Introduction

Top Abstract Introduction Experimental Procedures Results Discussion References

A_2AARs are one of four subtypes (A₁, A_2A, A_2B, and A₃) of GPCRs that respond to the purine adenosine, which is released from tissues in response to metabolic stress or ischemia. The A_2AAR is an important pharmacological target because of the generally anti-inflammatory effects elicited when it is activated (Sullivan and Linden, 1998; Linden, 2001). Like other GPCRs, the A_2AAR population is composed of receptors in two conformational states: those coupled to a heterotrimeric G protein, forming an R-G complex, and those that are uncoupled. GPCRs can be converted to uncoupled receptors upon binding of guanine nucleotides such as GTPS to the G protein. Coupled GPCRs have a higher affinity for agonist molecules than do their uncoupled counterparts.

We have shown previously that the radiolabeled agonist [¹²⁵I]APE binds to two affinity states of rat striatal A_2A adenosine receptors (K_D = 1.3 and 19 nM) and < 20% of striatal receptors are found in the high-affinity conformation (Luthin et al., 1995). [³H]CGS21680 also binds to two affinity states of rat striatal membranes (K_D = 3.9 and 51 nM) (Luthin et al., 1995) and to two affinity states in human brain preparations (Wennmalm, 1988). The high-affinity state of the recombinant human A_2AAR has not been easily observable because recombinant A_2AARs do not seem to form R-G complexes to a significant degree. Poor A_2A coupling was noted in COS-7 cells assayed with [¹²⁵I]APE (Luthin et al., 1995) and in Chinese hamster ovary cells assayed with [³H]NECA (Klotz et al., 1998). Similarly, little GTPS-sensitive [³H]CGS21680 binding is detected to A_2AARs transfected into COS-7 cells or human embryonic kidney 293 cells, suggesting that few receptors are coupled to G proteins (Piersen et al., 1994; Rosin et al., 1996). The inability to detect the high-affinity agonist binding conformation of the hA_2AAR may have resulted in an underestimation of the relative affinity of agonists for hA_2AAR compared with hA₁ARs and hA₃ARs (Sullivan et al., 2001).

The composition of G protein -subunits influences the potency of to stimulate guanine nucleotide exchange in assays with A_2AAR-G protein complexes such that G₄ is more potent than G₁ (McIntire et al., 2001). This prompted us investigate the influence of G protein -subunit composition on the degree of coupling to recombinant A_2A receptors. Recombinant baculoviruses encoding the hA_2AAR and three heterotrimeric G protein subunits were overexpressed in Sf9 cells. Expression of _s42 with hA_2AAR results in well-coupled receptors. We have used membranes from these Sf9 cells to investigate the affinities of various agonists for the high-affinity conformational state of hA_2AARs.

	Experimental Procedures

Top Abstract Introduction Experimental Procedures Results Discussion References

Materials. ZM241385 (Poucher et al., 1995) was a gift from Simon Poucher (Astra-Zeneca Pharmaceuticals, Cheshire, UK). Carrier-free ¹²⁵I-ZM241385 and [¹²⁵I]APE were synthesized and purified using high-performance liquid chromatography as described previously (Linden et al., 1984; Sullivan et al., 1999). ATL 146e was prepared as described previously (Rieger et al., 2001). MRS 1220 (Jacobson, 1998) was a gift from Kenneth Jacobson (National Institutes of Health; Bethesda, MD). CGS21680, NECA, XAC, CPA, and IB-MECA were purchased from Sigma/RBI (Natick, MA).

