Truncation of the A 1 Adenosine Receptor Reveals Distinct Roles of the Membrane-proximal Carboxyl Terminus in Receptor Folding and G Protein Coupling*

The carboxyl terminus (C-tail) of G protein-coupled receptors is divergent in length and structure and may represent an individualized cytoplasmic domain. By progressively truncating the A 1 adenosine receptor, a G i/o -coupled receptor with short cytoplasmic stretches, we identify two inherent functions of the C-tail, namely a role in receptor export from the endoplasmic reticulum (ER) and a role in G protein coupling. Deletion of the last 22 and 26 amino acids (of 36) reduced and completely abolished surface expression of the receptor, respectively. The severely truncated receptors were retained in the ER and failed to bind ligands. If overexpressed, even a substantial portion of the full-length receptor was retained in the ER in a form that was not functional. These data indicate that folding is rate limiting in export from the ER and that the proximal segment of the carboxyl terminus provides a docking site for the machinery involved in folding and quality control. In addition, the proximal portion is also important in G protein coupling. This latter role was unmasked when the distal portion of the C-tail (the extreme 18 amino acids, including a palmitoylated cysteine) had been removed; the resulting receptor was functional and transferred the agonist-mediated signal more efficiently than the full-length receptor. Signaling was enhanced because the coupling affinity increased (by 3-fold), which translated into a higher agonist potency. Thus, the distal immunoblots were scanned, and densitometric quantification was performed using an image analysis system (Bio-Rad). Data are presented as percent of the maximum intensity gauged from the individual concentration-response curves. Fractionation of HEK293 Cells in a Sucrose Density Gradient— Sta- ble HEK293 cells were grown to confluence on two 20-cm dishes, washed with phosphate-buffered saline, and detached by scraping. length and truncated receptor were subjected to analysis of variance (ANOVA). If the variance did not differ significantly between the data pools, potency differences were evaluated by an F-test using the method of the sum of squared residuals (33). Modeling the Conformation of the Receptor Carboxyl Terminus— To generate a putative structural model of the cytoplasmic carboxyl-ter-minal domain, HyperChem molecular visualization and simulation software (Hypercube, Inc.) was used. Based on the amino acid sequence and in analogy to the rhodopsin crystal structure, the proximal segment (HKFRVTFLKIWNDHFRC) was assigned an (cid:1) -helical structure. The peripheral segment (QPKPPIDEDLPEEKAED) was assigned a random coil structure. Ion pairing was simulated between lysine in position (cid:6) 32 (if counted from the far end) and glutamate in position (cid:6) 2 by approx- imating the charged atoms to a distance of 3Å. The energies of struc-tures were minimized in terms of bond and angle strain by subjecting the structure to geometry optimization using the AMBER force field as implemented in HyperChem 5.0. Optimization calculations proceeded until a conjugate gradient of 0.01 kcal mol (cid:6) 1 Å (cid:6) 1 was reached.

D(E)RY(W) motif and the third intracellular loop (i 3 ), in particular its amino-and carboxyl-terminal ends, are key elements. Between receptor types and even among related subtypes, sequence similarities in the cytoplasmic domains are scarce; this variability suggests that, among the cytoplasmic segments, some are reserved for receptor functions more divergent than G protein coupling.
The receptor carboxyl terminus varies considerably in length ranging from nil (in the GnRH receptor) to more than 200 amino acids (in the Ca 2ϩ /Mg 2ϩ -sensing receptor). Distinct functional properties have been assigned to individual carboxyl termini. In rhodopsin and the metabotropic glutamate receptor (subtypes IV, VI, and VII), the proximal portion represents an ancillary docking site for G␣ and/or the ␤␥-subunit (2,3). Furthermore, the C-tail 1 plays a part in receptor regulation, because it can hold a docking and/or phosphorylation site(s) for G protein-coupled receptor kinases (4). Several recent reports indicate that, in addition, the carboxyl terminus contains sequence elements required for export from the endoplasmic reticulum, membrane targeting, and endosomal sorting. Recognition sequences have been identified in the C-tail of various receptors that allow for the recruitment of PDZ domain-containing proteins, e.g. EBP-50/Na ϩ -H ϩ -exchange regulating factor (NHERF) (5), Homer (6), PICK-1 (7), and of calmodulin (8); the binding of other interaction partners, however, is less well defined (9,10). In addition, some of these reaction partners have been implicated in alternative signaling pathways, e.g. NHERF and activating transcription factor 4 (11,12); hence, the C-tail may support the direct association with effectors, bypassing the cognate G protein. This role was exemplified by comparing the signaling properties of glutamate receptor type II and type VII (13). Both subtypes are G i /G o coupled receptors and they inhibit Ca 2ϩ -currents in neurons but differ in their channel subtype specificity (P/Q-type versus N-and L-type). Although receptor activation relies on productive G protein coupling, swapping the carboxyl-terminal peptide domain transfers the specificity in channel regulation. Taken together, these findings emphasize that the C-tail of G protein-coupled receptors has evolved as an individualized functional domain.
The A 1 adenosine receptor is one of the four receptors (A 1 , A 2A , A 2B , and A 3 ) for the purine nucleoside adenosine. Adeno-sine is a primary messenger implicated in the endogenous defense against hypoxic damage and other forms of tissue stress, and the A 1 receptor is its major target in brain, heart, and kidney. The receptor couples with high affinity and fidelity to G i and G o , and most of its actions can be accounted for by G protein-mediated signaling (for a review, see Ref. 14). The A 1 receptor represents a very small, perhaps minimal, version of a class I receptor (326 amino acids; M r ϳ38,000 and, after deglycosylation, M r ϳ32,000); short loops connect the transmembrane helices, and the carboxyl-terminal domain has only 36 amino acids, including a palmitoylated cysteine residue (15). The division of the C-tail by the palmitoylated cysteine in a membrane-proximal and membrane-distal section is reflected in the primary structure. In the distal segment, the net charge is negative because of a number of acidic residues (7 of 17), whereas the proximal part has mostly positive charges and hydrophobic side chains. In the cytoplasmic extension of transmembrane helix 7, a cluster of hydrophobic residues is commonly found in many receptors. These have been implicated in the export of receptors from the endoplasmic reticulum as well as in the formation of an ancillary G protein docking site (2,8,16). Other sequence patterns, however, are not apparent in the A 1 receptor C-tail. Specifically, it has been shown that the receptor is a poor substrate for G protein-coupled receptor kinases because the C-tail lacks the appropriate phosphorylation sites. It is therefore unlikely that, upon activation, the receptor becomes phosphorylated and binds arrestin family members. In fact, the A 1 receptor is refractory to desensitization over a time frame of ϳ12 h and does not internalize (17,18). To define the role of the two parts of the C terminus, we have created a series of mutated receptors that differed by the length of the C-tail. Our data show that the segment adjacent to the seventh transmembrane helix is required for ER export. In contrast, the distal segment restrains receptor signaling. Removal of this portion increases signal transfer from the receptor to the downstream cognate G protein(s). Dulbecco's modified Eagle's medium, non-essential amino acids, ␤-mercaptoethanol, G418, and materials for cultivating bacteria were obtained from Invitrogen. Fetal calf serum was from PAA-Laboratories (Linz, Austria). L-glutamine, penicillin G, and streptomycin were from Sigma-Aldrich. Materials for polyacrylamide gel-electrophoresis were from Bio-Rad, and membranes for protein blotting were from Schleicher & Schuell. Anti-rabbit immune globulin conjugated to horseradish peroxidase was from Amersham Biosciences, and the chemiluminiscence substrate was from Pierce. A monoclonal antibody directed against c-myc was purchased from Oncogene (Cambridge, MA), and the mouse monoclonal antibody (clone 12CA5) against a peptide derived from the hemagglutinin A (HA) protein of the human influenza virus was from Roche Biochemicals. Protein G-Sepharose and protein A-Sepharose were obtained from Amersham Pharmacia Biotech. A monoclonal anti-HA antibody (16B12) was from Berkeley Antibody (Richmond, CA), and restriction enzymes were from Roche Applied Science. Rabbit antisera against the phosphorylated forms of the p42 and p44 isoforms of MAP kinase (9102) were from New England BioLabs (Beverly, MA), and anti-protein disulfide isomerase (PDI) polyclonal rabbit antibody was from StressGen (Victoria, BC, Canada). Goat allophycocyaninconjugated anti-rabbit antibodies (Alexa Fluor® 488) were from Molecular Probes (Eugene, OR), and goat anti-mouse Cy3® antibody conjugates were from Sigma-Aldrich.
