Abstract
Suramin analogues uncouple two Gi/Go-coupled receptors, the D2 dopamine receptor in rat striatum and the A1 adenosine receptor in human cerebral cortex, with distinct structure-activity relations. This discrepancy may reflect true differences in the affinity of the analogues for specific receptor/G protein complexes or may be attributable to differences in species or in the tissue source used. We addressed this question by using human embryonic kidney 293 cells that stably express the human A1 and rat A1 receptor and the human D2 receptor. Suramin is 10-fold more potent than its didemethylated analogue NF037 in inhibiting the interaction between G proteins and the rat A1 or human A1 receptor; in contrast, both compounds are equipotent in uncoupling the D2 receptor. These differences are observed regardless of whether (1) inhibition of high affinity agonist binding to the receptors or (2) agonist-stimulated GTPγS binding is used as readout, (3) the receptors are allowed to interact with the G protein complement in human embryonic kidney 293 cell membranes, or (4) the receptors are forced to interact with a defined G protein α subunit (i.e., after reconstituting pertussis toxin-treated membranes with exogenous rGiα-1). The apparent affinity of suramin depends in a linear manner on receptor occupancy, which shows that suramin and the receptor compete for the G protein. Finally, the affinity of the receptors for rGiα-1 (human A1 > rat A1 > human D2) is inversely correlated with the potency of suramin in uncoupling ternary complexes formed by these receptors and thus determines the selectivity of the suramin analogues for specific receptor/G protein tandems.
In most cells, G protein-coupled receptors interact with multiple distinct G protein oligomers, and the overall biological response to the agonist-activated receptor results from the concerted regulation of multiple G protein-dependent effector systems (Gudermann et al., 1996a). The ability of a receptor to activate multiple G proteins is specified by discrete, poorly conserved regions in the intracellular loops that connect the transmembrane helices and, in some cases, within the carboxyl terminus of the receptor (Gudermann et al., 1996b). Based on the observation that there is more than one active conformation of rhodopsin (Arnis et al., 1993, 1994), it has been proposed that other G protein-coupled receptors also may adopt several active conformations that interact with distinct G proteins; these may be selected by “biased” agonists, which will favor a conformation that preferentially interacts with one type of G protein (Gudermann et al., 1996a; Kenakin, 1996). This hypothesis is supported by two lines of experimental evidence. (1) Point mutations in a given receptor can elicit distinct effects on its coupling properties; for example, substitution of Ile486 by phenylalanine in the human thyroid-stimulating hormone receptor produces a receptor that constitutively activates both the adenylyl cyclase and phospholipase C cascade; in contrast, substitution of Phe631 by isoleucine only raises cAMP (Parma et al., 1995). (2) The PACAP receptor I can be stimulated by PACAP-27 and PACAP-38; however, on heterologous expression of the receptor, PACAP-27 activates adenylyl cyclase more potently than PACAP-38, whereas the reverse is true for stimulation of inositol trisphosphate formation (Spengleret al., 1993). Other arguments and additional experimental evidence in support of the hypothesis that multiple R* conformations exist have been reviewed recently (Gudermann et al., 1996a;Kenakin, 1996). A corollary of this concept is the assumption that compounds that block the interaction of R with G may be selective for specific R/G tandems. If compared with receptor antagonists, compounds that block the interaction of receptors and G proteins over receptor antagonists offer the advantage that they should provide for an additional level of selectivity in inhibiting signal transduction; provided that inhibitors with high selectivity can be found, they will block signaling of the activated receptor via one G protein-regulated pathway but will not perturb other receptor-generated signals within the cell.
Earlier work showed that suramin acted as an inhibitor of receptor/G protein coupling (Butler et al., 1988; Huang et al., 1990); circumstantial evidence for selective disruption of specific receptor/G protein tandems was provided by the observation that suramin inhibited the activation of pertussis toxin-substrate G proteins by δ-opioid receptors in membranes from NG 108–15 cells, whereas the stimulation of the guanine nucleotide exchange reaction of these Go/Gi proteins by serum factors, which acted on an unidentified receptor, was not blocked (Huang et al., 1990). In addition, we recently reported that the didemethylated suramin derivative NF037 discriminated between A1 adenosine receptor/G protein tandems in the human cerebral cortex and D2 dopamine receptor/G protein tandems in the rat striatum (Beindl et al., 1996). However, in rat cerebral cortex, the A1 adenosine receptor is resistant to the uncoupling effect of suramin unless the membranes are extracted with detergent to remove an inhibitory constraint imposed by an ancillary protein (Nanoff et al., 1997). Thus, the distinct activity profile of suramin and NF037 on human A1 adenosine and rat D2 dopamine receptor may have been due to species differences or may have arisen from the expression of the receptors in distinct cell types and/or in distinct microcompartments of the plasma membrane. Here, we eliminated these confounding variables by expressing the human and rat receptors in the same cell line; furthermore, the receptors were forced to interact with the same G protein α subunit. The results show that NF037 is selective for D2dopamine receptor/G protein tandems even if the receptors couple to the same Giα subtype. In addition, the potency of suramin and NF037 in uncoupling receptor/G protein complexes is inversely correlated to the affinity of the receptor for the G protein.
Experimental Procedures
Materials.