Cell Culture and Membrane Preparation. Sf9 cells were cultured in Grace's medium supplemented with 10% fetal bovine serum, 2.5 µg/ml amphotericin B, and 50 µg/ml gentamycin in an atmosphere of 50% N₂/50% O₂. Viral infection was performed at a density of 2.5 × 10⁶ cells/ml with a multiplicity of infection of two for each virus used. Infected cells were harvested 3 days postinfection and washed twice in insect PBS, pH 6.3. Cells were then resuspended in lysis buffer [20 mM HEPES, pH 7.5, 150 mM NaCl, 3 mM MgCl₂, 1 mM -mercaptoethanol, 5 µg/ml leupeptin, 5 µg/ml pepstatin A, 1 µg/ml aprotinin, and 0.1 mM PMSF] and snap-frozen for storage at 80°C. Cells were thawed on ice, brought to 30 ml of total volume in lysis buffer, and burst by N₂ cavitation (600 psi for 20 min). A low-speed centrifugation was performed to remove any unlysed cells (1000g for 10 min), followed by a high-speed centrifugation (17,000g for 30 min). The pellet from the final centrifugation was homogenized in buffer containing 20 mM HEPES, pH 8, 100 mM NaCl, 1% glycerol, 2 µg/ml leupeptin, 2 µg/ml pepstatin A, 2 µg/ml aprotinin, 0.1 mM PMSF, and 10 µM GDP using a small glass homogenizer followed by passage through a 26-gauge needle. Membranes were aliquoted, snap frozen in liquid N₂, and stored at 80°C. Membranes from cells stably expressing the human A₁ AR (Chinese hamster ovary K1 cells) or A₃ AR (human embryonic kidney 293 cells) were prepared as described previously (Robeva et al., 1996).

Western Blotting. For each membrane preparation, 100 µg of membrane protein was added to 2× electrophoresis buffer (20% glycerol, 150 mM Tris, 0.05 mg/ml bromphenol blue, 4% SDS) plus 1 mM -mercaptoethanol, loaded onto 10% Tris-Glycine Gradigels and electrophoresed at a constant voltage of 150 V for 90 min. Samples were transferred onto Westran polyvinylidene difluoride membranes (Schleicher and Schuell) using a constant current of 150 mA for 90 min. Nonspecific sites were blocked by incubating blots overnight at 4°C in a solution of TBST (50 mM Tris, 150 mM NaCl, and 0.5% Tween 20) containing 5% milk at pH 8. Blots were rinsed 4 × 5 min in TBST and then incubated with the primary antibody (NEN808 for _common and NEI800 for _common) in 2% milk in TBST. Blots were again rinsed 4 × 5 min with TBST before incubating for 90 min with donkey anti-rabbit IgG-horseradish peroxidase-linked F(ab')₂ at a dilution of 1:3000. Blots were rinsed 3 × 5 min in TBST, exposed to enhanced chemiluminescence reagents for 1 min, and placed on Kodak X-ray film for 15 s.

Radioligand Binding Assays. Radioligand binding to recombinant human A_2A receptors in Sf9 cell membranes was performed using either the radiolabeled agonist [¹²⁵I]APE (Luthin et al., 1995) or the radiolabeled antagonist ¹²⁵I-ZM241385. To detect the high-affinity, GTPS-sensitive state of A₁ and A₃ AR, we used the agonist [¹²⁵I]ABA (Linden et al., 1985, 1993). Binding experiments were performed in triplicate with 5 µg (A_2A) or 25 µg (A₁ and A₃) membrane protein in a total volume of 0.1 ml HE buffer (20 mM HEPES and 1 mM EDTA) with1 U/ml adenosine deaminase and 5 mM MgCl₂ with or without 50 µM GTPS. Membranes were incubated with radioligands at room temperature for 3 h (for agonists) or 2 h (for antagonists) in Millipore Multiscreen 96-well GF/C filter plates and assays were terminated by rapid filtration on a cell harvester (Brandel, Gaithersburg, MD) followed by four 150-µl washes over 30 s with ice-cold 10 mM Tris-HCl, pH 7.4, 10 mM MgCl₂. Nonspecific binding was measured in the presence of 50 µM NECA. For binding isotherms, nonspecific binding and free radioligand were fit by least-squares regression to a straight line. The extrapolated fit value of nonspecific binding for each free concentration of radioligand was subtracted from total binding to calculate specific binding. Equilibrium binding assays using [¹²⁵I]APE were carried out using isotope dilution (100 nM unlabeled I-APE and 5 nM [¹²⁵I]APE before serial dilutions) to create a range of radioligand concentrations useful for detecting both high- and low-affinity binding sites. Saturation binding assays using ¹²⁵I-ZM241385 did not require isotope dilution. B_max and K_D values were fit using nonlinear least-squares interpolation (Marquardt, 1963) for single or two-site binding models. For curvilinear 2-site Scatchard analyses (plots of [L]_bound versus [L]_bound/[L]_free), [L]_bound/[L]_free was calculated from specific binding using the quadratic equation Y = B + () / 2A, where Y = [L]_bound/[L]_free, A = K_D1 × K_D2, B = X(K_D1 + K_D2)  B_max1 × K_D2  B_max2, C = X × (X  B_max1  B_max2), and X = specific binding. Optimal parameter values were determined by nonlinear least-squares interpolation.