Purification of plasmids from bacterial lysates was performed with the purification kit from Qiagen. Primers for PCR were obtained from genXpress (Maria Wörth, Austria); DNA polymerases were from TaKaRa Bio Europe (Gennevilliers, France).
5Ј-Cyclopentyladenosine was from Sigma-Aldrich, N 6 -(R)-phenylisopropyl-adenosine (R-PIA) was from Roche Biochemicals, and DPCPX and xanthine amine congener (XAC) were from Research Biochemicals (Natick, MA). The pEGFP-C1 vector was purchased from Clontech. Adenosine deaminase and protease inhibitors were from Roche Biosciences. HEPES and CHAPS were from Biomol (Hamburg, Germany). Tissue culture plates were from Greiner (Kremsmü nster, Austria). All other reagents were of the highest available quality.
Generation and Expression of Full-length and Truncated Versions of the Rat A 1 Adenosine Receptor-The cDNA encoding the rat A 1 adenosine receptor (inserted in the pBC-A1dhfr vector, a kind gift of Dr. Martin J. Lohse, University of Wü rzburg, Germany) was amplified by PCR. Using the forward oligonucleotide primer 5Ј-AGATGGTACCAT-GTACCCATACGATGTCCAGATTACGCTCCGCCCTACATCTCGGCC-TTCCAG-3Ј, seven modified cDNA fragments were generated that included an extended 5Ј-region coding for an N-terminally appended HA-epitope plus an upstream KpnI restriction site. The reverse primers were complementary to the 3Ј-end of the complete receptor cDNA or, alternatively, were designed to amplify truncated pieces of receptor cDNA (sequences available upon request). In addition, an ApaI restriction site was introduced at the extreme 3Ј-end of the newly generated cDNA. Annealing of the oligonucleotide primers was carried out at 65°C, amplification of the cDNA at 72°C, and the reaction proceeded for 35 rounds. Amplified cDNA fragments were then subcloned into a pcDNA3 eukaryotic expression vector using the KpnI and ApaI cloning sites. The integrity of each construct was verified by fluorescence DNA sequencing.
The conditions for cell culture, transfection, and the selection of stable cell lines were as described (20). HEK293 and COS-7 cells were maintained in Dulbecco's modified Eagle's medium at 5% CO 2 and 37°C. The culture media was supplemented with 5% fetal calf serum, 2 mM L-glutamine, ␤-mercaptoethanol, and non-essential amino acids. Cells were transfected using the calcium phosphate precipitation method. Co-transfection of pEGFP-C1, a vector carrying cDNA for green fluorescent protein, served as a control for monitoring transfection efficiency. Stable cell lines were generated that expressed the full-length receptor and two truncated forms (A 1 R⌬18 and A 1 R⌬34). Stable clones were selected for resistance to G418 (at a concentration of 0.8 mg/ml) and were screened by radioligand binding; transcripts from the A 1 R⌬34 cDNA were identified by reverse transcription PCR.
To assess the dependence of receptor export on the amount of receptor synthesized, HEK293 cells were transfected with increasing doses of receptor cDNA of A 1 R, A 1 R⌬18, and A 1 R⌬34. The transfection efficiency was kept constant by supplementing the transfection mix with irrelevant DNA (derived from a bacterial expression vector) to a final amount of 10 g per dish. Cells were harvested 48 h after transfection, and membranes were prepared.
Reverse Transcription PCR-Transfected HEK293 cells were lysed in buffer containing 4 M guanidinium thiocyanate, 50 mM Tris-HCl (pH 7.5), and 1% ␤-mercaptoethanol. RNA was extracted with phenol/chloroform followed by precipitation with 3 M sodium acetate, and the pellet was washed with 70% ethanol. RNA was recovered by centrifugation, dissolved in RNase-free water, and treated with DNase I. Total RNA was reverse-transcribed into cDNA using random oligonucleotide primers. PCR-amplification of the receptor cDNA was performed with the appropriate oligonucleotide primers. Positive clones were identified by isolating DNA of the expected size in an agarose gel. As specificity controls, the amplification of reverse-transcribed RNA was attempted with non-matching primers (oligonucleotide primers designed to amplify the full-length receptor); in addition, the reverse transcriptase was omitted prior to the PCR.
Membrane Preparation and Protein Purification-Membranes from HEK293 cells were prepared as described (26). For reconstitution experiments and determination of the receptor-mediated guanine nucleotide exchange, plasma membranes were enriched by differential centrifugation (9,000 ϫ g and 36,000 ϫ g). Recombinant, myristoylated G␣ i-1 was produced in Escherichia coli and purified from bacterial lysates as described (27). Oligomeric G proteins were purified from porcine brain membranes (28), and free ␤␥-dimers were chromatographically resolved from the ␣-subunits (29) with minor modifications (22).
Radioligand Binding Experiments-Receptor-promoted G protein activation was determined by measuring the binding of [ 35 S]GTP␥S to HEK293 membranes expressing the A1 receptor as described (26). Cell membranes (ϳ20 g) were resuspended in an 30-l volume assay buffer containing 25 mM Hepes-NaOH, (pH 7.5), 1.5 mM MgCl 2 , 100 mM NaCl, 1 mM EDTA, and 0.01 mM GDP. Following preincubation (10 min at 25°C) with the agonist N 6 -cyclopentyladenosine (CPA at the indicated concentrations), the reaction was initiated by adding [ 35 S]GTP␥S to a final concentration of ϳ10 nM (specific activity 800 cpm/fmol). After 10 min, the binding reaction was stopped, and bound and free radioactivities were separated by filtration over glass fiber filters.