[35S]GTPγS, [125I]OH-PIPAT [(+)-trans-7-hydroxy-2[N-propyl-N-3-[125I]iodo-2′-propenyl)aminotetralin] and [125I] were purchased from New England Nuclear Research Products (Boston, MA). [125I]HPIA [(−)N6-3[125I](iodo-4-hydroxyphenyl-isopropyl)adenosine] was synthesized according to Linden (1984). Guanine nucleotides and adenosine deaminase were from Boehringer-Mannheim Biochemica (Mannheim, Germany). 1-O-n-Octyl-β-d-glucopyranoside (octylglucoside), CHAPS, and HEPES were from BIOMOL (Munich, FRG). Suramin, sulpiride, and XAC were obtained from Research Biochemicals (Natick, MA). The materials required for SDS-polyacrylamide gel electrophoresis were from BioRad (Richmond, CA). Fetal calf serum was from PAA Laboratories (Linz, Austria). Dulbecco’s modified Eagle medium, nonessential amino acids, β-mercaptoethanol, and G418 (geneticin) were obtained from GIBCO-BRL (Grand Island, NY). CPA, pertussis toxin, l-glutamine, penicillin G, and streptomycin were purchased from Sigma Chemical (St. Louis, MO). Buffers and salts were from Merck (Darmstadt, FRG). The cDNA coding for the rat A1 adenosine receptor in the plasmid vector pBC-A1R (Freund et al., 1994) and the HEK 293 cell clone expressing the human A1 adenosine receptor were kindly provided by M. J. Lohse (University of Würzburg). The human D2 (short isoform)plasmid vector and NF037 were generous gifts of C. Pifl (Institute of Biochemical Pharmacology, Vienna University) and of P. Nickel (Institut of Pharmaceutical Chemistry, University of Bonn), respectively. The vectors pEGFP-C1 and pRc-CMV were obtained from Clontech (Palo Alto, CA).
Generation of transient and stable cell lines.
COS-7 (African green monkey kidney fibroblasts) cells were plated at a density of 3 × 106 cells/10-cm dish and transiently transfected with 5 μg of the cDNAs pBC-A1dhfr containing the rat A1 adenosine receptor cDNA insert (Freundet al., 1994) and pCMV5 plasmid vector containing the D2short receptor cDNA using the calcium phosphate precipitation method (Chen and Okayama, 1988). The cells were harvested 48 hr after transfection; plasma membranes were prepared and used for radioligand binding assays. HEK 293 cells were plated at a density of 2.5 × 106 cells/10-cm dish and transfected with 7.5 μg of the plasmid pBC-A1dhfr (encoding the rat A1 adenosine receptor) and 0.75 μg of the resistance marker plasmid pRc-CMV carrying the neomycin phosphotransferase gene. Similarly, the plasmid encoding the short splice variant of the human D2 dopamine receptor was cotransfected with either pRc-CMV or pEGFP-C1, a vector carrying a red-shifted variant of wild-type green fluorescent protein cDNA from the jellyfish Aequoria victoria and a neomycin resistance cassette. The cells were grown in Dulbecco’s modified Eagle medium containing 10% fetal calf serum, 2 mml-glutamine, β-mercaptoethanol, nonessential amino acids, 100 units/ml penicillin G, and 100 μg/ml streptomycin at 5% CO2 and 37° for 16 hr. Thereafter, the medium was removed, and the cells were subjected to an osmotic shock by adding 15% glycerol in phosphate-buffered saline for a few seconds. Cells were grown for another 24 hr and subsequently selected by adding G418 (0.8 mg/ml) to the medium for 4–6 days. pEGFP-positive clones were identified by fluorescence microscopy. Positive clones appeared in bright green and were subjected to further selection to obtain clones with different expression levels. Three clones were selected that differed in D2 dopamine receptor density (ranging from ∼0.3 to 4 pmol/mg membrane protein).
Membrane preparation and protein purification.
Cells were grown to confluency in 10-cm tissue culture dishes, washed once with ice-cold phosphate-buffered saline, and scraped off their plastic support in HME buffer (25 mm HEPES·NaOH, pH 7.5, 2 mm MgCl2, 1 mm EDTA). After centrifugation at 20,000 × g for 10 min, the cell pellet was resuspended in HME, subjected to a freeze/thaw cycle with liquid nitrogen, and further homogenized by sonication. Membranes were sedimented by centrifugation (38,000 × g for 10 min) and resuspended in HME at a protein concentration of 8–10 mg/ml and stored in aliquots at −80°. Recombinant (R) Giα-1 and rGiα-2 were expressed in Escherichia coli BL21DE3 harboring a plasmid-encoding yeast myristoyl-CoA transferase and purified from bacterial lysates (Mumby and Linder, 1994). Oligomeric G proteins were purified from bovine or porcine brain, and free βγ dimers were chromatographically resolved from the α subunits (Casey et al., 1989).
Radioligand binding experiments.
