Selective Stabilization of the High Affinity Binding Conformation of Glucagon Receptor by the Long Splice Variant of G a s *

To analyze functional differences in the interactions of the glucagon receptor (GR) with the two predominant splice variants of G a s , GR was covalently linked to the short and the long forms G a s -S and G a s -L to produce the fusion proteins GR-G a s -S and GR-G a s -L. GR-G a s -S bound glucagon with an affinity similar to that of GR, while GR-G a s -L showed a 10-fold higher affinity for glu- cagon. In the presence of GTP g S, GR-G a s -L reverted to the low affinity glucagon binding conformation. Both GR-G a s -L and GR-G a s -S were constitutively active, caus- ing elevated basal levels of cAMP even in the absence of glucagon. A mutant GR that failed to activate G s (G23D1R) was fused to G a s -L. G23D1R-G a s -L bound glucagon with high affinity, but failed to elevate cAMP levels, suggesting that the mechanisms of GR-mediated G a s -L activation and G a s -L-induced high affinity gluca- gon binding are independent. Both GR-G a s -S and GR-G a s -L bound the antagonist desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]-glucagon amide with affinities similar to GR. The antagonist displayed partial agonist activity with GR-G a s -L, but not with GR-G a s -S. Therefore, the partial ag- onist activity of the antagonist observed not did display high affinity glucagon binding. Our results using GR-G a s fusion proteins demonstrate that G a s and G a s have different effects on GR-mediated signaling, and suggest that receptors pre-coupled to G a s -L account for GTP-sensitive high affinity glucagon binding. I-Glucagon -Glycosidase Construction of the GR-G a s Fusion DNAs— The GR-G a s fusion DNAs were generated using a combination of restriction fragment replace- ment and PCR-based techniques in a modified pGEM-2 (Invitrogen) cloning vector. All nucleotide sequences were verified by fluorescence- based sequencing (Perkin Elmer/Applied Biosystems model 377A DNA sequencer). GR-G a s -L was assembled in a four-part ligation consisting of: (i) the large Avr II- Not I restriction fragment from the GR synthetic gene in pGEM-2, which deletes the last eight codons (position 478–485) and the termination codon from the C terminus of GR; (ii) a Bse RI- Nsi I restriction fragment encoding residues 20–385 of G a s -L; (iii) a 78-base pair synthetic duplex with cohesive ends for Avr II and Bse RI encoding the C-terminal residues 478–485 of GR and the N-terminal residues 2–19 of G a s -L; and (iv) a 30-base pair synthetic duplex with cohesive ends for Nsi I and Not I encoding the C-terminal 386–393 residues of G a s -L, two successive stop codons, and a Not I site at the 3 9 -end open a s -S overlap extension PCR Pfu polymerase and the GR-G a s -L construct template. A set of primers synthesized to remove codons for amino acids 72–86 and to replace Glu 71 by Asp. In the first round of PCR, two overlapping DNA sequences were amplified independently. The sense primer at the deletion site and an antisense primer from the Nsi I site of G a s -L were used in one PCR reaction, and the antisense deletion primer and an extension primer from the Sac II site of GR were used in another reaction. In a second round of PCR, the two overlapping sequences were spliced together to form a single prod-uct using the terminal extension primers from the complementary strands. The resulting PCR-amplified fragment was digested with Acc I and Bgl II and ligated into GR-G a s -L in pGEM-2. After DNA sequenc- ing, a Spe I- Not I fragment with the correct G a s -S sequence was intro-duced in place of the analogous restriction fragment in GR-G a s -L in the eukaryotic expression vector pMT (25). In GR mutant G23D1R, the i2 loop (Ser 252 –Glu 261 ) and the i3 loop (Leu 334 –Ala 349 ) were replaced by an 11-amino acid peptide derived from the dopamine D4 receptor sequence G23D1R-G a s -L, Kpn Sac II restriction from GR-G a s -L pGEM-2 by the analogous restriction fragment from G23D1R in pcDNA3 Kpn I- Not I restriction GR-G a s -L in pMT. G a s -L into pMT, a Bse RI- Not I restriction fragment s -L pGEM-2, residues of a s -L, 67-base pair Eco RI- Bse RI duplex containing codons

tions between receptors and G␣ s might be modulated by differences in structure between the isoforms as well as the stoichiometry of these proteins in cells (7)(8)(9)(10). X-ray crystallographic studies have revealed that G␣ subunits consist of a GTPase domain that resembles the oncogene protein p21 ras and a helical domain (11)(12)(13). In the structures, the guanine-nucleotide binding site lies in a cleft between these two distinct domains. A 15-amino acid insert in G␣ s -L following residue 71 is located in the first of two linkers that connect the Ras-like domain and the helical domain. Linker 1 lies near two of three "switch" regions in the G␣ structure that undergo conformational rearrangement to allow nucleotide exchange. Thus, the 15-residue insert in G␣ s -L may influence the kinetics of receptor-mediated nucleotide exchange.