Receptor agonist and antagonist binding experiments were carried out as described (26). Cell membranes were resuspended in an assay volume of 30 -50 l containing HME buffer (25 mM HEPES-Na0H (pH 7.5), 1 mM EDTA, and 2 mM MgCl 2 ) and adenosine deaminase (0.8 units/ml). The amount of membrane protein added was ϳ10 -20 g if membranes were from stable cell lines or 50 -70 g if the membranes were from transiently transfected cells. The binding was allowed to proceed to equilibrium at 25°C and was terminated after 60 -90 min by filtration over glass fiber filters using a cell harvester (Skatron, Lier, Norway). Nonspecific [ 3 H]DPCPX binding was determined in the presence of 5 M CPA, and nonspecific [ 3 H]CCPA binding was determined in the presence of xanthine amine congener (1 M), which amounted to ϳ10 and ϳ5%, respectively, of total binding in the K D concentration range.
Reconstitution of the A 1 Adenosine Receptor-Cells were pretreated with pertussis toxin (100 ng/ml for 16 h), and membranes were prepared. Membranes (ϳ 25 g/tube) were incubated with increasing concentrations of G␣ i-1 in the presence of 4 mM CHAPS on ice. For the binding assay, the detergent concentration was diluted 5-fold; the final concentrations of G␣ i-1 were as indicated in Fig. 7B. Reconstitution was monitored by radioligand binding with [ 3 H]CCPA (ϳ1 nM). The experiment was also carried out in the presence of an excess of purified G␤␥; this did not significantly alter the affinity for the A 1 receptor, presumably because of the recruitment of uncomplexed ␤␥-dimer present in the PTX-treated membranes (30).
Immunocytochemistry-Cells from stably transfected HEK293 cells were used for identification of the epitope-tagged A 1 R adenosine receptor; alternatively, HEK293 cells were transiently transfected with receptor cDNA or a control plasmid 48 -60 h prior to fixation. Cells were seeded into 10-cm plastic dishes and grown to moderate density. Growth medium was removed, and cells were washed with PBS and treated with 4% paraformaldehyde. If indicated, the cells were then permeabilized with 0.4% Triton X-100 (20 min at room temperature). For antibody staining, clusters of ϳ100 -200 cells were insulated with a crayon. Then, cells were sequentially incubated with the following reagents: 2% bovine serum albumin; primary antibody (monoclonal mouse anti-HA (16B12) antibody or a purified polyclonal rabbit anti-PDI, both at a dilution of 1:200); and secondary anti-mouse Cy3-labeled or anti-rabbit allophycocyanin-conjugated antibodies (dilution 1:1000). Following each incubation, cells were washed three times with phosphate-buffered saline followed by a final rinse with water to remove salts and embedding with Mowiol. Exposure to fluorophore-labeled antibodies was performed in the dark, after which samples were wrapped in aluminum foil to prevent bleaching. Specificity controls included staining of untransfected cells, staining of receptor-positive cells with the secondary, fluorophore-labeled antibody alone, and employment of a nonspecific fluorophore-labeled secondary antibody in addition to the primary antibody. Confocal microscopy was performed under oil immersion using a laser scan microscope (Zeiss Axiovert 135 M).
Nonspecific background staining with the anti-HA antibody was examined on cells transfected alone with pEGFP; amplification of the fluorescent signal was set to a fixed value for cells expressing the tagged receptor. Under these conditions, the pEGFP-alone transfected cells gave a blank image except when the amplification was drastically enhanced, which visualized cell contours.
Immunoprecipitation of the A 1 Adenosine Receptor-Membranes prepared from cells stably expressing the full-length and truncated A 1 adenosine receptors (A 1 R⌬18 and A 1 R⌬34) were solubilized with 10 mM CHAPS (the ratio of detergent to membrane protein was 4:1) in HME buffer containing 150 mM NaCl and protease inhibitors on ice. The membranes contained ϳ2 pmol (ϳ80 ng) of full-length and A 1 R⌬18 receptor proteins. The insoluble material was collected by centrifugation at 40,000 ϫ g for 30 min. The supernatant was diluted 2.5-fold with HME plus 100 mM NaCl. The soluble extract was incubated with anti-HA antibody (12CA5, ϳ 5 g) for 8 h at 4°C. Protein G-Sepharose was added and, after 2 h at 4°C, the material was collected by centrifugation. The pellet was washed briefly three times with HME buffer containing 4 mM CHAPS and 150 mM NaCl. The final pellet was dissolved in SDS sample buffer containing 6 M urea, and proteins were denatured for 12 h at room temperature. Samples were applied to a 12% SDS-polyacrylamide gel (containing 4 M urea). After separation and electrophoretic transfer onto nitrocellulose membranes, blots were developed with the 16B12 anti-HA antibody. The immunoreactive bands were visualized with horseradish peroxidase-linked secondary antibodies and an appropriate chemiluminiscence substrate. As a control, membranes from cells transfected with the pEGFP vector alone were subjected to immunoprecipitation.
Determination of cAMP Formation-Receptor-dependent inhibition of cAMP formation in stable HEK293 cells was assessed as described previously (20). Cells labeled with [ 3 H]adenine (2 Ci/well) were stimulated with forskolin (25 M), and receptor-dependent inhibition was determined after a 20 min incubation with CPA or R-PIA at the indicated concentrations. The assay was carried out in the presence of the phosphodiesterase inhibitor rolipram (100 M) and adenosine deaminase (0.8units/ml), which were added 30 min prior to the assay. After termination by cell lysis, [ 3 H]cAMP was isolated by sequential chromatography on Dowex AG 50W-X4 and neutral alumina. Alternatively, COS-7 cells (about 5-7.5 ϫ 10 6 cells) were transiently transfected with the appropriate A 1 adenosine receptor constructs and a plasmid encoding the human A 2A adenosine receptor. 16 h after transfection, cells were detached with EDTA and re-plated in (typically 15) 35-mm dishes. Cells were allowed to attach and were labeled with [ 3 H]adenine. Formation of cAMP was stimulated with the selective A 2A adenosine receptor agonist CGS21680 (1 M). In an individual experiment, cells expressing the intact and truncated receptor were assayed in parallel.
Determination of Receptor-mediated MAPK Phosphorylation-Stably transfected HEK293 cells were grown to 50% density on 6-cm dishes and then kept in serum-free medium for 24 h prior to the MAP kinase assay as described (31). The starving medium was supplemented with adenosine deaminase (0.8 units/ml) and replaced with fresh medium 1 h before stimulation of the cells. The level of MAPK phosphorylation was evaluated by immunoblotting with an antiserum directed against the phosphorylated p42 and p44 isoforms of MAPK. The immunoblots were scanned, and densitometric quantification was performed using an image analysis system (Bio-Rad). Data are presented as percent of the maximum intensity gauged from the individual concentration-response curves.