Equilibrium binding with the A1 adenosine receptor agonist [125I]HPIA and with the dopaminergic D2 agonist [125I]OH-PIPAT were carried out in a final volume of 40 μl containing 50 mm Tris·HCl, pH 8, 1 mm EDTA, 5 mm MgCl2, 1 mm ascorbic acid, 8 μg/ml adenosine deaminase, 10 μg of membrane protein, and the concentrations of suramin and NF037 as indicated (ascorbic acid and adenosine deaminase are not required for determining binding to A1 adenosine and D2dopamine receptors, respectively, but were present in all incubations to obtain identical incubation conditions). The binding reaction was carried out for 90 min at 25° and terminated by filtration over glassfiber filters using a cell harvester (Skatron, Lier, Norway). Nonspecific binding was determined in the presence of 1 μm XAC (for A1 adenosine receptors) or 10 μm sulpiride (for D2 dopamine receptors) and amounted to ∼5–10% of total binding in theKD concentration range. In experiments using membranes from clones with a high receptor expression level (1.5–3.9 pmol/mg of membrane protein) or low radioligand concentrations, the amount of membrane protein added and the assay volume was adjusted (up to 250 μl) to avoid depletion of the radioligand (bound <10% of total). Specific binding of agonist or antagonist radioligands ([3H]DPCPX and [125I]epideprid for A1adenosine and D2 dopamine receptors, respectively) was not detectable in membranes prepared from untransfected HEK 293 cells. Radioligand binding to membranes from transfected cells was displaced by unlabeled receptor ligands with the appropriate pharmacological specificity, and specific binding for both agonists and antagonist radioligands was saturable;Bmax values for antagonist binding were ∼1.2-fold higher than those for agonist radioligands, indicating that the majority of the receptors were capable of interacting with G proteins endogenous to the HEK 293 membranes (not shown).
Receptor-mediated [35S]GTPγS binding.
Receptor-promoted binding of [35S]GTPγS was determined essentially as described previously (Nanoff et al., 1995). In brief, membranes from HEK 293 cells (∼10 μg) were suspended in 40 μl of buffer containing 25 mmHEPES·NaOH, pH 7.5, 1.5 mm MgCl2, 150 mm NaCl, 1 mm EDTA, 0.01 mmGDP, and the concentrations of dopamine, CPA, and suramin analogues indicated in the respective figures. After a preincubation of 10 min at 25°, the assay was initiated by adding 10 μl of buffer containing [35S]GTPγS to yield a final concentration of 1 nm (specific activity, 2000 cpm/fmol). The assay was terminated after 10 min by adding 0.5 ml of ice-cold stop buffer containing 10 mm Tris·HCl, pH 8.0, 100 mmNaCl, 20 mm MgCl2 and 0.1 mm GTP. Bound and free nucleotides were separated by filtration over glassfiber filters.
Determination of adenylyl cyclase activity.
Adenylyl cyclase activity in HEK 293 membranes expressing the recombinant receptors was assayed in 0.1 ml containing 50 mm HEPES·NaOH, pH 8.0, 0.05 mm [α-32P]ATP (∼200 cpm/pmol), 5 mm MgCl2, 0.1 mm rolipram, 10 mm creatine phosphate, membrane protein (25 μg), 1 mg/ml creatin kinase, 8 μg/ml adenosine desaminase, and 1% bovine serum albumin. Inhibitory regulation of adenylyl cyclase by the D2 dopamine and A1 adenosine receptor agonists was determined in the presence of 1 μm prostaglandin E1 and 10 nm GTPγS. The reaction was carried out for 20 min at 25°; cAMP was separated from ATP by sequential chromatography on Dowex and Alumina (Johnson and Salomon, 1991).
Pertussis toxin treatment and reconstitution of HEK 293 cell membranes with rGiα-1.
HEK 293 cells expressing the rat A1 adenosine, human A1adenosine, or human D2 dopamine receptors were incubated with 100 ng/ml pertussis toxin for 24 hr, and membranes were prepared as described. To insert exogenously added G protein into the membranes, the stable reconstitution protocol (Freissmuth et al., 1991a) was adapted as follows: PTX-treated membranes were incubated with 4.5 ng pf rGiα-1/μg membrane protein in HME containing 1% octylglucoside. After 1 hr on ice, membranes were diluted 1:10 in detergent-free buffer and centrifuged at 38,000 × g for 12 min. Pellets were resuspended in HME and stored in aliquots at a concentration of ∼10 mg/ml at −80°. The amount of rGiα-1 incorporated into the membranes was determined by immunoblotting. To assess the potency of rGiα-1 to restore high affinity agonist binding, rGiα-1 was combined with a 4-fold molar excess of purified βγ dimers in 1% octylglucoside (or 10 mm CHAPS); appropriate dilutions were added to the membranes to give 0.5% octylglucoside (or 5 mm CHAPS) and preincubated on ice for 15 min. Subsequently, radioligand binding assays were carried out after diluting the detergent 2-fold.
Immunoblots.
Membrane proteins (∼25 μg/lane) were separated on SDS-polyacrylamide gels (10% acrylamide, 0.13% bisacrylamide) and transferred to nitrocellulose membranes that were probed with AS7, an antiserum recognizing Giα-1and Giα-2 (McClue et al., 1992) or with the Giα-1-specific antiserum I1C (Selzeret al., 1993). The immunostained bands were visualized by enhanced chemoluminescence using an anti-rabbit IgG antibody conjugated to horseradish peroxidase (Amersham, Arlington Heights, IL). Purified recombinant G protein α subunits were used as standards. To verify that comparable amounts of membrane proteins had been applied in individual lanes, blots also were probed with a rabbit antiserum directed against the G protein β subunit (Hohenegger et al., 1995).