Glucagon is known to exhibit both high and low binding affinities for GRs, which is thought to be due to the existence of a heterogeneous population of receptors in cells, one having a high affinity and the other a low affinity for ligand (14 -16). An expanded ternary complex model of receptor activation postulates that GPCRs exist in an equilibrium between two interconverting states that are stabilized by association with G proteins, even prior to ligand binding. The differential expression of the two forms of G␣ s in various tissues suggests the possibility for functional consequences for glucagon-dependent activation of GRs that may account for the observed differences in receptor affinities.
To test this hypothesis, we prepared fusion constructs of GR with G␣ s -S and G␣ s -L and assessed the fusion proteins for signal transduction properties in transiently transfected COS-1 cells. The abilities of glucagon and the glucagon antagonist desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide to bind to the expressed GR-G␣ s -S and GR-G␣ s -L were measured. In addition, cAMP measurements were carried out to evaluate the ability of GR-G␣ s -S and GR-G␣ s -L to couple to adenylyl cyclase. Both GR-G␣ s -S and GR-G␣ s -L caused glucagon-dependent adenylyl cyclase activation and were constitutively active. The antagonist bound to GR-G␣ s -S and GR-G␣ s -L with equal affinities, but behaved as a weak partial agonist when bound to GR-G␣ s -L but not with GR-G␣ s -S. Finally, a non-signaling GR mutant (G23D1R) in which the second (i2) and third (i3) intracellular loops were each replaced with the first (i1) intracellular loop sequence from the D4 dopamine receptor, was fused to G␣ s -L and studied. Interestingly, the G23D1R-G␣ s -L was not constitutively active, but did display high affinity glucagon binding. Our results using GR-G␣ s fusion proteins demonstrate that G␣ s -S and G␣ s -L have different effects on GR-mediated signaling, and suggest that receptors pre-coupled to G␣ s -L account for GTP-sensitive high affinity glucagon binding.

EXPERIMENTAL PROCEDURES
Materials-The cDNA for the rat G␣ s long splice variant (G␣ s -L) was kindly provided by Dr. Randall R. Reed (17). The construction of the rat GR synthetic gene (GenBank TM /EMBL Data Bank accession no. U14012) was reported previously (18). Affinity-purified anti-G␣ s antibody RM, raised against the C-terminal decapeptide RMHLRQYELL of G␣ s was kindly provided by Dr. Allen M. Spiegel (19). The preparation and characterization of the anti-GR antibody DK-12 was reported previously (20,21). The synthesis by solid phase methods and characterization of the glucagon antagonist desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide have been described (22,23). 125 I-Glucagon (507 Ci/mmol) was from NEN Life Science Products. N-Glycosidase F, endoglycosidase H, and GTP␥S were from Roche Molecular Biochemicals.