Fractionation of HEK293 Cells in a Sucrose Density Gradient-Stable HEK293 cells were grown to confluence on two 20-cm dishes, washed with phosphate-buffered saline, and detached by scraping. Cells were pelleted, taken up in hypotonic HME buffer including protease inhibitors (Complete protease inhibitor mixture, Roche Applied Science), and disrupted by a freeze-thaw cycle (i.e. snap freezing in liquid nitrogen followed by rapid thawing). Subsequently, membranes were further fragmented with a glass pestle homogenizer in a volume of 2.5 ml. The nuclear fraction was sedimented by low speed centrifugation (1000 ϫ g); the postnuclear supernatant containing about 2.5 mg protein was mixed with sucrose to give a concentration of 40% and layered on the bottom of a 36-ml centrifugation tube. A discontinuous sucrose gradient was generated by sequentially adding 15 2.6-ml cushions of sucrose (range, 12-38%). Membranes were brought to their isopycnic flotation density by centrifugation for 16 h at ϳ100,000 ϫ g in a Beckman ultracentrifuge. Fractions, about 1 ml each, were collected by aspiration from the top using a Haake-Buchler pump. To avoid overloading, two gradients were generated in parallel.
Fractions were screened for protein content (Bradford assay with the Bio-Rad protein dye) and radioligand binding for the distribution of the A 1 adenosine receptor and the plasma membrane marker enzyme Na ϩ -K ϩ -ATPase.  (32). Nonspecific binding was assessed in the presence of 30 M unlabeled ouabain and amounted to ϳ10% of the total binding in fractions enriched in Na ϩ -K ϩ -ATPase. [ 3 H]DPCPX binding was carried out on 100-l fraction aliquots at a final [ 3 H]DPCPX concentration of 15 nM as described above. We verified that the carryover of sucrose into the radioligand binding mix did not affect binding under the assay conditions described.
For immunoblotting, membranes were precipitated from equal fraction aliquots by adding trichloroacetic acid (final concentration, 5%). The precipitated protein was collected by centrifugation and washed three times with ice-cold acetone. After complete evaporation of the residual acetone, the proteins were dissolved in SDS-containing sample buffer. Equal aliquots were subjected to gel electrophoresis and immunoblotted using anti-PDI and anti-HA antibodies. The specificity of the anti-PDI antibody was verified in experiments with HEK293 cells before and after overnight treatment with tunicamycin, an inhibitor of N-glycosylation. Treatment with tunicamycin induces an ER-stress response and results in the induction of PDI.
Statistical Analysis-The data obtained in individual experiments were fitted to a three-or four-parameter equation, and the obtained estimates for maximal response were used to normalize the measured values. The pooled data obtained in each assay format with the full-length and truncated receptor were subjected to analysis of variance (ANOVA). If the variance did not differ significantly between the data pools, potency differences were evaluated by an F-test using the method of the sum of squared residuals (33).
Modeling the Conformation of the Receptor Carboxyl Terminus-To generate a putative structural model of the cytoplasmic carboxyl-terminal domain, HyperChem molecular visualization and simulation software (Hypercube, Inc.) was used. Based on the amino acid sequence and in analogy to the rhodopsin crystal structure, the proximal segment (HKFRVTFLKIWNDHFRC) was assigned an ␣-helical structure. The peripheral segment (QPKPPIDEDLPEEKAED) was assigned a random coil structure. Ion pairing was simulated between lysine in position Ϫ32 (if counted from the far end) and glutamate in position Ϫ2 by approximating the charged atoms to a distance of 3Å. The energies of structures were minimized in terms of bond and angle strain by subjecting the structure to geometry optimization using the AMBER force field as implemented in HyperChem 5.0. Optimization calculations proceeded until a conjugate gradient of 0.01 kcal mol Ϫ1 Å Ϫ1 was reached.

Expression of the Carboxyl-terminally Truncated A 1 Receptor-
The carboxyl-terminal tail of the A 1 adenosine receptor consists of 36 amino acids; it is divided by a palmitoylated cysteine residue at position ⌬18 into a proximal and distal part (15). Truncation of the receptor by more than 18 amino acids resulted in a deficiency in radioligand binding. Fig. 1A Fig. 1B. B max values were estimated from saturation curves obtained with the antagonist and, in addition, with a selective agonist radioligand ([ 3 H]CCPA); the ratio of agonist over antagonist binding represents the fraction of receptors that are G protein-coupled. The level of antagonist and agonist binding was significantly reduced when A 1 R⌬22 was compared with the full-length receptor or A 1 R⌬18. However, the ratio of agonist over antagonist binding was similar for the intact and both truncated receptor forms. This indicates that the truncated receptors (A 1 R⌬18 and A 1 R⌬22) are capable of G protein coupling. Although there was some variability in the determination of K D values (because of the low level binding to A 1 R⌬22), it is nevertheless clear that the decrease in binding to A 1 R⌬22 was due to a diminished B max rather than a change in K D , suggesting that truncation primarily affects the expression level but not the conformation of the ligand binding pocket.
It is generally assumed that receptor/G protein coupling is restricted to surface-located (and maybe internalized) receptors. To our knowledge, it has never been documented that receptors interact with G proteins during biosynthesis, i.e. before they are inserted into the plasma membrane; thus, A 1 R⌬18, A 1 R⌬22, and, presumably, A 1 R⌬9, must be targeted to the cell surface. Receptors that fail to bind ligand may be trapped along the biosynthetic route. A limitation to transfection experiments, particularly if one wishes to compare different constructs, is that the level of expressed protein may vary. To rule out the possibility that the severely truncated constructs suffered a deficiency in expression, we carried out the following experiments.
First, a DNA-dosing experiment was performed with the shortest receptor version that bound radioligand (A 1 R⌬22) and two longer versions (full-length and A 1 R⌬18). HEK293 cells were transiently transfected with increasing doses of expression vector supplemented with non-coding DNA such that the total amount of DNA per culture dish was constant (10 g). Fig.  1C shows the average values of specific DPCPX binding to membranes prepared from the transfected cells. Upon transfection with the full-length construct and A 1 R⌬18, DPCPX binding increased with the cDNA dose and reached high levels. The increase was not linear but leveled off, suggesting that the full-length construct and A 1 R⌬18 are folded/exported through a saturable, affinity-controlled reaction. There was little difference between these two constructs in the cDNA dose required and the maximal binding level attained. By contrast, A 1 R⌬22 gave low-level binding with no significant dose-dependent changes; the number of functional receptors could not be titrated by increasing protein expression. If, however, the binding/DNA relationship for A 1 R⌬22 was characterized by a small but steady increment in binding, this would rather suggest that, compared with the longer constructs, protein expression was inefficient.
Antibody Detection of the Intact and Truncated Forms of the Receptor-To further substantiate the assumption that, in the case of A 1 R⌬22 (and the shorter constructs), low binding levels indicated an export problem rather than inefficient expression, we raised stable HEK293 cell lines expressing the full-length receptor or one of the truncated variants and immunostained the proteins by virtue of their N-terminal epitope tag.
Immunoreactivity was visualized at the cell surface of HEK293 cells transfected with receptor cDNA encoding the full-length receptor and A 1 R⌬18 (Fig 2A, top panel). After several attempts on cells transfected with A 1 R⌬22 and A 1 R⌬34, we considered the signal too weak to reliably outline cellular structures. Cell contours could be visualized only upon drastic amplification of the signal, and this was also seen in vectortransfected control cells.