Results
Uncoupling of A1 adenosine and D2 dopamine receptors after heterologous expression.
High affinity binding of agonists to G protein-coupled receptors depends on the formation of a ternary complex of agonist, receptor, and G protein (Hepler and Gilman, 1992).After stable expression of the human D2dopamine and the rat and human A1 adenosine receptors in HEK 293 cells, the coupling of the receptors with G proteins in the membrane was assessed by using agonist radioligands. Suramin and NF037 did not block binding of appropriate antagonist radioligands to the receptors (not shown; see Beindl et al., 1996) but inhibited equilibrium binding of the A1-selective agonist [125I]HPIA and the D2-dopaminergic agonist [125I]OH-PIPAT (Fig.1); suramin (▪ in Fig. 1) was >10-fold more potent than NF037 (• in Fig. 1) in suppressing ternary complex formation of the rat (Fig. 1A) and human (Fig. 1B) A1 adenosine receptor. In contrast, the compounds were equipotent in inhibiting binding of the D2dopamine receptor agonist [125I]OH-PIPAT (Fig.1C). The same difference was seen if the rat A1adenosine receptor and human D2 dopamine receptor were transiently expressed in a cell line of nonhuman origin, namely, COS-7 cells (not shown). In addition, the apparent affinity of suramin and of NF037 was highest for human D2 receptor/G protein complexes and lowest for human A1receptor/G protein complexes (Table 1).
The agonist-liganded receptor catalyzes the GDP/GTP exchange reaction of the G protein; agonist-stimulated binding of [35S]GTPγS therefore can be used as an alternative readout to assess receptor/G protein coupling. The A1-selective agonist CPA stimulated [35S]GTPγS binding with EC50 values of 4.1 ± 2.4 and 161 ± 57 nm (not shown) in membranes harboring the human and the rat A1 adenosine receptor, respectively; after a 10-min incubation period, the receptor-promoted binding was ∼2.5-fold higher than the basal binding (Fig. 2, A–C). In membranes expressing the D2 dopamine receptor at low levels (used to generate the data shown in Fig. 1C), the dopamine-induced increment in [35S]GTPγS binding was too low (∼1.2-fold) to obtain a reliable signal-to-noise ratio for assessing the inhibitory effect of suramin and NF037. Hence, membranes from a cell clone that expressed the D2dopamine receptor at high levels (3.9 pmol/mg) were used where dopamine-stimulated basal [35S]GTPγS binding ∼2-fold (see Fig. 2D) with an EC50 value of 0.12 ± 0.02 μm. Fig. 2A summarizes experiments carried out with membranes harboring the rat A1adenosine receptor. The basal rate of [35S]GTPγS binding was determined in the presence of receptor antagonists (1 μm XAC or 5 μm sulpiride) to eliminate nucleotide exchange catalyzed by the unliganded receptor. Suramin and NF037 decreased basal [35S]GTPγS binding by ∼50% (Fig. 2A,open symbols); these findings are consistent with the ability of the compounds to directly block the release of GDP from G protein α subunits (Freissmuth et al., 1996). In contrast, suramin (IC50 = 1.5 ± 0.3 μm) was more potent than NF037 (IC50 = 15.9 ± 2.2 μm) in blocking [35S]GTPγS binding promoted by the activated rat A1adenosine receptor (Fig. 2B). Similarly, the apparent affinity of suramin was higher than that of NF037 when inhibition of [35S]GTPγS binding promoted by the agonist-liganded human A1 adenosine receptor (Fig. 2C) was determined. In contrast, the two compounds were equipotent in inhibiting the D2 dopamine receptor-stimulated guanine nucleotide exchange reaction. Higher concentrations of suramin and NF037 are required to inhibit receptor-promoted [35S]GTPγS binding than high affinity agonist binding (compare Figs. 1 and 2). This discrepancy is presumably due, in part, to the different assay conditions; that is, the catalytic turnover of the agonist-liganded receptor in the presence of a mixture of GTPγS and GDP (Fig. 2) versus stoichiometric interaction to form a ternary complex in the absence of guanine nucleotides (Fig. 1). Importantly, differences in receptor occupancy by the agonists contribute to the rightward shift of the inhibition curves (see also below); agonist radioligands were present at concentrations close to their KD values, whereas CPA and dopamine were used at saturating concentrations (300 nm and 1 μm, respectively) to promote [35S]GTPγS binding. If the human A1 adenosine receptor was activated with 10 nm CPA, the IC50value of suramin was 2.89 ± 0.77 μm (not shown), whereas it amounted to 8.78 ± 1.81 μm at 300 nm CPA (Fig.2C).
Stable reconstitution of high affinity agonist binding to membranes from pertussis toxin-treated cells by rGiα-1.