	Results

Top Abstract Introduction Experimental Procedures Results Discussion References

To produce membranes containing hA_2AAR that are partially coupled to G proteins, Sf9 cells were simultaneously infected with a baculovirus encoding the hA_2AAR or four baculoviruses encoding the receptor and the heterotrimeric G protein subunits _s, ₂ or ₄, and ₂. The amount of receptor expression was determined from saturation binding isotherms using both the agonist [¹²⁵I]APE and the antagonist ¹²⁵I-ZM241385 as radioligands. The data from antagonist saturation experiments (found in Table 1) shows that the neither the presence of G proteins nor addition of GTPS significantly changed the B_max for ¹²⁵I-ZM241385.

View this table: [in this window] [in a new window]   TABLE 1 Saturation binding parameters of hA2AAR expressed with or without G proteins Binding parameters of a selective A2A antagonist (125I-ZM241385) or agonist ([125I]APE) to Sf9 cell membranes expressing A2A receptors with or without G proteins (s,4 or 1,2) and in the presence or absence of 50 µM GTPS as described in Experimental Procedures. KH and KL represent the KD values of the high- and low-affinity states, respectively. The percentage of coupled receptors is determined by dividing the number of receptors in the high-affinity state (Bmax,H) by the total number of receptors (Bmax for 125I-ZM241385). Values represent means ± S.E.M.; n = 3 to 5.

The data from agonist saturation binding summarized in Table 1 shows that the A_2A receptor, when expressed alone, has a relatively low affinity for [¹²⁵I]APE (K_D = 27 nM). Scatchard analysis indicates that all agonist binding is optimally fit to a single affinity site. When G proteins are coexpressed, the receptor displays two different affinities for the agonist. Table 1 shows that G₁- and G₄-containing membranes have similar high-affinity (K_H) binding dissociation constants for [¹²⁵I]APE of 0.32 nM and 0.46 nM, respectively. The low-affinity (K_L) dissociation constants also are similar (K_L = 10.5 nM and K_L = 26 nM). Neither K_H nor K_L are significantly different between membranes expressing G₁ or G₄.

The total number of receptors measured varied with the choice of ligand. The B_max using ¹²⁵I-ZM241385 is significantly greater than that using [¹²⁵I]APE. This is probably caused by partial dissociation of the agonist ligand from the low-affinity site during washing of glass fiber filters. This phenomenon would cause the low-affinity B_max values to appear artificially depressed without affecting the observed K_D. Thus, to determine the fraction of receptors found in the high-affinity state, we divided the high-affinity B_max by the total number of receptors as determined by saturation binding of ¹²⁵I-ZM241385 (B_{max, ZM}). This analysis shows that the G subunit influences the efficiency of R-G coupling in this system. The fraction of receptors found in the high-affinity state in the presence of G₄ (40%) was significantly higher than that seen when ₁ was present (8%) (Table 1). This difference cannot be attributed to a difference in the amount of expressed - or -subunits, because the amount of protein detected in Sf9 membranes by Western blotting with antibodies that detect  or  subunits was similar (Fig. 1). The anti- antibody detects ₁ and ₄ with the same affinity because it recognizes a common epitope. Because the G₄-containing membranes had a greater fraction of coupled receptors, we used them to characterize agonist high-affinity binding sites in subsequent experiments.