Construction of the GR-G␣ s Fusion DNAs-The GR-G␣ s fusion DNAs were generated using a combination of restriction fragment replacement and PCR-based techniques in a modified pGEM-2 (Invitrogen) cloning vector. All nucleotide sequences were verified by fluorescencebased sequencing (Perkin Elmer/Applied Biosystems model 377A DNA sequencer). GR-G␣ s -L was assembled in a four-part ligation consisting of: (i) the large AvrII-NotI restriction fragment from the GR synthetic gene in pGEM-2, which deletes the last eight codons (position 478 -485) and the termination codon from the C terminus of GR; (ii) a BseRI-NsiI restriction fragment encoding residues 20 -385 of G␣ s -L; (iii) a 78-base pair synthetic duplex with cohesive ends for AvrII and BseRI encoding the C-terminal residues 478 -485 of GR and the N-terminal residues 2-19 of G␣ s -L; and (iv) a 30-base pair synthetic duplex with cohesive ends for NsiI and NotI encoding the C-terminal 386 -393 residues of G␣ s -L, two successive stop codons, and a NotI site at the 3Ј-end of the open reading frame in the fusion. GR-G␣ s -S DNA was generated by the overlap extension PCR protocol using Pfu polymerase (24) and the GR-G␣ s -L construct as template. A set of primers was synthesized to remove codons for amino acids 72-86 and to replace Glu 71 by Asp. In the first round of PCR, two overlapping DNA sequences were amplified independently. The sense primer at the deletion site and an antisense primer from the NsiI site of G␣ s -L were used in one PCR reaction, and the antisense deletion primer and an extension primer from the SacII site of GR were used in another reaction. In a second round of PCR, the two overlapping sequences were spliced together to form a single product using the terminal extension primers from the complementary strands. The resulting PCR-amplified fragment was digested with AccI and BglII and ligated into GR-G␣ s -L in pGEM-2. After DNA sequencing, a SpeI-NotI fragment with the correct G␣ s -S sequence was introduced in place of the analogous restriction fragment in GR-G␣ s -L in the eukaryotic expression vector pMT (25). In GR mutant G23D1R, the i2 loop (Ser 252 -Glu 261 ) and the i3 loop (Leu 334 -Ala 349 ) were replaced by an 11-amino acid peptide derived from the dopamine D4 receptor sequence (26). To prepare G23D1R-G␣ s -L, a KpnI-SacII restriction fragment from GR-G␣ s -L in pGEM-2 was replaced by the analogous restriction fragment from G23D1R in pcDNA3 (26). A KpnI-NotI restriction fragment was then transferred to GR-G␣ s -L in pMT. To transfer G␣ s -L into pMT, a BseRI-NotI restriction fragment from GR-G␣ s -L in pGEM-2, corresponding to residues 20 -393 of G␣ s -L, and a 67-base pair EcoRI-BseRI synthetic duplex containing the codons for N-terminal residues 1-19 of G␣ s -L were joined with an EcoRI-NotI restriction fragment from pMT in a three-part ligation reaction.
Characterization of GR-G␣ s Fusions-GR-G␣ s fusion genes were expressed in COS-1 cells by transient transfection using LipofectAMINE (Life Technologies, Inc.). In addition, GR was independently co-expressed with G␣ s -L by transfecting COS-1 cells with equimolar ratios of vectors. Preparation of membranes from transfected COS-1 cells was carried out as described previously (18). Membrane protein concentration was determined using the Bio-Rad DC protein assay kit. N-Glycosidase F digestion and immunoblot analyses of expressed GR-G␣ s protein fusions were carried out essentially as described (18,20). Competition binding experiments with 125 I-glucagon and adenylyl cyclase activity assays to determine intracellular cAMP levels as a function of glucagon concentration were performed as previously reported (18). Competitive binding studies with 125 I-glucagon were also performed in membranes expressing GR and G␣ s fusion proteins in the absence or presence of 100 M GTP␥S. Binding and adenylyl cyclase activities were also measured in COS-1 cells that had been co-transfected with GR and G␣ s -L. Ligand binding and adenylyl cyclase studies were performed at least three times to verify reproducibility. The data were fit to a four-parameter logistic function and the IC 50 and EC 50 values calculated from the inflection point of the best-fit curve with Sigma Plot version 4 (Jandel Scientific Software).