Conversely, if cells had been permeabilized with Triton X-100 before immunostaining, the truncated forms of the receptor A 1 R⌬22 and A 1 R⌬34 gave a strong signal that was located intracellularly and spared the cell nucleus ( Fig. 2A  bottom panel). This immunoreactive material was also present in cells expressing the intact receptor and A 1 R⌬18 but was absent from vector-transfected cells. In the former, intracellular staining was accompanied by membrane staining. Thus, if transcribed under the control of the cytomegalovirus (CMV) promoter, the receptor is overproduced, and the accumulated protein exceeds the capacity of folding and export by the ER and the Golgi stacks. There was no appreciable difference in the amount of immunoreactivity between cells expressing the intact and truncated receptor forms ( Fig. 2A, bottom row). In cells expressing A 1 R⌬22 and A 1 R⌬34, immunoreactivity was detected exclusively in the cell interior with no concomitant membrane staining, which suggests retention of these receptors in (an) intracellular compartment(s).
Finally, we assessed receptor expression by immunoprecipitating the shortest receptor form (A 1 R⌬34) as well as fulllength A 1 R and A 1 R⌬18, the functionally intact versions. Immunoprecipitates were collected from solubilized homogenates derived from the stable HEK293 cell lines. The precipitated samples were separated on an urea-SDS-polyacrylamide gel and blotted, and the membranes were probed with the anti-HA antibody (Fig. 2B). A single immunoreactive band, which was absent from control precipitates (vector-transfected cells), was identified in cells transfected with the full-length cDNA with A 1 R⌬18 and A 1 R⌬34. In each case, the immunoreactive band (indicated by an arrowhead in Fig. 2B) migrated to a position consistent with the apparent molecular weight of the A 1 adenosine receptor (M r ϳ38,000). This confirms that the truncated receptor, which lacks the entire C terminus (A 1 R⌬34), also becomes translated. Taken together, immunocytochemistry and Western blotting firmly indicated that the protein encoded by the A 1 R⌬34 cDNA was synthesized but retained in an intracellular compartment. The apparent quantitative differences in the immunoprecipitates in Fig. 2B did not represent the experimental average; rather, the amounts of expressed and immunoprecipitated receptor protein were similar, as would be predicted by the antibody staining of permeabilized cells (not shown). Because the receptor did not exit from the intracellular compartment, posttranslational modification of the receptor was incomplete; we believe that incomplete processing accounted for the observation that the predicted differences between the individual receptor forms (in the range of ϳ2-4 kDa) were not resolved by electrophoretic separation. The native A 1 adenosine receptor is a glycoprotein and bears complex type carbohydrate chains (34), whereas the immunoprecipitated receptor band presumably comprised a mixture of differently glycosylated receptor forms.
Localization of the Retained Receptor Proteins-We surmised that the extensively truncated receptors (A 1 R⌬26 to A 1 R⌬34) failed to pass the ER quality control and were therefore retained in the ER. This conjecture was verified by comparing the distribution of the A 1 receptor protein with that of the ERresident PDI. Permeabilized cells expressing the full-length receptor, A 1 R⌬18, and A 1 R⌬34 were incubated with the anti-HA antibody and simultaneously with a polyclonal PDI antibody. The anti-HA antibody was visualized with a fluorescein isothiocyanate-labeled secondary antibody (Fig. 3, red) and the anti-PDI antibody with a Cy3-labeled anti-rabbit IgG antibody (Fig. 3, green). The top row in Fig. 3 shows antibody staining of the three forms of the receptor. As in Fig. 2B, only the intact receptor and A 1 R⌬18 gave membrane staining, which was absent with A 1 R⌬34. With each receptor form, immunoreactivity was also detected in the cell interior, but the staining did not reveal significant quantitative differences between the constructs. The images in the middle row (Fig. 3) show the distribution of PDI. PDI immunoreactivity surrounded the nucleus, filled much of the cell interior, but spared the protrusions, thus meeting the expected localization and extension of the endoplasmic reticulum. Upon overlay of the two images (Fig. 3, bottom row), cells expressing surface located receptors (fulllength and A 1 R⌬18) revealed red membrane staining, green ER-staining and, in addition, merged yellow. This is confirmation that folding in the ER and/or export are rate-limiting; even those receptors that can be correctly folded and targeted to the surface accumulate in the ER. Immunostaining of A 1 R⌬34, however, spared the plasma membrane, and the overlay showed merged yellow with almost no fluorescent green or red. Where red patches were visible upon image overlay, there may have been other compartments than the ER that contained the receptor, presumably lysosomes; however, their identity remains uncertain.
Subcellular Fractionation of HEK293 Cells-The image overlay presented in Fig. 3 suggests that A 1 R⌬34 as well as a major proportion of the full-length receptor were retained in the ER. We tested whether the retained full-length receptor was capable of ligand binding. To separate plasma membranes from the endoplasmic reticulum, we subjected HEK293 cells stably expressing the full-length receptor to sucrose density gradient centrifugation. Plasma membranes and ER membranes were identified by the distribution of two marker enzymes, namely Na ϩ -K ϩ -ATPase and PDI, respectively. A large proportion of membrane protein was located in a single heavy peak ( ϳ 1.16 g/ml), and only smaller amounts floated to lower densities. The protein peak (Fig. 4A) coincided with the peak in PDI immunoreactivity (Fig. 4B) and was likely accounted for by rough ER microsomes, which is in accordance with similar findings reported in the literature (35,36). Ouabain binding (to plasma membrane-bound Na ϩ -K ϩ -ATPase) was distributed over a range of lighter fractions and gave two binding peaks (Fig. 4A). Some overlap of ouabain binding and PDI immunoreactivity was found that was probably due to the comigration of the smooth ER and plasma membranes in sucrose gradients, which is a common observation (37).
Radioligand binding to the A 1 adenosine receptor coincided with ouabain binding, indicating that it was found exclusively in plasma membrane-containing fractions. When the receptor was tracked by immunoblotting, two immunoreactive bands conforming to the receptor size were found exclusively in the heavy PDI-positive fractions, which were devoid of receptor binding activity and the plasma membrane marker. Thus, the full-length receptor was retained in a form that had not acquired the ability to bind ligand and was, therefore, not functional. Because the amount of receptor protein was below the detection limit, we were unable to immunoblot the HA-tagged receptor in the plasma membrane fractions. If these were combined and concentrated, the receptor protein was also detected in the plasma membrane fractions (Fig. 4D).
Two immunoreactive receptor bands that differed by ϳ3 kDa in size were detected in the ER fractions. The disparate appearance of the receptor bands most likely reflected the state of posttranslational modification of the receptor. In the process of folding, glycoproteins are subjected to several cycles of glucosylation (to an asparagine-linked N-acetylglucos-amine) and deglucosylation until the proper folding state is achieved (reviewed in Ref. 38). Hence, a straightforward explanation for the two receptor bands found in the ER fractions is that the state of glucosylation varies. This is in accordance with our conclusion (cf. above and Fig. 3) that a portion of the receptor protein failed to exit the ER because it was incompletely processed. In contrast, the complex glycosylated receptor present in the lighter fractions occupied a rather broad band (Fig. 4D).
