR subtypes are ~50–70% identical in the membrane-spanning domains. Until recently, information regarding the interaction of these receptors and ligands at a molecular level has been limited to studies of native receptors and synthetic ligands which had been modified to enhance receptor affinity or selectivity (Jacobson, 1990). With the cloning of ARs, the structural features of the receptor proteins responsible for their ligand binding properties may be determined. Recently, it was shown that replacement of histidine residues located in the transmembrane domains of the bovine A₁AR had significant effects on agonist and antagonist binding by the receptor (Olah et al., 1992). Specifically, the histidine at position 278 in TM7 appeared critical for both agonist and antagonist binding by the receptor. Replacement of histidine 251 in TM6 resulted in a decrease in antagonist affinity while agonist affinity was unaltered. These data are supported by the results of previous experiments which employed chemical modification of rat brain membranes and indicated the involvement of histidines in ligand recognition by the A₁AR (Klotz et al., 1988). The 2 histidine residues examined in the mutational analysis are the sole histidines present in the membrane-spanning regions of the A₁AR and they are conserved in all A₁, A_2a, and A_2bARs thus far cloned. The A₃AR contains a histidine in TM7, although this amino acid is not present in TM6. It was shown recently that mutation of threonine at position 277 of the human A₁AR (TM7) to an alanine resulted in a marked decrease in the binding of NECA, an AR agonist containing a substitution at the 5′-position of the ribose group (Townsend-Nicholson and Schofield, 1994). This mutation had much less of an effect on the affinity of R-PIA and S-PIA, agonists containing functional groups at the N⁶-position of the adenine ring. Interestingly, this threonine residue in the human A₁AR is adjacent to a histidine in TM7 that is analogous to the residue studied in the bovine A₁AR (Olah et al., 1992). No other data obtained from mutational analysis of ARs are currently available regarding the ligand binding domain of this receptor family.

An approach to the further exploration of AR structural requirements for ligand binding is the exploitation of the distinctions which exist in the pharmacologic profiles of the various AR subtypes. Dramatic differences exist in the ligand specificity of the A₁AR and A₃AR despite the apparent similarity in the functional consequences of activation of either subtype, i.e. inhibition of adenylyl cyclase. These distinctions are observed for both agonist and antagonist ligands. With regards to agonist binding, the bovine A₁AR binds N⁶-substituted adenosine analogs with high affinity, e.g. [¹²⁶I]APNEA binds with a K_d (equilibrium dissociation constant) of ~0.5 nM and displays a 100–200-fold greater affinity for the N⁶-substituted compound, R-PIA, than for the 5′-substituted analog NECA (Olah et al, 1989; Olah et al, 1992). In contrast, the rat A₃AR binds [¹²⁵I]APNEA with a K_d of 15 nM and R-PIA and NECA with equal affinity (Zhou et al., 1992). The bovine A₁AR also displays ~1 nM affinity for substituted xanthine antagonist compounds such as [³H]XAC (Schwabe et al., 1985; Olah et al., 1992). In marked contrast, the rat A₃AR displays little affinity for any xanthine analog at concentrations up to 0.01 mM (Zhou et al., 1992). The sheep (Linden et al., 1993) and human (Salvatore et al., 1993) A₃ARs have been cloned recently and their pharmacologic properties extensively characterized. As compared with the distinct binding differences typical of the A₁AR, the sheep and human A₃ARs are also characterized by nearly equivalent affinities for R-PIA and NECA. The rat, sheep, and human A₃ARs are differentiated to a greater degree by their affinities for certain xanthine analogs.

In this report, the basis for agonist pharmacological differences displayed by the wild-type bovine A₁AR and rat A₃AR have been examined at a structural level by the creation of chimeric receptors consisting primarily of A₁AR sequence with substitution from distinct regions of the A₃AR. The binding properties of these chimeric receptors suggest that specific transmembrane domains are involved in the recognition of distinct regional substituents on adenosine analogs.