RESULTS
Characterization of Expressed GR-G␣ s Fusions-The rat GR was fused with the short and long forms of G␣ s (G␣ s -S and G␣ s -L) to construct GR-G␣ s -S and GR-G␣ s -L. The 3Ј-terminal codon of the open reading frame of the synthetic GR gene was joined directly to codon 2 of the cDNAs encoding the two G␣ s variants. The nonsense codon of the GR gene and the initiator-Met codon of the G␣ s cDNAs were removed. G␣ s -L differs from G␣ s -S by 16 amino acids. G␣ s -L contains a Glu in place of Asp at position 71 and an insert of 15 amino acids starting with position 72 (17). The 15-amino acid insert in G␣ s -L is located within the first of two linkers connecting the Ras and helical domains in the structure of G␣ s (Fig. 1). G␣ s -L was also fused to GR mutant G23D1R, in which the i2 and i3 loops were both replaced with the i1 loop of the D4 dopamine receptor, to form G23D1R-G␣ s -L. G23D1R did not couple to cellular G␣ s as described previously (26). In addition, experiments were carried out in which GR and G␣ s -L were co-expressed from separate vectors in COS-1 cells.
Immunoblot analysis of the membrane preparations from cells expressing GR-G␣ s -L was carried out with DK-12 anti-GR antibody ( Fig. 2A) and RM anti-G␣ s antibody (Fig. 2B). The fusion protein GR-G␣ s -L immunoreacted with DK-12 to yield a band with an apparent molecular mass of 102 kDa after Nglycosidase treatment ( Fig. 2A) (20,21). This result was expected since the apparent molecular mass of G␣ s -L is 52 kDa (4), and the GR migrates as a broad band with an apparent molecular mass of 55-75 kDa that collapsed to a band at 50 kDa after N-glycosidase F treatment. GR-G␣ s -S migrated with an apparent molecular mass of 95 kDa, which is consistent with the apparent molecular mass of G␣ s -S (45 kDa) plus GR ( Fig. 2A) (4).
The RM antibody, which is directed against a C-terminal peptide of G␣ s , did not react with GR. However, as expected it detected in COS-1 cells the endogenous short and long forms of G␣ s , which migrate with apparent molecular masses of 45 and 52 kDa, respectively (Fig. 2B, lanes 1 and 2). The immunoblots consistently showed that the COS-1 cell membranes contained a greater proportion of the short form relative to the long form , respectively, were also detected by RM antibody. The membrane preparations also contained minor bands that were immunoreactive with RM antibody at 35, 40, and 60 kDa that could be degradation products derived from the fusion proteins. The fusion proteins were not recognized by anti-GR antibody ST-18, which was generated against a Cterminal peptide of GR. This suggests that the ST-18 epitope is eclipsed in the fusion or that the free C-terminal carboxylic acid group is required for immunoreactivity (data not shown) (18). Both GR and GR-G␣ s -L were insensitive to endoglycosidase H digestion, indicating that they were properly processed and transported to the cell surface (20) (Fig. 3A).
Competitive Ligand Binding Experiments-Increasing concentrations of glucagon were used to compete with 125 I-glucagon for binding to membranes from COS-1 cells containing the expressed fusion proteins. GR bound glucagon with the expected average IC 50 value of 46 nM (18,20,26) (Table I). GR-G␣ s -S displayed an affinity for glucagon that was essentially the same as non-fused GR (Fig. 4). In contrast, GR-G␣ s -L bound glucagon with an IC 50 value of 3.2 nM (Fig. 4, Table I). GR and G␣ s -L were co-expressed in COS-1 cells to determine if simply increasing the proportion of G␣ s -L to GR increased glucagon-binding affinity. The overexpression of G␣ s -L was confirmed by immunoblot analysis using RM antibody to detect an increase in the intensity of the 52-kDa band, which repre-sents total combined endogenous and recombinant G␣ s -L expression (data not shown). The expression of GR in combination with overexpression of G␣ s -L resulted in an increase in glucagon binding affinity. However, the IC 50 value obtained for the fusion protein GR-G␣ s -L (IC 50 value ϭ 3.2 nM) was still about 5-fold lower than that obtained for GR in the presence of excess G␣ s -L (IC 50 value ϭ 16 nM) ( Table I).