Inhibition of cAMP Formation and Activation of the MAPK Phosphorylation Cascade in Intact Cells-Only the proximal part of the receptor is relevant for folding and export of the receptor, whereas the distal part beginning with the palmitoylated cysteine may have evolved as a regulator of receptor signaling. We have tested the effect of truncation of the distal carboxyl terminus by comparing the full-length receptor with A 1 R⌬18. First, we employed COS-7 cells for receptor expression; cells were transiently transfected with A 1 receptor plasmids (full-length or A 1 R⌬18) plus a plasmid encoding for the human A 2A adenosine receptor. The advantage of this approach is that co-transfection largely restricts stimulation (by the selective A 2A agonist CGS21680) and inhibition of cAMP formation to the same subset of cells; the increase in cAMP formation over basal was 8-fold on average (full-length, 7.7 (4.8 -12.3); A 1 R⌬18, 8.2 (4.7-14.0)) and was independent of the type of A 1 receptor coexpressed. Concentration-dependent inhibition of cAMP with the selective agonist R-PIA leveled at similar minima (42 versus 38%), but the inhibition curves revealed a higher agonist potency at the truncated receptor than that at the full-length receptor ( Fig. 5 and Table I). Because the difference between IC 50 values was small (2-3-fold) and barely reached statistical significance, inhibition of cAMP formation was re-evaluated in stable HEK293 cell lines. Similarly as in the experiments on transient transfection, the stable cells revealed a lower IC 50 value for A 1 R⌬18 than that for the full-length receptor and, in addition, the agonist effect was more pronounced with the truncated receptor ( Fig. 5B and Table I). Because the expression level was higher for the full-length (4.6 pmol/mg) than the truncated receptor (1.6 pmol/mg), the stronger inhibition could not be attributed to greater receptor number. G i -coupled receptors stimulate phosphorylation of ERK (p42/ p44 MAPK; see Ref. 39). If the increased potency of the truncated receptor were due to improved coupling, the difference would also become apparent in agonist-simulated ERK-phosphorylation. In the stable HEK293 cells, phosphorylation of ERK occurred in response to A 1 receptor activation; the time course of activation was similar for the full-length and truncated receptors, reaching a maximum after 5-10 min and declining thereafter (data not shown). There was no indication of constitutive receptor activity in the full-length or truncated receptor for DPCPX alone, which acts as an antagonist, and inverse agonist on the A 1 receptor; this did not affect basal ERK-phosphorylation over the entire time course (not shown). In the stable HEK293 cells, the difference between the receptors in the maximal increase in MAPK phosphorylation was small (fold stimulation over basal: full-length, 4.7 Ϯ 1.7; A 1 R⌬18, 6.9 Ϯ 3.4). Nonetheless, a summary of the normalized concentration-response curves documented that, in cells ex-pressing the truncated receptor, significantly lower agonist concentrations were necessary than those necessary in cells expressing the intact form ( Fig. 6 and Table I).
Receptor/G Protein Coupling-We tested whether enhanced signaling via the truncated receptor was preserved in isolated plasma membranes. G protein coupling was evaluated as follows: (i) by means of receptor agonist-promoted binding of [ 35 S]GTP␥S; and (ii) by reconstitution of the uncoupled receptor. The 2-fold approach is justified by the sequence of steps in the signal transfer reaction; it requires different contacts for stabilizing the ternary high-affinity complex (consisting of ag- FIG. 4. Differentiating between the membrane-bound A 1 adenosine and overexpressed receptor protein, which is retained in the ER. A postnuclear supernatant of the cell homogenate derived from HEK293 cells that stably express the full-length receptor was fractionated on a sucrose density (12-40%) gradient; 26 fractions were collected, and equal aliquots were assessed by radioligand binding and immunoblotting. A, charted are protein content (‚), specific radioligand binding of [ 3 H]D-PCPX to the A 1 receptor (E), specific radioligand binding of [ 3 H]ouabain to the Na ϩ -K ϩ -ATPase (q), and specific gravity (based on the calculated estimate, represented by a dashed line). B and C, distribution of the ER-marker protein PDI (B) and the receptor immunoreactivity (C) is shown. The migration position of PDI corresponded to its reported molecular mass (Mr ϳ58,000); HA immunoreactivity migrated to two positions (Mr Յ 40,000). D, immunoblotting of the A 1 receptor in a concentrated pool of fractions containing plasma membranes. onist, receptor, and G protein) and the receptor-promoted guanine nucleotide exchange reaction, respectively (20,40). In isolated membranes from the stable HEK293 cells, agonist-dependent binding of GTP␥S was determined in the presence of high concentrations of GDP (10 M) and high ionic strength (150 mM NaCl). Therefore, the resulting apparent CPA affinity was lower than that in intact cell experiments. Nonetheless, the concentration-dependent increase in GTP␥S binding revealed a higher agonist potency for the truncated than for the full-length version of the receptor (Table I).
Although agonist-dependent binding of [ 35 S]GTP␥S encompasses both reaction steps, i.e. formation of the ternary complex plus guanine nucleotide exchange, only the former is assessed by reconstituting the receptor with G␣ i-1 . For the reconstitution experiment, membranes were derived from cells that had been pre-treated with pertussis toxin (100 ng/ml), which completely eliminated high-affinity agonist binding; under these conditions, addition of the G␣ i-1 -subunit alone suffices to reconstitute high-affinity agonist binding (30). Upon incubation with G␣ i-1 and dilution of the detergent, reconstitution was assayed by agonist radioligand binding ([ 3 H]CCPA). As can be seen from Fig. 7B and Table I, the affinity differed by ϳ3-fold between the truncated and intact receptor.
As predicted by the high-affinity ternary complex model, the differences in G protein affinity translated into agonist affinity differences in binding experiments. In direct and indirect binding experiments, the agonist affinity was invariably higher for A 1 R⌬18 (data not shown). However, if agonist affinity was determined on the uncoupled receptor, that is, in the presence of GTP␥S, the competition curves were superimposable (Fig.  7C) confirming that the C terminus has no direct effect on the structure of the ligand binding pocket. DISCUSSION In contrast with other G i /G o -coupled receptors of the rhodopsin-related class, e.g. D 2 -like dopamine receptors and ␣ 2 -adrenergic receptors (19,20,21), the A 1 adenosine receptor lacks an extended third intracellular loop that can provide for accessory docking sites in receptors with a short carboxyl terminus. Nonetheless, the A 1 receptor does bind accessory proteins (e.g. a brain-coupling cofactor, the D 1 dopamine receptor hsc73; see Refs. [22][23][24]. In addition, it was shown to be delivered to specialized membrane compartments in polarized cells, e.g. the apical surface in polarized epithelia (25) and the soma of differentiated PC12 cells, 2 indicating that the intracellular portions contain recognition domains for the apical transport and retention machinery. However, for the experiments described here we have used cells (HEK293 and COS-7) that are devoid of the known intracellular interaction partners and do not support a polarized distribution. Our results show that progressive truncation of the A 1 adenosine receptor carboxyl terminus causes two independent effects, i.e. enhanced signaling and a deficiency in receptor folding/export. Deletion of more than the extreme 22 amino acids generates a polypeptide that is retained within the cell. In contrast, if truncated by only 18 amino acids, the receptor is inserted into the plasma membrane and G protein coupling is improved. This translates into an increased agonist potency in the activation of the cognate receptor-signaling pathways.