EXPERIMENTAL PROCEDURES

Mutagenesis and cDNA Expression

Bovine A₁AR (Olah et al., 1992) and rat A₃AR (Zhou et al., 1992) cDNAs were used as templates for the construction of mutant receptors. Chimeric receptors consisting primarily of A₁AR sequence with replacement of either solely the third intracellular loop or the fifth transmembrane domain and adjoining third intracellular loop with that of the A₃AR are referred to as IC3 and 5ic3, respectively. Chimeric receptors were created using a three-step PCR method with oligonucleotides consisting of both A₁AR and A₃AR sequence specifying the splicing junction. Thus, IC3 was created employing the oligonucleotides 5′-CTGCTCATGGTCCTCATCTATCTGGACATCTTC-3′ and 5′-AGGATGAGGGCCAGCGACTTGGCGGTCTT-3′. Oligonucleotides used for the construction of 5ic3 were 5′-GTCATCAGCATGGAGTACATGGTCTTCTTC-3′ and 5′-AGGATGAGGGCCAGCGACTTGGCGGTCTT-3′. A mutated A₁AR, referred to as A1FF, containing a substitution of 6 amino acids of TM5 with the analogous amino acids of the A₃AR was created using a PCR approach and a single oligonucleotide, 5′-AGCACCCAGGTGATGAAGCTGAAGAAGACCATGTACTCCAT-3′. The sequence of all mutated receptors was confirmed by double-stranded cDNA sequencing following the subcloning of the cDNA fragment into the expression vector pCMV5 (Dr. D. Russell, University of Texas Southwestern). The exact amino acid sequence of these receptors is shown in Fig. 1.

Fig. 1. Partial amino acid sequence of wild-type ARs and mutant receptors.

Sequence is in single letter code for the bovine A₁AR, and the sequence of the rat A₃AR and mutant constructs are aligned with the A₁AR. Putative transmembrane domains five (TM5) and six (TM6) are marked by a solid line. Hyphens indicate amino acids identical to those of the A₁AR, and the dot is a gap in sequence for optimal alignment. Construction of mutant receptors is described under “Experimental Procedures.”

For radioligand binding studies, the wild-type A₁AR, wild-type A₃AR, and mutant receptors were subcloned into the expression vector pCMV5. Transient transfection of COS-7 cells was performed via the DEAE-dextran method (Cullen, 1987) employing 15 μg of the wild-type A₁AR-pCMV5 and A₃AR-pCMV5 constructs or 25 μg of the mutant receptor-pCMV5 constructs. COS-7 cells were maintained in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum. For all studies, cells were harvested ~72 h post-transfection.

Radioligand Binding

Radioligand binding assays were performed using conditions and buffers which were found previously to give optimum results for both the A₁AR (Olah et al., 1992) and A₃AR (Zhou et al., 1992). Flasks of transfected COS-7 cells were washed twice with 10 ml of ice-cold 10 mM Tris, 5 mM EDTA, pH 7.4. Cells were then scraped into 5 ml of this buffer and disrupted by 20 strokes in a glass homogenizer. The homogenate was centrifuged at 43,000 × g for 10 min and the membrane pellet resuspended in 50 mM Tris, 10 mM MgCl₂, 1 mM EDTA, pH 8.26, at 5 °C. Adenosine deaminase was added to give a final concentration of 2 units/ml. Saturation binding assays were performed using 10 μM R-PIA to define nonspecific binding. Competition assays were performed using ~1.0 nM [³H]XAC or [^I25I]AB-MECA as the radioligand. All assays were conducted at 37 °C for 1 h and terminated by filtration using a Brandel cell harvester and rapid washing with 50 mM Tris, 10 mM MgCl₂, 1 mM EDTA, and 0.01% CHAPS over glass fiber filters pretreated with 0.3% polyethylenimine. Data were analyzed by a previously described computer modeling system (De Lean et al., 1982). IC₅₀ values obtained from computer analysis of competition curves were converted to K_i values using the Cheng-Prusoff equation (Cheng and Prusoff, 1973). For each construct, the appropriate K_d value obtained for [³H]XAC or [¹²⁵I]AB-MECA from saturation binding analysis was used in the conversion. Protein concentrations were determined by the method of Bradford (Bradford, 1976).

Materials

For the PCR construction of chimeric receptors, Vent DNA polymerase and restriction enzymes from New England Biolabs were employed. Double-stranded plasmid DNA sequencing was performed with the Sequenase 2.0 kit from U. S. Biochemical Corp. [³H]XAC was from DuPont NEN, and [¹²⁵I]APNEA was synthesized as described previously (Stiles et al, 1985). The synthesis and radioiodination of AB-MECA has been described (Olah et al., 1994). Unlabeled ligands were purchased from the following suppliers: R-PIA and NECA, Boehringer Mannheim; cyclopentyladenosine, Sigma; and N⁶-benzylad-enosine, Research Biochemicals International. The synthesis of cyclohexyl-NECA (Olsson et al., 1986) and benzyl-NECA (Gallo-Rodriguez et al., 1994) was as described. All cell culture supplies were from Life Technologies, Inc.