Mutation of the loops i2 and i3 of GR has little effect on the Membranes from COS-1 cells that had been transfected with GR, GR-G␣ s -S, and GR-G␣ s -L were treated with N-glycosidase F (N-Gase F) to remove N-linked carbohydrates. Samples (10 g of protein/lane) were separated by SDS-polyacrylamide gel electrophoresis, transferred to Immobilon-P membranes, and probed with anti-GR DK-12 antibody (A) or anti-G␣ s antibody RM (B) as described under "Experimental Procedures." Immunoreactive bands were visualized by chemiluminescence. Lanes labeled (Ϫ) and (ϩ) correspond to samples untreated and treated with N-glycosidase F, respectively. A, GR was visualized as a broad band at an apparent molecular mass of 55-75 kDa (lane 1, Ϫ). The deglycosylated GR migrated with an apparent molecular mass of about 48 kDa (lane 2, ϩ). The glycosylated (lanes 3 (Ϫ) and 5 (Ϫ)) and deglycosylated (lanes 4 (ϩ) and 6 (ϩ)) forms of GR-G␣ s -L and GR-G␣ s -S migrated with the expected molecular masses. Higher molecular weight bands appeared to be dimers of the GR or fusion proteins. Dimerization of GR was described previously (18). B, the identical sample as in panel A was probed with anti-G␣ s antibody RM. In COS-1 membranes expressing GR (lanes 1 and 2), the endogenous G␣ s splice variants, G␣ s -S and G␣ s -L were visualized at apparent molecular masses of 45 and 52 kDa, respectively. G␣ s -S was more abundant than G␣ s -L. GR-G␣ s -L and GR-G␣ s -S (lanes 3 and 5) migrated as broad bands that collapsed to single bands with apparent molecular masses of 100 and 93 kDa, respectively, after digestion with N-glycosidase F (lanes 4 and 6). Probable fusion protein dimers were visible with molecular masses above 115 kDa. The endogenous G␣ s splice variants were also visible in lanes 3-6. Other RM immunoreactive bands were presumably derived from proteolysis of the fusion proteins.   16 ]glucagon amide, showed Ͻ0.001% adenylyl cyclase activity relative to glucagon when assayed in liver membranes (22) and weak partial agonist activity (0.01% relative to glucagon) when assayed in COS-1 cells.
b The concentration of unlabeled glucagon or antagonist required to displace 50% of receptor-bound 125 I-glucagon. Values given are the means of at least three independent determinations. c Effective glucagon concentration at 50% stimulation of adenylyl cyclase. Values given are the means of at least three independent determinations. d Not determined.
ligand-binding pocket of the receptor. GR mutant G23D1R bound glucagon with an affinity similar to that of the wild-type GR (26). However, G23D1R was unable to activate endogenous G␣ s in COS-1 cells as measured by cAMP accumulation, or in HEK-293T cells as measured by calcium flux (26). Despite the inability of G23D1R to activate G␣ s , tethering G23D1R to G␣ s -L in the fusion protein G23D1R-G␣ s -L led to a 10-fold increase in affinity for glucagon. The IC 50 value for competitive displacement of labeled glucagon from G23D1R-G␣ s -L (IC 50 value ϭ 6.3 nM) was similar to that of the wild-type GR (Table I).
Competitive binding of glucagon by the fusion proteins GR-G␣ s -S and GR-G␣ s -L was also measured in the presence of GTP␥S, a non-hydrolyzable GTP analogue. The affinity of glucagon for GR and for GR-G␣ s -S was not affected by the presence of 100 M GTP␥S. This result suggests that GR-G␣ s -S displayed a single low affinity state that was independent of GTP. However, GR-G␣ s -L displayed two glucagon-binding affinities. In the presence of 100 M GTP␥S, the binding affinity of glucagon for GR-G␣ s -L shifted to an IC 50 value about 10-fold higher than that observed in the absence of GTP␥S (Fig. 5).