View larger version (26K): [in this window] [in a new window]   Fig. 1.   Western blot for G (left) and G (right) in membrane preparations of Sf9 cells infected with baculoviruses encoding the recombinant hA2AAR alone (lane 1), receptor plus s, 1, and 2 (lane 2), or receptor plus s, 4, and 2 (lane 3). Electrophoresis, transfer, and blotting were performed as described under Experimental Procedures. The result shown is typical of triplicate experiments.

The equilibrium binding of [¹²⁵I]APE to membranes derived from Sf9 cells quadruply infected with hA_2AAR-_s42 complexes with or without GTPS is shown in Fig. 2A. Coinfection with G protein subunits substantially increased the amount of specific binding at a given concentration without changing nonspecific binding. In the absence of GTPS, specific binding to R-G complexes is fit significantly better to a two-site binding model (Fig. 2A, ) than to a single site equation (Fig. 2A, - -  - -). The two [¹²⁵I]APE affinity states of hA_2AAR-_s42 complexes are more clearly evident in the Scatchard plot shown in Fig. 2B. Treatment of membranes expressing these receptor-G protein complexes with GTPS (50 µM) eliminates the high-affinity site completely, resulting in equilibrium binding that is optimally fit to a single-site equation (K_D = 32 nM) and characterized by a linear Scatchard plot. This affinity is similar to the low-affinity population of receptors detected in the absence of GTPS (26 nM) or the single low-affinity site detected when the receptor is expressed alone (27 nM). This GTPS sensitivity of binding confirms that the high-affinity site is due to the interaction of the receptor with G proteins.

View larger version (27K): [in this window] [in a new window]   Fig. 2.   [125I]APE saturation binding to hA2AAR-s42 complexes. Sf9 cells were infected with recombinant baculoviruses expressing the human A2AAR and the G protein subunits s, 4, and 2. Membranes were prepared 72 h after infection and radioligand binding was performed as described under Experimental Procedures. A, equilibrium binding of [125I]APE in the absence () or presence () of GTPS. Nonspecific binding was assayed by addition of saturating concentration (50 µM) NECA in the absence () or presence () of GTPS. The solid line through solid squares represents a two-site fit of the equilibrium binding data, whereas the dotted line represents a one-site fit. B, Scatchard plot of equilibrium binding in the absence () or presence () of GTPS. Binding parameters from triplicate experiments are summarized in Table 1.

We next conducted kinetic experiments to determine k₁ and k₁ for [¹²⁵I]APE binding hA_2AAR-_s42 complexes. The kinetics of [¹²⁵I]APE (0.34 nM) association is illustrated in Fig. 3A. At this concentration, [¹²⁵I]APE binds with a k_obs of 0.0673 ± 0.0025 min¹. The binding is well fit by a single exponential equation, which is confirmed by the linearity of the transformed data shown in the Fig. 3, inset. The slope of this line defines a pseudo-first-order rate constant (k₁ × [[¹²⁵I]APE] + k₁) and is calculated based on the assumption that free radioligand does not change significantly over time. During this experiment, only 15% of the radioligand was bound at equilibrium. The dissociation of [¹²⁵I]APE is shown in Fig. 3B. The data is well fit by a single exponential equation, which is consistent with the proposition that a large fraction of the total binding is to the high-affinity site. We calculated k₁ to be 0.0291 ± 0.002 min¹ corresponding to a t_1/2 of 24 min. A kinetic analysis of [¹²⁵I]APE dissociation from the low-affinity site could not be accurately performed because of rapid dissociation of the radioligand.