Both effects may be attributed to an altered secondary structure of the proximal carboxyl terminus; we suggest that this membrane-adjacent segment fulfills distinct functions during receptor biosynthesis and in G protein coupling but that G protein access to the proximal segment is hindered by the distal receptor C-tail. The peptide sequence of the proximal A 1 receptor C-tail (up to the palmitoylated cysteine in position ⌬18) comprises a number of regularly spaced hydrophobic residues (7 of 17), has a positive net charge, and, upon projection into a helical wheel diagram, can be predicted to form an amphipathic helix. In contrast, the peripheral 18 amino acids include a set of acidic (7 of 17) and proline (4 of 17) residues, but, otherwise, the segment is devoid of a known sequence pattern.
Folding Is Rate-limiting in ER Export-On the basis of a commonly shared sequence pattern, the proximal portion has been suggested to hold a cue for ER-export (41) and/or folding (16). The motif in question is also present in the A 1 receptor (Table II) and is destroyed by progressive truncation. It is worth pointing out that only the receptor residing in the plasma membrane was capable of ligand binding. In contrast, all receptors that were found within the ER compartment were inactive regardless of whether the full-length receptor or the truncated receptors were investigated. If a G protein-coupled receptor is properly folded, by definition the receptor transmembrane helices should adopt a circular arrangement (the 2 H. Just and H. Pankevych, unpublished observations.  6. Agonist-dependent phosphorylation of ERK 1/2 mediated by the full-length and the truncated form of the A 1 adenosine receptor. Stable HEK293 cells were serum-starved and exposed to R-PIA at the indicated concentrations for 10 min. Subsequently, cells were lysed, and equal amounts of protein were separated on a polyacrylamide gel. Immunoblots were obtained with an anti phospho-ERK antibody (top panel). The stained bands were quantified by densitometry (bottom panel). Each point represents the average intensity (ϮS.E.) from at least five experiments expressed as a percentage of maximal phosphorylation. q, full-length A 1 receptor; E, A 1 R⌬18.
Baldwin conformation). In a receptor (such as the A 1 adenosine receptor; Ref. 18) in which the binding cavity is buried between the transmembrane helices, the circular arrangement of the helices should suffice to recognize a ligand for which not even a contiguous backbone is required; functional receptors were shown to assemble from separately expressed split parts of the polypeptide chain (42). Because the receptor in the ER did not bind, we conclude that its folding is incomplete. To the best of our knowledge, it has not been determined previously whether G protein-coupled receptors that are retained in the ER are functional. In contrast, transporter proteins and ion channels can be retained in an active, hence properly folded form, a prominent example being the most common mutation of CFTR-1 (⌬F508; 43); similarly, point mutations within the second transmembrane domain cause intracellular retention of an otherwise fully active ␥-aminobutyric acid (GABA)-transporter-1 (44). Thus, retention and folding can, in principle, be separated. In keeping with this interpretation, examples can be found in the literature where fragments of a G protein-coupled receptor that comprise less than seven transmembrane helices and are devoid of functional activity locate to the cell surface, albeit with low efficiency (42,45).
Export Deficiency-It is generally held that folding is required for plasma membrane proteins to pass the ER quality control and be exported. However, individual instances reveal that this does not suffice but that additional criteria have to be met (46 -48). Based on our data, we conclude that, in the A 1 adenosine receptor, folding is the critical and rate-limiting step. Even with the full-length receptor, only a fraction of the synthesized receptor became functional, but a vast proportion of protein accumulated in the ER and was inactive. Although the proportion of retained receptors may depend on the recep-tor expression level, it is unlikely that retention of the fulllength A 1 receptor is exclusively due to overexpression. Upon programmed expression of proteins driven by an endogenous promoter, ER export may fail, and the proteins will be diverted to a degradation pathway (49); a well studied example is wildtype CFTR-1. Re-routing of wild-type CFTR can be observed in cells endogenously expressing CFTR as well as upon heterologous expression of the protein (50). It is noteworthy that the A 1 receptor surface density in the stable HEK293 cell line is comparable with that in brain where, in a fraction of cells, the density must be higher than 1 pmol/mg. The receptor abundance in brain is attributable, at least in part, to enhanced transcription (51), which has to be complemented by efficient biosynthesis and export. Because we observed saturation of export, it is likely that in HEK293 cells the export capacity is limited.
If the receptor carboxyl terminus was shortened by more than 22 amino acids, the truncated receptors (e.g. A 1 R⌬26, ⌬30, or ⌬34) were completely retained; retention was nearly complete if only the last 22 amino acids had been deleted. In contrast with the full-length receptor, increasing the dose of the expression vector for A 1 R⌬22 did not enhance its surface expression. In conjunction with the results provided by immunostaining of the receptor constructs (cf. Figs. 2 and 3), this finding indicates that retention of the severely truncated receptors cannot be due to deficient expression. Rather, our results suggest that (i) the mechanisms leading to surface targeting must differ substantially between A 1 R⌬22 and the longer constructs; and (ii) a peptide sequence relevant to export is probably contained within the proximal C-tail of the A 1 receptor.
This is in accordance with two models proposed in the liter- FIG. 7. Enhanced signaling by the A 1 adenosine receptor is due to increased affinity for the cognate G protein ␣-subunit, G␣ i-1 . A, agonist-promoted binding of [ 35 S]GTP␥S to cell membranes from stable HEK293 cells expressing the full-length (q) and truncated (E) A 1 receptor (A 1 R⌬18). The data shown represent means Ϯ S.E. from three independent experiments and are given as a percentage of the maximal agonist-dependent increment in each experiment. GTP␥S binding to cell membranes in the absence of receptor agonist amounted to 42 Ϯ 11 fmol/mg and was independent of the receptor construct used. Maximal agonist stimulation exceeded basal by 100% (full-length receptor) and 50% (A 1 R⌬18). B, reconstitution of full-length receptor (q) and A 1 R⌬18 (E) with G␣ i-1 in membranes from HEK293 cells pre-treated with pertussis toxin. A summary of the normalized reconstitution experiments wherein reconstitution was assessed by agonist high-affinity binding is shown; error bars represent S.E. The reconstitution efficiency for A 1 R and A 1 R⌬18 was 160 Ϯ 45 fmol/mg and 87 Ϯ 9 fmol/mg, respectively; nonspecific binding amounted to between 15 and 25% of the maximal binding value. C, displacement of [ 3 H]DPCPX by R-PIA from the full-length receptor (q) and A 1 R⌬18 (E) in the presence of 10 M GTP␥S. Shown are the data from a representative experiment; the slope of the displacement curves was close to unity.