The binding affinity of the glucagon analogue desHis 1 -[Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide for the GR-G␣ s fusion proteins was also evaluated. The glucagon analogue binds to native GRs in rat liver membranes with an affinity similar to that of glucagon and effectively inhibits glucagon action (22). In COS-1 cell membranes expressing recombinant GRs, desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide displayed a slightly lower binding affinity (IC 50 ϭ 71 nM) than glucagon. The glucagon analogue had the same binding affinity for GR-G␣ s -S, GR-G␣ s -L, and GR (Table I). Bringing the GR and G␣ s signaling partners in close proximity in the fusion proteins did not lead to higher binding affinities for the antagonist even in the absence of GTP.
Glucagon-dependent Adenylyl Cyclase Activity-The fusion proteins were assayed for the ability to mediate an increase in intracellular cAMP upon stimulation with glucagon. The fusion proteins GR-G␣ s -S and GR-G␣ s -L caused glucagon-dependent adenylyl cyclase activation, which led to an increase in intracellular cAMP concentrations in transfected COS-1 cells. The EC 50 value of glucagon-dependent adenylyl cyclase activation in COS-1 cells expressing GR was 5.8 nM (Table I). Essentially the same EC 50 value was obtained for GR-G␣ s -L (5.1 nM). Consistent with a lower glucagon binding affinity, GR-G␣ s -S exhibited an EC 50 value that was 4 times higher than that measured for GR-G␣ s -L or non-fused GR (22.4 nM).
In addition, expression of both fusion proteins GR-G␣ s -S and GR-G␣ s -L produced elevated basal levels of cAMP in the absence of glucagon, which is characteristic of constitutive receptor signaling (Fig. 6). The constitutive activity of the fusion proteins GR-G␣ s -S and GR-G␣ s -L was not suppressed by desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide. The fact that desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide had no effect on the constitutive activity of the fusion proteins confirms that the elevated basal activity is not caused by endogenous glucagon. In the classification of antagonists, this result would mean that desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide is a neutral antagonist, not a negative antagonist or inverse agonist (27,28). DesHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide showed no measurable adenylyl cyclase activity (Ͻ0.001% relative to glucagon), when assayed in liver membrane preparations. However, in COS-1 cells expressing recombinant GRs, a small cAMP elevation, 0.01% relative to glucagon, was observed upon treatment with desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide, consistent with very weak partial agonist activity (Fig. 7). Interestingly, the glucagon analogue induced adenylyl cyclase activity above baseline only in cells expressing GR-G␣ s -L, but not in cells expressing GR-G␣ s -S (Fig. 7). This observation suggests that the partial agonist activity in intact cells induced by desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide may be attributed to receptors that are coupled to G␣ s -L, and not G␣ s -S.
Glucagon-dependent cyclase activity was also assayed in  (Table I) (Table I) and represents the average of at least three independent determinations. COS-1 cells expressing the mutant receptor-G␣ s fusion, G23D1R-G␣ s -L. G23D1R was shown to be incapable of signaling due to the replacement of its second and third intracellular loops with the first intracellular loop of the unrelated D4 dopamine receptor (26). G23D1R-G␣ s -L bound glucagon with a higher affinity than G23D1R and displayed an IC 50 value (6.3 nM) close to that of GR-G␣ s -L. However, unlike GR-G␣ s -L, G23D1R-G␣ s -L was not constitutively active in the absence of glucagon, and treatment with glucagon did not cause adenylyl cyclase activation (Fig. 6). DISCUSSION The effect of GTP to lower the affinity of certain hormones for membrane receptors was described in the pioneering work of M. Rodbell (29). The molecular basis of the "GTP effect" is now generally understood. Heterotrimeric G proteins bind to GPCRs to stabilize a high affinity agonist-binding conformation. In the presence of agonist, GTP induces the dissociation of G protein from its receptor, which then assumes a low affinity agonist-binding conformation. GRs in hepatocyte membranes display the classical GTP effect, but they are also known to exist in two interconverting ligand affinity conformations that are modulated by G s even prior to glucagon binding (30). G␣ s exists in two predominant splice variants, G␣ s -S and G␣ s -L. We aimed to test whether differences in GR coupling to these splice variants may account for the observed heterogeneity in GR ligand-binding affinity.