View larger version (26K): [in this window] [in a new window]   Fig. 3.   Time courses of 0.34 nM [125I]APE association with and dissociation from A2AAR-s42 complexes. Sf9 cells were infected with recombinant baculoviruses expressing the human A2AAR and the G protein subunits s, 4, and 2. Membranes were prepared 72 h after infection and radioligand binding was performed as described under Experimental Procedures. A, specific binding of [125I]APE at time t after addition of radiolabeled agonist to membranes. Inset, a linear regression of transformed data. B, specific binding of [125I]APE at time t after dissociation is induced by addition of saturating antagonist (50 µM ZM241385) to radioligand/membrane solution which had be incubated for 3 h. Inset, log transformation of the data in B with a linear regression of the transformed data, illustrating dissociation from a single site. The data shown is typical of three to five experiments.

From k_obs and k₁ of the high-affinity site, and using (k_obs = k₁ × [[¹²⁵I]APE]+ k₁) we calculated k₁ as 1.12 × 10⁸ ± 0.09 × 10⁸ min/M. The K_D for [¹²⁵I]APE was calculated (K_D = k₁/k₁) to be 0.37 ± 0.02 nM, which is similar to the K_D determined by equilibrium binding (0.46 nM). Thus, there is good agreement between equilibrium and kinetic binding parameters. By a similar analysis, we determined the kinetic binding parameters for hA_2AAR-_s12 complexes: k₁ = 1.40 × 10⁸ ± 0.12 × 10⁸ min/M, k₁ = 0.0295 ± 0.003 min¹, and K_D = 0.21 ± 0.06 nM. These values are not significantly different from those for hA_2AAR-_s42.

We also observed two affinity sites of the receptor based on competition binding assays with agonists. Figure 4 demonstrates that when ¹²⁵I-ZM241385 binds to membranes expressing the hA_2AAR and G proteins, competition by the agonist, ATL 146e, for binding sites is biphasic. To accurately determine both K_i values from these biphasic curves, we used a curve-fitting algorithm that directly calculated both K_i values based on the K_D of the radioligand and correcting for the depletion of both the radioligand and the competing compound during binding (see Experimental Procedures). When fit to a two-site model, the competition curve yields two K_i values for the potent A_2A ligand ATL 146e: 0.18 and 58 nM. A comparison of [¹²⁵I]APE and ATL 146e reveals a potency difference of 2.6- and 2.1-fold, respectively, for the high- and low-affinity hA_2AAR binding sites. Addition of GTPS in competition assays completely eliminates the high-affinity K_i, leaving a low-affinity K_i (58 nM), which is consistent with the K_i for ATL 146e when competing for ¹²⁵I-ZM241385 binding sites in membranes expressing the receptor alone (68 nM, Table 2). The high-affinity K_i of ATL 146e in competition for ¹²⁵I-ZM241385 (0.18 nM) is not significantly different from the K_i calculated from competition for [¹²⁵I]APE (0.20 nM).

View larger version (22K): [in this window] [in a new window]   Fig. 4.   Competition of ATL 146e for binding sites on hA2AAR. Binding is plotted as a fraction of control specific binding. Data for ATL 146e competition for 125I-ZM241385 binding on Sf9 membranes expressing the A2A receptor alone () (Ki = 68 nM) or A2AAR-s42 complexes () (Ki, H = 0.18 nM and Ki,L = 58 nM) and [125I]APE binding on membranes expressing A2AAR-s42 complexes () (Ki = 0.20 nM) are representative of three to six replicate experiments.

View this table: [in this window] [in a new window]   TABLE 2 Comparison of high- and low-affinity Ki values for selected agonist compounds at the hA2AAR KiL (expressed as mean ± S.E.M.) values were determined from the IC50 of the agonist in competition with 125I-ZM241285 on A2A receptors expressed in Sf9 cells. High-affinity KiH values were determined from the IC50 of the drug in competition with [125I]APE on A2A receptors expressed with heterotrimeric G proteins (s,4,2) in Sf9 cells.