The different affinity of suramin and NF037 for ternary complexes formed by the A1 adenosine and D2 dopamine receptor may have been due to an interaction of the receptors with distinct G proteins. This possibility is substantiated by the following observations: both A1 adenosine and D2 dopamine receptors physiologically recruit adenylyl cyclase inhibition as one of the signaling pathways to elicit a biological response, and a marked (>50%) inhibition of prostaglandin E1-stimulated cAMP formation was observed for both receptor types in intact transfected HEK 293 cells (Roka F and Nanoff C, unpublished observations). However, attenuation of adenylyl cyclase activity was observed only on activation of the D2 dopamine receptor in HEK 293 membranes, whereas A1 adenosine receptors were inactive (data not shown). We therefore determined the expression of Giα in the transfected cells. Two forms of Giα were detected in HEK 293 membranes (Fig.3A). The top band commigrates with purified rGiα-1 and is recognized by I1C, an antiserum specific for Giα-1 (see Fig. 3B). The bottom band is detected only by the antiserum that reacts with Giα-1 and Giα-2 but not by antiserum I1C and commigrates with rGiα-2. The levels of Giα-2 and of the G protein β subunits (migrating as a β35/β36 doublet) were comparable in all cell lines. In contrast, there was some variability in the level of Giα-1; the membranes from the cell line that expressed the A1 adenosine receptor (Fig. 3A,lanes 4 and 5) had more Giα-1 than the control cells (Fig. 3A,lane 2), whereas those harboring the D2 dopamine receptor contained less Giα-1 (Fig. 3A, lane 3). However, this is most likely not related to an effect of receptor expressionper se; other cell clones that expressed more D2 receptors than that shown in Fig. 2A had Giα-1 levels comparable to those of untransfected control cells (not shown). This suggests that the variation in the amount of Giα-1 may be due to clonal selection.
To force the receptors to interact with a defined G protein α subunit, we disrupted the coupling of the receptors to the endogenous Giα subunits by pretreating the cells with pertussis toxin and subsequently incorporating exogenously added rGiα-1 into the membrane. ADP-ribosylation of Giα subunits, which occurs at a cysteine residue at position −4 from the carboxyl terminus, retards the migration of the proteins, which can be detected by gel electrophoresis under appropriate conditions (Linder et al., 1990); Giα-1, which was visualized in pertussis toxin-treated membranes (Fig. 3B), was shifted to a slightly lower mobility. Pertussis toxin-treated membranes were incubated with rGiα-1 in the presence of detergent followed by a dilution far below the critical micellar concentration of the detergent and its removal by centrifugation; this stable reconstitution resulted in incorporation of substantial amounts of the protein into the membranes (Fig. 3B, lanes labeled Rec). As expected, pertussis toxin treatment eliminated the high affinity binding of agonist radioligands (□ in Fig. 4). In membranes stably reconstituted with exogenous rGiα-1, high affinity agonist binding to the uncoupled receptors was restored. As shown in Fig. 4 (compare • and ○), the saturation isotherms showed varying reconstitution efficiencies for the different receptors; however, the affinities for the agonist radioligands were similar in native and reconstituted membranes. KD values (three experiments) were 0.7 ± 0.3 and 0.7 ± 0.2 nm for binding of [125I]OH-PIPAT to the D2dopamine receptor in control and reconstituted membranes, and 1.8 ± 0.3 and 2.2 ± 0.4 nm and 0.7 ± 0.1 and 0.7 ± 0.1 nm for binding of [125I]HPIA to the rat and human A1 adenosine receptor in control and reconstituted membranes, respectively.
Uncoupling of A1 adenosine and D2 dopamine receptors after stable reconstitution.
The reconstituted membranes in which the receptors were forced to interact with identical G protein α subunits were used to evaluate the ability of suramin and NF037 to inhibit receptor agonist binding (Fig.5). The selectivity of the two compounds toward the individual receptor/G protein tandems was essentially unchanged (compare Figs. 1 and 5). In addition, the IC50 estimates obtained for inhibition of [125I]OH-PIPAT binding to the D2 dopamine and of [125I]HPIA binding to the human A1 adenosine receptor were identical for suramin and NF037 in native and reconstituted membranes (Table1). Only after reconstitution of the rat A1 adenosine receptors with exogenous rGiα-1 complement was the inhibitory potency of both compounds moderately shifted to higher IC50values.
From the data summarized in Table 1, it is clear that suramin (and NF037) displayed the highest affinity for D2dopamine receptor/G protein complexes regardless of whether it was assessed in native or in reconstituted membranes; in addition, suramin and NF037 were more potent inhibitors of rat A1adenosine receptor/G protein tandems than those formed by the human homologue. If the site of action of suramin and NF037 is at the receptor/G protein interface, the ability to dissociate agonist binding relies on a competition between the receptor and the suramin analogue for binding to the G protein docking site. In this case, one would predict that the ability of the suramin analogues to discriminate among specific receptor/G protein tandems should be inversely correlated with the affinity of the receptors for the G protein. We therefore have assessed the ability of receptors to interact with rGiα-1 by restoring high affinity agonist binding to pertussis toxin-treated membranes. Membranes were reconstituted with increasing concentrations of rGiα-1. Because βγ dimers are required for efficient interaction of the α subunit with the receptor (Freissmuthet al., 1991b), the association of rGiα-1 with βγ dimers endogenous to the membrane may be limiting for estimating the affinity of the α subunit for the receptor. This confounding effect, however, was eliminated by combining rGiα-1 with a 4-fold molar excess of purified βγ dimers to reform the oligomer (rGiα-1.βγ) before the incubation. After detergent dilution, agonist radioligand binding was measured at a fixed concentration (see Experimental Procedures). Fig.6 shows a concentration-dependent restoration of agonist binding to the human (○) and rat A1 adenosine receptor (•) and the D2 dopamine receptor (▿). [125I]HPIA binding to membranes carrying either the human or the rat A1 adenosine receptor was restored to ≥75% of the values obtained in untreated control membranes; at the highest concentrations of Giα-1 added (300 nm), the reconstitution efficiency amounted to only ∼40% for the D2 dopamine receptor as evaluated by [125I]OH-PIPAT binding. The EC50 values for Giα-1 in the presence of βγ dimers were estimated to be 5.9 ± 1.7, 44.4 ± 9.1, and >400 nm in restoring agonist binding to the human and rat A1 adenosine and the D2 dopamine receptor, respectively. This is the inverse of the rank order of the selectivity that suramin and NF037 displayed in uncoupling the individual receptor/G protein tandems. The same difference in affinity between human and rat A1 adenosine receptor was also observed if CHAPS was used as the detergent (instead of octylglucoside) to dilute the G protein subunits; however, agonist (and antagonist) binding to D2 dopamine receptors was greatly reduced if the membranes were exposed to CHAPS.