RESULTS AND DISCUSSION

The third intracellular loop of the A₁AR was replaced with the analogous structure from the A₃AR to create the chimeric receptor IC3. This was designed to determine if the coupling between this region of the AR subtypes and their associated G protein may influence agonist or antagonist binding. As previously described (Olah et al., 1992), wild-type A₁AR expressed in COS-7 cells bound both classes of radioligands with high affinity with K_d values for [¹²⁵I]APNEA and [³H]XAC being 0.63 ± 0.07 nM (n = 8) and 0.95 ± 0.11 nM (n = 8), respectively. The binding parameters for IC3 were nearly identical to those of the wild-type A₁AR with K_d values for [¹²⁵I]APNEA and [³H]XAC being 0.79 ± 0.20 nM (n = 6) and 1.07 ± 0.16 nM (n = 9), respectively. Furthermore, levels of receptor expression, as determined from [³H]XAC saturation isotherms, were similar for the wild-type A₁AR and IC3 (8.02 ± 2.27 and 8.53 ± 2.27 pmol/mg, respectively). These data indicate that the coupling of the respective receptors to their G proteins is not responsible for the lower affinity of A₃AR for agonists compared to the A₁AR.

Molecular modeling studies have suggested that TM5 of adenosine receptors is important in their recognition of ligands (IJzerman et al., 1992). We, therefore, constructed the chimeric 5ic3 receptor in which TM5 and third intracellular loop of the A₁AR was replaced with the analogous region of the A₃AR. The affinity of 5ic3 for [³H]XAC was similar to that observed for the previous constructs with a K_d of 1.9 ± 0.37 nM (n = 8). 5ic3 was expressed at significant levels (2.8 ± 0.72 pmol/mg), although the B_max (maximum binding capacity) for this chimeric receptor was ~35% of that obtained for the wild-type A₁AR. Although 5ic3 bound [¹²⁵I]APNEA with relatively high affinity (K_d = 3.79 ± 0.64 nM, n = 6), the affinity was ~5-fold lower than that displayed by the wild-type A₁AR.

The fact that [³H]XAC has similar affinity for all three receptors studied suggests that TM5 of the A₁AR and A₃AR is not the critical factor in determining antagonist affinity. [³H]XAC affinity differs by ~ 100,000-fold between the A₁AR and the rat A₃AR, yet switching of this domain does not significantly decrease the affinity of the A₁AR for [³H]XAC. Competition binding assays with CPX (see below) also provide evidence that regions of TM5 of the A₁AR are not essential in antagonist binding.

To further explore agonist recognition in the two chimeric constructs, a series of agonist compounds was employed in competition binding assays versus [³H]XAC (Table I). [³H]XAC was selected as the radioligand, since it bound with very similar affinity to the wild-type A₁AR and two chimeric receptors. In addition, an antagonist such as [³H]XAC binds to a single affinity state of the receptor. In Table I, K_i values for agonist binding to the wild-type A₃AR are given as a reference for the relative affinities of these compounds at this AR. Competition curves involving the A₃AR required the use of [¹²⁵I]APNEA as the radioligand, since antagonists do not bind to this receptor (Zhou et al., 1992). With regards to the data in Table I, it should be noted that the absolute K_i values of competitors will vary depending on the class of radioligand employed in the binding assay even with examination of the same receptor subtype. However, the rank potency order will remain characteristic of that receptor. For example, an approximate 100-fold variation in the affinity of R-PIA for the bovine A₁AR, depending upon the use of [¹²⁶I]APNEA or [³H]XAC as the radioligand, has been described previously (Olah et al., 1992). The initial set of compounds examined were adenosine analogs containing solely N⁶-substitutions. The selected compounds bound to the wild-type A₁AR with a wide range of affinities. R-PIA had ~150-fold greater affinity for the receptor than N⁶-benzyladenosine with cyclopentyladenosine being intermediate in potency (Table I). The chimeric receptors, IC3 and 5ic3, displayed affinities for all N⁶-substituted compounds similar to those observed for the wild-type A₁AR (Table I). These results are in agreement with those obtained previously from the [¹²⁵I]APNEA saturation binding studies. Extensive structure-activity experiments have documented that the affinity of several N⁶-substituted analogs for the A₃AR is typically 100–500-fold lower than that observed at the A₁AR (van Galen et al., 1994). The high affinity binding of N⁶-substituted analogs to 5ic3 suggests that the region of the A₁AR binding pocket which recognizes this class of agonists remains essentially intact even with TM5 replaced with that of the A₃AR.