To determine the affinities of other adenosine receptor agonists at G protein-coupled A_2A receptors, we performed competition experiments using low concentrations (0.3-0.5 nM) of [¹²⁵I]APE as the radioligand. At the concentrations used, the radioligand was bound predominantly (>95%) to the high-affinity site. We chose several agonists that are widely used experimentally because they have been reported to be highly selective for binding to one of the four adenosine receptor subtypes: A_2A-selective, ATL146e and CGS21680; A₃ selective, IB-MECA and Cl-IBMECA; and A₁ selective, CPA and CCPA. We also examined the nonselective agonist, NECA. As expected, these agonists have a much higher affinity for the G protein coupled hA_2AAR than for the A_2A receptor expressed without any G proteins. A typical experiment is shown in Fig. 5 and the results of 60 experiments are summarized in Table 2.

View larger version (22K): [in this window] [in a new window]   Fig. 5.   Competition of adenosine receptor agonists for [125I]APE binding to A2AAR-s42 complexes. Binding is plotted as a fraction of control specific binding. Data for ATL 146e (), NECA (), CGS 21680 (), and IB-MECA () are shown. The data was fit by nonlinear regression to a one-site binding model. Binding parameters from replicate experiments are summarized in Table 2.

We also examined four antagonists in competition assays: theophylline, MRS1220, XAC, and ZM241385. The results are summarized in Table 3. The K_i values of antagonists do not vary significantly between the coupled and uncoupled receptors. These results are consistent with the expectation that antagonists do not differ substantially in their affinities for coupled and uncoupled receptors.

To examine the selectivity of selected compounds for various adenosine receptor subtypes, we performed competition radioligand binding experiments to membranes expressing the human A₁AR or A₃AR using the agonist [¹²⁵I]ABA. At the [¹²⁵I]ABA concentrations used, >80% of radioligand binding in these systems is GTPS sensitive and therefore predominantly represents the high-affinity binding sites of these receptors. Resultant competition curves were fit with both one- and two-site models. In all cases, the one-site model better fit the data as determined by F tests (Motulsky and Ransnas, 1987). K_i values were determined as described under Experimental Procedures. The high-affinity K_i values from these experiments are summarized in Table 4. We also calculated the selectivity ratio of each agonist at high-affinity A₁ and A₃ receptors relative to both their low and high A_2A affinities. The results indicate that agonists that are considered to be selective for A₁ or A₃ receptors are less selective over A_2A receptors than previously noted.

	Discussion

Top Abstract Introduction Experimental Procedures Results Discussion References

We have developed a system in which the G protein-coupled state of the A_2AAR can be accurately detected using a quadruple infection of Sf9 cells with baculoviruses encoding the hA_2AAR and three G protein subunits. Previous studies using transfected COS-7 cells (Piersen et al., 1994) or HEK293 cells (Rieger et al., 2001; Sullivan et al., 2001) display little or no GTPS-sensitive agonist binding, possibly because of the low levels of G_s expressed in these cell lines relative to the highly expressed receptor or to other factors that limit coupling of receptors to G_s.

In this study, we show that the identity of the  subunit can influence the coupling efficiency of the A_2AAR. Overexpressing _s12 with the A_2AAR in Sf9 cells resulted in only partially coupled A_2AAR (8%), but substantially greater coupling was observed (40%) if ₁ was replaced by ₄. This difference was not due to variations in protein expression levels as determined by Western blots. The results are consistent with the recent report showing in reconstitution experiments that ₄ is significantly more potent than ₁ in stimulating agonist-induced guanine nucleotide exchange in Sf9 membranes expressing hA_2AAR (McIntire et al., 2001). Our data imply that the difference in guanine nucleotide exchange is a consequence of changes in the stability of the agonist-receptor-G protein complex, which is dependent on the composition of the G subunit. G composition also influences the coupling of M₂ muscarinic receptors such that guanine nucleotide exchange on _o42 is greater than on _o12 (Hou et al., 2001). The influence of G protein composition on A_2A receptor coupling has not been as extensively studied as the A₁ receptor. A₁ receptors couple preferentially to G proteins containing ₂ or ₃ over subunits that contain ₁ (Figler et al., 1997). Additional experimentation will be required to determine whether, like the A₁ receptor, the coupling of A_2A receptors is influenced by the composition of G.