Effect of receptor occupancy on the apparent affinity of suramin and NF037.
Suramin and NF037 do not compete for binding of antagonists to the A1 adenosine and D2 dopamine receptors, nor do they inhibit the binding of agonists in the absence of a productive interaction between receptor and G protein (Beindl et al., 1996). If the receptors were allowed to couple to G proteins, suramin inhibited agonist binding in a quasicompetitive manner (i.e., the IC50 values increased at higher concentrations of the agonist radioligand; Beindl et al., 1996; see also below). This phenomenon may result from a competition of the agonist-liganded receptor with suramin for binding to the G protein or, alternatively, from the direct action of suramin on the receptor to prevent the agonist-promoted transition of the receptor to the active conformation R*. In this case, variations in the membrane concentration of the receptor should not affect the IC50 values of suramin analogues. This was tested by using membranes from the three clones of HEK 293 cells expressing different D2dopamine receptor densities (0.3, 1.3, and 3.9 pmol/mg). The IC50 value of suramin and NF037 was determined in the presence of the agonist radioligand [125I]OH-PIPAT at a concentration close to theKD value (0.5 nm). As shown in Fig.7A for NF037, the IC50 value was shifted to the right with increasing expression levels of the receptors. The same was true for suramin and the decrease in the apparent affinity of suramin and of NF037 was related in a linear manner to the amount of bound agonist (Fig. 7B, solid symbols). Control experiments were carried out with the D2-dopaminergic antagonists sulpiride and haloperidol; as expected, receptor density did not affect IC50 values of the receptor antagonists (data not shown).
If the clone expressing intermediate levels of D2dopamine receptors was incubated with increasing concentrations of [125I]OH-PIPAT, the IC50estimates of suramin and of NF037 varied with the concentration of the radioligand (Fig. 7B, open symbols). Again, in the plot of IC50 versus receptor occupancy, the affinity estimates fall onto a straight line. The slope of the regression line is comparable within experimental error with that calculated for the IC50 values that were observed by varying receptor density (Fig. 7B, solid symbols). Hence, the number of agonist-liganded receptor present was responsible for the rightward shift of the inhibition curves (Fig. 7A) and the increase in the IC50estimates (Fig. 7B). The dependency of IC50estimates on the activator concentration is determined by the Cheng-Prusoff relation [Ki = IC50/(1 + A/KD A)]; on rearranging, the equation yields IC50 =Ki /KD A*A + Ki , stating that the IC50 value of an inhibitor depends in a linear manner on the concentration of the activator A and is determined by both the dissociation constantKD A of the activator and that of the inhibitorKi . Thus, the y-axis intercept yields an estimate of theKi (∼0.16 μm; see Fig. 7B) and the slope is given by the ratio ofKi /KD A. The slope of the regression line in Fig. 7B is ∼0.2; thus, theKD A estimate for the activator (A) is ∼0.8 μm. Obviously, because this number is calculated by a division with two derived parameters, it is inherently imprecise. However, thisKD A estimate (∼0.8 μm) is 3 orders of magnitude higher than the KD value for [125I]OH-PIPAT binding to the D2 dopamine receptor (∼0.7 nm; see Fig. 4); in contrast, theKD A estimate for the activator is consistent with the affinity estimated for the interaction between agonist-liganded D2 dopamine receptors and exogenously added rGiα-1 (Fig.6). Taken together, these findings imply that the activator (A) for which suramin and NF037 compete is not the agonist [125I]OH-PIPAT but the agonist-liganded receptor.
In the control experiments, in which haloperidol and sulpiride were allowed to compete with [125I]OH-PIPAT at radioligand concentrations covering the range 0.3–3 nm, the intercepts yielded Ki estimates of ∼0.6 and ∼5 nm for haloperidol and sulpiride, respectively (data not shown); theKD A of [125I]OH-PIPAT was estimated from these experiments to be in the range of 0.55–1.0 nm (i.e., consistent with theKD determined in saturation binding experiments; see Fig. 4).