Table I. Agonist competition for radioligand binding.

K_i values for wild-type (WT) A₁AR and the chimeric receptors, IC3 and 5ic3, expressed in COS-7 cells were obtained in competition assays using [³H]XAC as the radioligand. Data for wild-type A₃AR were obtained using a CHO cell line stably expressing the rat A3AR and [¹²⁵I]APNEA was used as the radioligand. The first three agonists contain substitutions exclusively at the N⁶-position. NECA has solely a 5′-substitution and benzyl-NECA and cyclohexyl(CH)-NECA are disubstituted (N⁶, 5′) compounds. All experiments were performed three to six times. K_i values represent the mean ± S.E.

Receptor	Ki
	R-PIA	CPA	N6-BA	NECA	Benzyl-NECA	CH-NECA
	nM
WT A1AR	24.2 ± 2.9	126.9 ± 3.4	3990 ± 200	5600 ± 1000	1780 ± 140	40.9 ± 6.5
WTA3AR	158.0 ± 52.0	240.0 ± 36.0	120.0 ± 20.0	113.0 ± 34.0	6.8 ± 2.5	16.0 ± 5.4
IC3	28.8 ± 5.0	121.7 ± 6.3	4150 ± 800	6300 ± 700	1880 ± 340	78.0 ± 27.2
5ic3	21.9 ± 3.2	124.2 ± 23.8	3040 ± 1100	99.2 ± 19.3	58.8 ± 16.8	0.45 ± 0.09

As noted above, 5′-substituted adenosine analogs display differential affinities for the A₁AR and A₃AR. Therefore, a series of such compounds was analyzed in competition binding assays versus [³H]XAC (Table I). The K_i values for NECA, benzyl-NECA and cyclohexyl-NECA were nearly identical for the wild-type A₁AR and IC3. These receptors displayed similar affinities for NECA and benzyl-NECA with cyclohexyl-NECA being ~100-fold more potent than the parent compound NECA. Unlike the situation observed for N⁶-substituted ligands, 5ic3 demonstrated much higher affinity for the 5′-substituted compounds. Based on K_i values, the 5ic3 chimeric displayed 56-, 30-, and 90-fold higher affinity for NECA, benzyl-NECA, and cyclohexyl-NECA, respectively, than did wild-type A₁AR. This enhancement of the affinity of the 5′-substituted compounds results in a pharmacologic profile of the 5ic3 chimeric closely resembling that of the A₃AR. The wild-type A₁AR demonstrates ~ 100–200 fold greater affinity for R-PIA as compared with NECA using [¹²⁵]APNEA (Olah et al, 1992) or [³H]XAC (Table I, Olah et al., 1989) as radioligand, and these two agonists are equipotent at the wild-type A3AR in displacing [¹²⁶I]APNEA binding (Zhou et al., 1992). Thus, the 5-fold difference in affinity between R-PIA and NECA displayed by 5ic3 (Table I) indicates that TM5 of the A₃AR is involved in the specific recognition of 5′-substituted adenosine analogs. The unique binding profile of the 5ic3 construct can be attributed to TM5, as solely an exchange of the third intracellular loop (IC3) had no effect on agonist binding.