We used _s42 to carefully examine the high-affinity binding site of hA_2AARs. Several lines of evidence support the conclusion that the high-affinity binding site observed in this study results from G protein coupling: 1) the agonist used ([¹²⁵I]APE) was previously shown to bind to both coupled and uncoupled A_2AAR receptors on rat striatal membranes (Luthin et al., 1995); 2) two affinity states were detected in both saturation and competition studies; 3) high-affinity binding to quadruply infected Sf9 membranes was completely inhibited by the addition of 50 µM GTPS to binding assays; 4) the high-affinity site was absent in membranes expressing the receptor alone; and 5) agonists, but not antagonists, bound differentially to coupled and uncoupled receptors.

It is significant that agonist radioligands of the A_2A receptor such as [¹²⁵I]APE and the widely used compound [³H]CGS21680 bind with high enough affinity to detect both uncoupled and coupled receptors. Consequently, attempts to fit radioligand binding data to a single site may result in detection of a composite apparent binding site that is intermediate in its affinity for coupled and uncoupled receptors. This will result in discrepancies in agonist binding constants that depend on various factors, including the receptor density, the fraction of coupled receptors, the concentration of radioligand, and the time and temperature of filter washing. The absolute affinity of the radioligand for the low-affinity site will influence its detection because the lowest affinity ligands are most prone to washing off the receptor during the wash phase of the filtration process. In this regard, it is notable that the rank affinity of agonists radioligands for the low-affinity site is: [¹²⁵I]APE (26 nM) < [³H]NECA (85 nM) < [³H]CGS21680 (944 nM). Collectively, this may explain the wide differences in reported agonist binding affinities for the A_2A receptors between laboratories and the fact that the high-affinity K_i values found in this study are significantly lower than those reported previously (Robeva et al., 1996; Klotz et al., 1998) because previous studies have not had the benefit of a well-coupled receptor system and thus could not explicitly detect the high-affinity state.

We measured both the low- and high-affinity K_i values of several widely used adenosine receptor agonists. It is notable that the ratio of binding affinity for the high- and low-affinity sites was highly variable among the agonists we examined and ranged from 862-fold with IB-MECA to 42-fold with NECA (Table 2). This is a significant observation because it indicates that even the relative affinities and rank order potency of agonists vary depending on the coupling state of receptors. Stated another way, these data indicate that binding to uncoupled A_2A receptors is not highly predictive of binding to coupled receptors. The reason for the variance in the ratio between high- and low-affinity K_i values is interesting and is the subject of ongoing work in our laboratory.

For comparing the selectivity of agonists for the four adenosine receptor subtypes, it makes sense to compare coupled receptors to coupled receptors, or uncoupled receptors to uncoupled receptors. In contrast to recombinant hA_2AARs, recombinant hA₁ARs and hA₃ARs do couple sufficiently well to G proteins to readily allow detection of the high-affinity agonist binding conformation characterized by agonist binding that is largely inhibited by GTPS (Gao et al., 1999). This may be related to the fact that G_i/o is more abundant in mammalian cells than is G_s. In addition, the G_i/o proteins seem to couple very tightly to their cognate receptors (Munshi et al., 1991; Gao et al., 1999). Consequently, the selectivity of agonists has been assessed in many instances based on competition of radioligand binding to well-coupled A₁ and A₃ receptors, but to poorly coupled A_2A receptors. Based on a comparison of well-coupled A₁, A_2A, and A₃ receptors, we have shown in this study that among agonists generally used as A_2A-selective ligands (ATL146e and CGS21680), the limited apparent selectivity for human A_2A receptors is actually substantially higher than previously thought (Sullivan et al., 2001). Moreover, the high-affinity K_i values for agonists in our experiments are more in line with their functional EC₅₀ values in other studies (Walker et al., 1997; Sullivan et al., 2001) than the low-affinity K_i values determined here and by others.