In an analogous experiment, the occupancy of the human A1 adenosine receptor in HEK 293 membranes was varied by using [125I]HPIA concentrations covering the range of 0.15–7.5 nm. The IC50 value of suramin increased in a linear manner with receptor occupancy (Fig. 7C, •). For the purpose of comparison, the data obtained by varying occupancy of the D2 dopamine receptors with increasing [125I]OHPIPAT have been replotted as a fraction of Bmax (Fig. 7C, ▪). It is evident that the slope of the regression line determined for uncoupling the A1 adenosine receptor (∼18) was considerably steeper than that determined for inhibition of [125I]OH-PIPAT binding, whereas they-axis intercepts are similar within experimental error. Division of the y-axis intercept (0.18 μm) by the slope (∼18) yielded a value of ∼10 nm for theKD A of the activator. This is in reasonable agreement with the affinity of the agonist-liganded human A1 adenosine receptor for rGiα-1 determined in the reconstitution experiment (see Fig. 6).
Discussion
The current results unequivocally demonstrate that suramin and its didemethylated analogue NF037 discriminate among receptor/G protein tandems formed by the A1 adenosine and the D2 dopamine receptor regardless of whether the inhibition of high affinity agonist binding or of agonist-stimulated binding of [35S]GTPγS was determined. Suramin was more potent than NF037 in uncoupling A1adenosine receptors, whereas the two compounds were equipotent in preventing the interaction of the D2 dopamine receptor with G proteins. These observations complement and extend previous work that was carried out on A1adenosine receptors in human brain cortex and D2dopamine receptors in rat striatum. Here, we used both the rat and human A1 adenosine receptor and therefore rule out species differences as a trivial explanation for the distinct activity profiles of the two compounds. Because the receptors were stably expressed in the same cell line, an effect of cellular heterogeneity also can be ruled out. In agreement with this conclusion, the activity profile of the compounds was indistinguishable from the findings obtained in HEK 293 cells when the receptors were transiently expressed in COS-7 cells to obtain a nonhuman tissue readout system (not shown). Finally, the receptors may have been targeted to different subcellular compartments that possibly differed in composition of G protein subunits; when heterologously expressed in a cell line derived from a polarized epithelium, α2-adrenergic and A1 adenosine receptors are localized in the basolateral and apical membrane, respectively (Saunders et al., 1996; Wozniak and Limbird, 1996). This potential source of error was eliminated by pretreating the cells with pertussis toxin and stably reconstituting high affinity agonist binding to the membranes with a defined G protein α subunit (rGiα-1). Thus, the higher affinity of NF037 for uncoupling the D2 dopamine receptor (compared with its ability to uncouple A1 adenosine receptors) is maintained even when the receptors are forced to interact with identical G protein α subunits. Uncoupling of the D2 dopamine receptor/G protein complex by suramin analogues gave inhibition curves with varying slopes (Figs. 1C and 5C). On pertussis toxin treatment and reconstitution with Giα-1, the inhibition curves were shallower than in the control membranes. A steep slope (Hill coefficient ∼ −2) suggests interference with a reaction different from the 1:1 mode of receptor/G protein coupling, such as through the formation of receptor dimers. Dimerization of G proteincoupled receptors might result in enhanced signaling efficacy as opposed to the monomeric form of receptor (Hebert et al., 1996). On the basis of evidence obtained with other types of G protein-coupled receptors (Hebert et al., 1996; Cvejic and Devi, 1997), it is attractive to speculate that the D2 dopamine receptor in HEK 293 cells undergoes dimerization leading to steep inhibition curves with the suramin analogues. On pertussis toxin treatment of the membranes and reconstitution with Giα-1 shallow inhibition curves (Hill coefficient ∼ −1) would suggest that the ability to dimerize is lost after manipulation of the membranes. Nevertheless, the slope of the inhibition curves but not the IC50 values was independent of the fractional receptor occupancy generated in the inhibition experiments (see Fig.7A). Thus, although we have no direct evidence to explain the changes in slopes, we believe that this discrepancy does not interfere with our conclusions. Finally, we stress that experiments with purified α subunit and [3H]suramin demonstrate a binding stoichiometry of 1:1 (Hill coefficient ∼ 1.0; Hoheneggeret al., 1998).
If the pertussis toxin-treated membranes were stably reconstituted with rGiα-1, large differences in reconstitution efficiencies were observed, and the rank order was human A1 > rat A1 > human D2 receptor; the amount of G protein incorporated into the membrane was clearly not limiting because the α subunit was present in vast excess over the receptor level. We therefore hypothesized that the differences in reconstitution reflected the rank order of affinity of the individual receptors for rGiα-1. If correct, it was likewise sensible to assume that the affinities of individual receptors for the same G protein is the major determinant for the potency of suramin in differentially inhibiting receptor/G protein coupling. The difference in G protein affinity was confirmed by titrating the ability of exogenously added rGiα-1 to reconstitute high affinity binding to pertussis toxin-treated membranes. We previously determined the affinity of the human A1 adenosine receptor expressed in E. coli in reconstitution experiments with individual forms of recombinant G protein α subunits (Jockers et al., 1994); the affinity currently observed for the interaction of the human A1 adenosine receptor in pertussis toxin-treated HEK 293 membranes with rGiα-1.βγ (∼6 nm) was in reasonable agreement with that estimated in the earlier work (∼15 nm). It was, on the other hand, somewhat surprising that the affinity of the D2 dopamine receptor for rGiα-1.βγ was so low.