To define which amino acids of TM5 may participate in the interaction with 5′-substituted adenosine analogs, a mutant A₁AR, A1FF, was created which contained a replacement of a 6-amino acid segment located in the exofacial half of TM5 with that of the A₃AR. Two of the amino acids in this cassette are conserved in the A₁AR and A₃AR (Fig. 1). The selection of this region was based on the probability that amino acids of the transmembrane-spanning helices directed toward the extracellular surface may be more likely to participate in receptor-ligand interactions (Baldwin, 1993). Membranes from COS-7 cells expressing the A1FF construct were prepared and employed in radioligand saturation assays. Since this chimera has properties of both the A₁-and A₃AR, we tested a series of radioligands to optimize receptor detection and nonspecific binding. The highest level of binding with the greatest component of specific binding was detected using [¹²⁵I]AB-MECA which is a N⁶-5′-disubstituted adenosine analog which has been recently characterized as a high affinity radioligand (K_d = ~1.0 nM) for the rat A₃AR (Olah et al., 1994). This radioligand was therefore selected to characterize A1FF. [¹²⁵I]AB-MECA bound to A1FF saturably and with high affinity (Fig. 2). For the binding of [¹²⁵I]AB-MECA to A1FF, K_d and B_max values of 0.58 ± 0.08 nM and 0.43 ± 0.09 pmol/mg, respectively, were obtained (n = 6). At the K_d, specific binding was typically 85% of total binding. The wild-type A₁AR and A₃AR expressed in COS-7 cells bound [¹²⁵I]AB-MECA with K_d values of 0.59 ± 0.18 nM (n = 6) and 1.22 ± 0.22 nM (n = 3), respectively (Fig. 2). [¹²⁵I]AB-MECA detected the wild-type A₁AR and A₃AR at levels of 1.20 ± 0.48 and 1.82 ± 0.70 pmol/mg, respectively. The chimeric receptor, 5ic3, also displayed high affinity binding of [¹²⁵I]AB-MECA with a K_d of 0.36 ± 0.07 nM (n = 4).

Fig. 2. Representative [^12SI]AB-MECA saturation binding in wild-type (WT) A₁AR and A₃AR and mutant receptors.

Membranes were prepared from COS-7 cells transfected with the appropriate cDNA-pCMV5 construct and radioligand binding assays performed as described under “Experimental Procedures.” Solid circles and solid squares represent total and nonspecific binding, respectively. Nonspecific binding was defined in the presence of 10 μM R-PIA. Plots are representative of experiments performed four (WT A₁AR, WT A₃AR, 5ic3) or six (A1FF) times.

The pharmacologic profile of A1FF was explored in competition for [¹²⁶I]AB-MECA binding using the prototypical N⁶-and 5′-substituted ligands, R-PIA and NECA, and compared with that of the wild-type A₁AR and A₃AR and chimeric 5ic3 (Table II). At the wild-type A₁AR, R-PIA was of markedly greater affinity (~ 70-fold) than NECA, whereas at the wild-type A₃AR, nearly identical K_i values were obtained for the two agonists. These findings are consistent with the potency order that is characteristic of these receptors. As observed previously in [³H]XAC competition assays (Table I), the chimeric 5ic3 bound NECA with greater affinity than did the wild-type A₁AR. Unlike the data obtained from [³H]XAC competition, 5ic3 displayed a decreased affinity for R-PIA compared with the wild-type A₁AR. This difference may result from the use of an agonist versus antagonist radioligand in the competition binding assays as [¹²⁵I]APNEA saturation binding also indicated a decline in affinity for N⁶-substituted agonists by 5ic3 as compared with wild-type A₁AR. In eight experiments, A1FF demonstrated an affinity for NECA very similar to that observed for 5ic3 which was ~4-fold greater than that obtained for the wild-type A₁AR. A1FF and wild-type A₁AR displayed similar affinities for R-PIA and CPX, indicating the enhancement of affinity produced by the amino acid substitution is specific for 5′-sub-stituted analogs. The unaltered CPX affinity supports the [³H]XAC binding data, indicating that TM5 of the A₁AR is not critically involved in antagonist recognition.

Table II. Competition for [¹²⁵I]AB-MECA binding.

K_i values for wild-type (WT) A₁AR and A₃AR and mutant receptors (5ic3 and A1FF) expressed in COS-7 cells were obtained using ~1.0 nM [¹²⁵I]AB-MECA as radioligand. Values represent the mean ± S.E. with the number of experiments indicated in the parentheses.

	Ki
	R-PIA	NECA	CPX
	nM
WT A1AR	0.22 ± 0.04 (7)	16.1 ± 6.1 (7)	4.36 ± 0.47 (3)
WT A3AR	225 ± 31.8 (6)	233 ± 35 (6)	NDa
5ic3	1.83 ± 0.14 (3)	3.31 ± 0.11 (3)	ND
A1FF	0.38 ± 0.07 (8)	4.01 ± 0.94 (8)	6.5 ± 3.0 (2)

^aND, not determined.