TABLE II
Peptide sequence alignments of the proximal C-terminal domains of the vasopressin (V 2 ), dopamine (D 1 ), and adenosine (A 1 ) receptors Letters in italics highlight the carboxyl-terminal receptor fragments in which the export-relevant sequence patterns are contained. Letters in bold typeface indicate the sequence whose deletion results in decreased surface expression of the A 1 receptor. Numbers in the bottom row indicate amino acid position in the A 1 receptor if counted backward from the far end.

Receptor
Amino acids I  32  31  30  29  28  27  26  25  24  23  22  21  20  19  18  17  16  15  14 ature (18,41). Each one is based on an export-relevant peptide sequence pattern that was identified in the V 2 vasopressin and the D 1 dopamine receptor carboxyl terminus, respectively. For the V 2 receptor, it has been suggested that the formation of intramolecular contacts (within loop 4 and between loops 4 and 1) determines export, whereas the D 1 receptor is said to rely on an ER-associated chaperone (DriP78). In theory, the two models may be reconciled. A comparison of the peptide sequences involved (cf. the peptide sequence alignments in Table II) shows that the patterns are similar and overlap in these (and many other G protein-coupled) receptors. Both motifs are susceptible to the exchange of hydrophobic amino acids for hydrophilic ones. This suggests that they have to be compatible with an amphipathic structure and implicitly infers that the secondary structure of the peptide in which they are contained is likewise important. In the A 1 receptor the HXXXHXXHH V 2 receptor pattern may be represented at positions Ϫ32 to Ϫ24, whereas at least two alternative alignments are possible for the pattern (FXXXFXXXF) derived from the D 1 receptor because its representation is imprecise (FXXXFXXXW or FXXXWXXXF). However, the putative V 2 motif does not extend beyond position ⌬24, whereas removal of the extreme 22 amino acids (in A 1 R⌬22) dramatically impaired receptor surface expression. It is therefore more likely that the integrity of the D 1 motif and, hence, docking of the putative receptor chaperone (DriP78) is essential. Indeed, we found that the receptor binding domain of DriP78 combines with the A 1 receptor C-tail in vitro (data not shown). A requirement for (cytosolic) chaperones accounts for the observation that, by increasing receptor biosynthesis, surface expression of the fulllength receptor and A 1 R⌬18 was increased and their ER export was saturable. Thus, if the capacity of the chaperones is exceeded, a relative lack of chaperone molecules ensues, which results in receptor retention.
Enhanced Signaling-A plausible explanation for enhanced signaling by a receptor whose carboxyl terminus has been shortened is a diminished ability to desensitize (1,(52)(53)(54). However, as mentioned above, the A 1 receptor is not a substrate for G protein-coupled receptor kinases and fails to desensitize (and internalize) during an hour-long agonist exposure (18). Hence, we postulate that the increased propensity to transfer a signal, i.e. signaling at lower agonist concentrations, is a direct consequence of the structural alteration of the carboxyl terminus. We stress that, although the enhancement is only moderate (ϳ3-fold), the observed effect is not due to an artifact created by clonal selection of transfected cells. The difference was found similarly upon transient transfection and was independent of the cell line. The magnitude of the increase in agonist potency was reflected in each of the assay formats used, namely effector regulation in intact cells as well as G protein activation in isolated membranes, including reconstitution of the receptor with recombinant G␣ i-1 .
In intact cells, two signaling pathways, namely inhibition of cAMP formation and activation of the ERK signaling cascade, were evaluated. Although the former is likely mediated by a direct interaction of the inhibitory ␣-subunit with adenylyl cyclase, ERK-activation, if promoted by a G i -coupled receptor, involves both G␣ and ␤␥ (55)(56)(57). Thus, the full-length and truncated receptors activate the same set of signaling events, but the increased potency extends similarly to effector regulation mediated by the ␣ and ␤␥-subunits. This is in agreement with our conclusion that the increased affinity governing the receptor/G protein (R/G) interaction causes enhanced signaling via the truncated A 1 R⌬18 receptor. We have proposed previously that the stability of the ternary complex is the cue for efficient signal transfer; if the affinity is too low, the ternary complex forms upon receptor activation but thereafter decays too quickly for catalytic G protein activation (58). Conversely, if the affinity increases, the ability to generate active G protein species is enhanced. In extremis, a very high R/G coupling affinity even results in spontaneous activity of the receptor, or, conversely, a very low affinity may convert a partial agonist into a mere antagonist (59,60). Thus, coupling affinity may translate into both agonist potency and efficacy on a G proteincoupled receptor.
The Cytoplasmic Extension of Helix 7 Is Involved in G Protein Activation-It is likely that the proximal C-tail of the A 1 receptor folds into an ␣-helical structure, thus resembling the conformation of the carboxyl terminus in rhodopsin (61,62). In rhodopsin and other receptors, the C-tail is known to contribute a G protein docking site (2,3). Moreover, because the conformational switch associated with receptor activation (R 3 R*) presumably extends to the carboxyl terminus, a direct effect in G protein activation is likely (64,65). By analogy, the C-tail of the A 1 receptor and other receptors as well may be involved in G protein coupling. If so, the secondary structure of the proximal portion is essential, and structural changes should result in altered signaling properties. Indeed, in receptors with carboxyl-terminal sequences closely related to that of the A 1 receptor, mutations within the membrane-adjacent domain were found to enhance receptor activity (66,67). Our data show that Hydrophobic side chains are in cyan, proline in red, lysine in blue, and acidic residues in yellow. The top panel presents a view parallel to the membrane; in the bottom panel the view is directed from the cell interior toward the membrane. For the sake of clarity, only selected amino acid side chains are depicted. We ruled out that a significant intramolecular strain is imposed on this hypothetical conformation of the carboxyl-terminal domain. deletion of the distal C-tail (in A 1 R⌬18) facilitates induction (or selection) of the high-affinity receptor conformation. A straightforward explanation is that the distal C-tail impedes access of the cognate G protein. In contrast, the palmitate membrane anchor, which was eliminated in the truncated receptor (A 1 R⌬18), is not important in signaling per se, either in the A 1 receptor (15,30) or in rhodopsin (63).
By inspecting the sequence, one may assume that the C-tail makes up a ␤-turn by virtue of the proline residues (PPIP at positions Ϫ13 to Ϫ16) following the palmitoylated cysteine. Hence, the orientation of the peripheral segment is not just a straight extension of the proximal part. In Fig. 8, we propose a model of the A 1 receptor C-tail where the distal segment folds back and makes contact with the proximal portion. This is possible if the proximal segment is represented as an ␣-helix and is positioned parallel to the membrane surface with the hydrophobic residues orientated toward the membrane (Fig. 8,  top). The proline residues (Fig. 8, highlighted in red) allow for the peptide chain to turn and thus make it possible that ion bridges form between basic side chains in the proximal and acidic side chains in the distal segment. For the sake of clarity, only the last two amino acids (aspartate and glutamate, Fig. 8, yellow) and two lysine residues in the proximal segment (Fig. 8, blue) are depicted. The model suggests that a physical restraint is eliminated upon carboxyl-terminal truncation, resulting in unimpeded G protein access or an enhanced ability to undergo the conformational switch; either way promotes formation of the ternary complex.