The interaction between the D2 dopamine receptor and rGiα-1.βγ may have been impeded by the presence of octylglucoside. We therefore exploited the Cheng-Prusoff relation to independently estimate the affinity of the human D2 dopamine receptor for its cognate G protein in HEK 293 membranes. This approach is valid if suramin competes with the agonist-liganded receptor for binding to the G protein α subunit. All available evidence supports this assumption: (1) suramin analogues bind directly to G protein α subunits (Freissmuth et al., 1996), (2) they do not affect binding of antagonists or agonists to the receptor in the absence of receptor/G protein coupling (Beindl et al., 1996), (3) the inhibition of receptor/G protein coupling can be overcome by raising the concentration of active receptor in the membrane (see Fig. 7), and (4) if the site of action of suramin is on the G protein, the regression lines in the plot of IC50 versus receptor occupancy are expected to yield similar y-axis intercepts; this was indeed observed. The affinity values estimated for the interaction between receptor and G protein were ∼0.8 μm and ∼10 nm for agonist-liganded D2 dopamine receptor and the human A1 adenosine receptor, respectively, and hence consistent with the findings in the reconstitution experiments. Based on our experimental observations, we conclude that the affinity of the individual receptors for rGiα-1 is inversely correlated to the potency of suramin in uncoupling the receptors. This conclusion predicts that a selective action of suramin on receptor/G protein coupling can also result from the difference in affinity of individual agonist-liganded receptors for the same G protein. This may also explain the earlier observation that in membranes from NG108–15 cells, suramin inhibited activation of pertussis toxin-sensitive G proteins by δ-opioid agonists but not by serum factors (Huang et al., 1990).
The structural basis for the different activity profile of suramin analogues in uncoupling A1 adenosine (suramin > NF037) and D2 dopamine receptors (NF037 = suramin) is not known. The contact points by which receptors interact with their cognate G proteins are formed by those segments of the receptor that are juxtaposed to the transmembrane spans. These discontiguous segments cooperatively support binding of the receptor to the G protein oligomer (Ernst et al., 1995;Gomeza et al., 1996) and determine the G protein specificity of the receptor (Wong et al., 1994; Liu and Wess, 1996; for review, see Gudermann et al., 1996b). The amino acids DRY (ERY in rhodopsin) at the beginning of the second intracellular loop are invariant and are required for G protein activation (Scheeret al., 1996). Apart from this triplet, only very few amino acids are conserved within the intracellular loops; hence, a clearcut consensus sequence that would allow to predict the G protein specificity of a given receptor cannot be deduced. In addition, the ability of a receptor to activate multiple G proteins is specified by distinct portions within the intracellular loops; the α2A-adrenergic receptor can couple to both Gsα and Giα. These two coupling modes, however, require distinct amino acid stretches in the second and third intracellular loops (Eason and Ligett, 1996). It is even more striking that different amino acids in the third intracellular loop are required to support coupling of the α1B-adrenergic receptor to the closely related α subunits Gαq, Gα-14, and Gα-16 (Wuet al., 1995). These findings predict that the surface the receptors cover on a given G protein α subunit varies in individual receptor/G protein tandems. This is indeed the case; if the five last amino acids in the carboxyl terminus are exchanged between Gsα and Gqα, some, but not all, receptors are capable of recruiting this mutated α subunit in a manner similar to their cognate G protein (Conklin et al., 1996). Hence, the contact sites that are formed in individual receptor/G protein tandems must be different to account for this observation. It is attractive to speculate that in the A1 receptor/Giα-1 tandem, the receptor covers a larger area of the G protein α subunit than in the D2 receptor/Giα-1tandem. This hypothesis would explain both the higher affinity of the A1 receptor for Giα-1 and the lower relative potency of NF037 in disrupting the A1 receptor/G complex; in this model, suramin, which has two additional methyl groups, competes more efficiently than NF037 with the A1 adenosine receptor for binding to the G protein, whereas the difference in surface covered by the two compounds does not affect the formation of the D2receptor/G protein complex. Taken together, our data show that two factors contribute to the selectivity of inhibitors of receptor/G protein tandem formation, namely (1) differences in affinity of individual receptors for the G protein (which determines the apparent IC50 value of an inhibitor) and (2) differences in the contact site between individual receptors and the G protein (which gives rise to a distinct structure activity relation for inhibitors). Both aspects are relevant in the development of G protein inhibitors that may eventually be useful in vivo.
Footnotes
-
Send reprint requests to: Dr. Christian Nanoff, Institute of Pharmacology, Vienna University, Währinger Str. 13a, A-1090 Vienna; Austria. E-mail:christian.nanoff{at}univie.ac.at
-
This work was supported by Grants P12750 (F.W.F.) and P12125 (M.F., C.N.) from the Austrian Science Foundation and by a concerted action “ENBST” within the EC Biomed program.
- Abbreviations:
- PACAP
- pituitary adenylyl cyclase activating polypeptide
- GTPγS
- guanosine-5′-(3-O-thio)triphosphate
- HEPES
- 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid
- XAC
- xanthine amine congener
- HEK
- human embryonic kidney
- SDS
- sodium dodecyl sulfate
- CPA
- N6-cyclopentyladenosine
- Received October 9, 1997.
- Accepted January 5, 1998.
- The American Society for Pharmacology and Experimental Therapeutics