We also examined agonists reported be highly A₁ and A₃ selective. We confirmed that the radioligand binding in these assays is to a single GTPS-sensitive site. CCPA and CPA both have subnanomolar affinities for the coupled A₁AR and have been described as highly selective agonists of the A₁ receptor. Previous studies have cited selectivity ratios for A₁ over A_2A receptors of over 300-fold for CPA and over 2000-fold for CCPA (Klotz et al., 1998). However, in this study we have shown that both of these compounds have only a 30- to 50-fold selectivity for A₁ over A_2A at their high-affinity binding sites. This finding would seem to explain the effect of CCPA to increase interleukin-10 release and decrease tumor necrosis factor- concentrations in endotoxemic mice at high concentrations (Hasko et al., 1996), actions that have been shown to be A_2A dependent in vitro (Bouma et al., 1994; Sullivan and Linden, 1998).

Similarly, examination of the high-affinity K_i values for IB-MECA and Cl-IB-MECA shows that neither is as selective for A₃ over A_2A receptors as has been reported (Gallo-Rodriguez et al., 1994; Klotz et al., 1999). Both of these agonists have only a ~5-fold selectivity for A₃ over A_2A for binding to the high-affinity site. This finding has critical importance for work involving discrimination of the physiological roles of the A₃ and A_2AARs in mediating the protective effects of adenosine. Various groups have relied on IB-MECA to define the role that the A₃AR receptor plays modulating inflammatory responses (Hasko et al., 1996, 1998; Sajjadi et al., 1996). This was reasonable based on the relative potencies of these compounds reported in the prior literature. However, because the A_2A and A₃AR work through opposing mechanisms (G_s versus G_i), it seems unlikely that both receptors could have the generally anti-inflammatory effects noted in the same cells. The potency of IB-MECA at the A_2AAR confirms our recent observation that the anti-inflammatory effects of IB-MECA on tumor necrosis factor- production in human monocytes can be potently blocked by the selective A_2A antagonist, ZM241385 (Sullivan and Linden, 1998).

Comparing K_i values of compounds is an excellent method for determining selectivity of agonist binding at various receptors. However, this does not give a complete picture of agonist action because these values do not incorporate the efficacy of the various agonists at each receptor. A given agonist could have equal potencies at two given receptors, but if its efficacy at one is much greater than at the other, the agonist would seem to be more potent at one receptor in functional assays. Because no work has been published on the relative efficacies of these compounds, we do not yet have a complete understanding of these agonists' actions.

In summary, we report here on the establishment of a method for expressing G protein coupled hA_2AARs by quadruple infection of Sf9 cells and sensitivity of this coupling to G composition. We have confirmed the existence of this high-affinity state by Scatchard analysis, competition assays, and kinetic experiments. Using this system, we have determined the affinity of several commonly used adenosine receptor agonists at the coupled A_2AAR. This work has demonstrated that IB-MECA and Cl-IB-MECA can no longer be assumed to be highly selective agonists of the hA₃AR or CCPA, a highly selective agonist of the hA₁AR. Consequently, findings based on the use of these agonists must be critically evaluated with respect to possible involvement of the A_2AAR in responses that have been previously ascribed to A₁ or A₃ receptors. We anticipate that the use of methods to express well coupled adenosine receptors will be valuable for detecting novel potent and selective AR agonists.