The agonist competition binding assays with 5′-substituted analogs provide new insights into agonist-receptor interactions of the A₁- and A₃AR. First, the 5′-group of an adenosine analog must interact with distinct and different regions of the binding pocket of the receptor than does the N⁶-group. The complete (5ic3) or partial (A1FF) substitution of TM5 of the A₁AR with that of the A₃AR produced a small decrease or had no effect on the affinity of N⁶-substituted agonists, yet dramatically increased the affinity of compounds containing a 5′-substitution. The fact that the two classes of adenosine agonists are differentially affected by both the 5ic3 and A1FF mutations indicates that the N⁶-substituent and 5′-substituent must interact with distinct regions of the receptor. Furthermore, an analysis of structure-activity relationships at the A₃AR has revealed that 5′-substitution of several adenosine receptor agonists results in these compounds having increased A₃AR affinity, and this enhancement is much more pronounced at the A₃AR as compared with the A₁ or A₂ARs (van Galen et al., 1994). Disubstitution at the N⁶- and 5′-positions results in compounds displaying the highest A₃AR affinity, an effect which is not typical of the structure-activity relationship at the A₁AR (van Galen et al., 1994).

Second, the present findings suggest that TM5 of the A₃AR may be the specific region of the receptor protein which interacts with the 5′-substituted group. The substitution of this region into the A₁AR results in a mutant A₁AR having much greater affinity for this class of compounds than does the wild-type A₁AR. Based on the enhancement of affinity, it appears that TM5 of the A₃AR interacts more favorably with this class of agonists than does that of the A₁AR. The data obtained with the A1FF mutant suggest that amino acids specifically in the exofacial segment of TM5 are involved in the recognition of 5′-substituted adenosine analogs. As evidenced by competition for [¹²⁵I]AB-MECA binding (Table II), replacement of amino acids in the A₁AR limited to this region (A1FF) results in an enhancement of NECA affinity to the same level observed with the replacement of the entire TM5 (5ic3). The precise role of this segment of TM5 in ligand recognition is not presently defined. A recent mutational study of human A₁AR ligand binding indicated that a threonine in TM7 was preferentially involved in NECA binding as compared with the recognition of N⁶-substituted agonists and the antagonist, [³H]CPX (Townsend-Nicholson and Schofield, 1994). Based on these findings and those presently described, the precise amino acid(s) of the A₁AR or A₃AR which may interact with the 5′-substituent cannot be defined. It is possible that both of the regions thus far examined (exofacial segment of TM5 and the defined region of TM7) form a binding site for the 5′-functional group. For several G protein-coupled receptors, including the 5-HT_1a (Guan et al., 1992), (β₂-adrenergic (Suryanarayana et al., 1991; Suryanarayana and Kobilka, 1993), and neurokinin 1 (Fong et al., 1993) receptors, mutational analysis has permitted the identification of a single amino acid which may interact with a specific functional group of a particular class of ligands. However, as the entire three-dimensional structure of a membrane spanning domain(s) may create the ligand binding pocket (Hibert et al., 1991), it is perhaps this overall conformation that is required for high affinity ligand binding.

Although the data from structure-activity experiments (van Galen et al., 1994) and mutational analysis suggest that 5′-substitution enhances ligand affinity at the A₃AR, perhaps due to an interaction with TM5, the presence of this functional group exclusively does not allow a ligand to preferentially bind to the A₃AR as opposed to the A₁AR. For example, the parent compound NECA has approximately 15-fold greater affinity for the A₁AR as compared with the A₃AR (Table II). Therefore, other moieties of the ligand molecule in addition to the 5′- or N⁶-group must interact with distinct regions of ARs. For example, one A₁AR ligand binding model suggests that hydroxyls of the ribose group of agonist ligands coordinate with the conserved histidine present in the seventh transmembrane domain of the receptor (IJzerman et al., 1992). The involved regions may differ among AR subtypes, and for all compounds examined this presently undefined interaction is apparently more favorable for the A₁AR. In effect, 5ic3 and A1FF display affinities for 5′-substituted analogs greater than either the wild-type A₁AR or A₃AR (Tables I and II), as these mutants contain the optimal components for ligand binding from both receptor subtypes.

In contrast to agonist binding, the substitution of the fifth transmembrane domain of the wild-type A₁AR with that of the wild-type A₃AR had minimal effect on antagonist binding. As noted above, the affinities of the A₁AR and rat A₃AR for antagonists varies ~100,000-fold. The present findings indicate that TM5 of the A₁AR does not interact with antagonists and that the amino acid differences existing in this specific region of the A₃AR are not responsible for the insensitivity of the latter subtype to alkylxanthine compounds. Thus, multiple regions of ARs are important in binding ligands, with TM5 being particularly important for 5′-substituted analogs.