One useful approach to study the interactions of GPCRs and G proteins has been to join a G␣ subunit directly to the C terminus of a GPCR and express a single fusion protein signaling complex (31)(32)(33)(34)(35)(36)(37). This strategy allows unambiguous assign-ment of the functional consequences of agonist binding and subsequent receptor activation and signaling. Ideally, the fusion proteins ensure a defined 1:1 stoichiometry of GPCR to a specific G␣ subtype and enhance the efficiency and specificity of activation by constraining two predetermined signaling partners in relatively close proximity. Early attempts to find differences between the long and the short forms of G␣ s in its interaction with the ␤ 2 -adrenoreceptor (␤ 2 -AR) led to conflicting observations (38,39). However, fusion of the ␤ 2 -AR with each of the two forms of G␣ s demonstrated that the properties of either protein in the fusion did not change, and revealed that the subtle structural differences in the two variants of G␣ s had significant effects on ␤ 2 -AR-mediated signaling (33). When expressed in Sf9 cells, the ␤ 2 -AR in ␤ 2 -AR-G␣ s -L, but not in ␤ 2 -AR-G␣ s -S, possessed characteristics of a constitutively active receptor that could be inhibited by an inverse agonist of ␤ 2 -AR (33).
The GR belongs to a unique class of receptors within the GPCR family and bears little sequence resemblance to the ␤ 2 -AR, which falls within the rhodopsin-like group of receptors (1)(2)(3). We previously demonstrated that the i2 and i3 loops of the GR are essential for G protein coupling (26). Replacement of both cytosolic domains with the i1 loop sequence of the D4 dopamine receptor in the receptor mutant, G23D1R, resulted in complete loss of G protein signaling, as measured by cAMP accumulation and calcium flux assays (26). The expressed GR can mediate both cAMP elevation and calcium flux by coupling to G s (26).
Fusion of GR to G␣ s -S and to G␣ s -L allowed efficient coupling between both of the tethered partners, as demonstrated by the ability of transfected cells expressing the fusion to undergo FIG. 6. Adenylyl cyclase activity of COS-1 cells expressing GR and GR-G␣ s fusions. COS-1 cells were transiently transfected with GR, GR-G␣ s -L, GR-G␣ s -S, or G23D1R-G␣ s -L fusion genes and incubated with increasing concentrations of glucagon as described under "Experimental Procedures." Cell extracts were assayed for cAMP using a method that measures the ability of cAMP in each sample to compete with [8-3 H]cAMP for a high affinity cAMP-binding protein. The pmol of cAMP produced by 10 5 transfected COS-1 cells are plotted versus the log of peptide concentration. Each symbol represents the mean from duplicate experiments, and the curve represents a single experiment. The concentration required to elicit 50% full agonist response (EC 50 ) was determined from the inflection point of the best-fit curve derived from at least three experiments and is given in Table I. Experiments were repeated at least three times to verify reproducibility. GR-G␣ s -L and GR-G␣ s -S were constitutively active. GR and the fusion G23D1R-G␣ s -L displayed normal basal adenylyl cyclase activities. glucagon-stimulated adenylyl cyclase activity (Fig. 6). This established that the intrinsic properties of either protein were preserved within the fusion construct, and that the physical connection did not hinder or restrict the conformational changes and domain movements that must accompany the activation process in both proteins. In the case of the ␤ 2 -AR-G␣ s fusion, it was shown that downstream post-activation events such as desensitization and the repalmitoylation mechanisms were compromised, presumably because complete separation of receptor from G protein was not possible in the fusion protein (32,36). Moreover, another study reported that restricting the mobility of G␣ s relative to receptor in ␤ 2 -AR-G␣ s affected its GTPase activity (40).
The C-terminal tail of GR is relatively long compared with that of ␤ 2 -AR. We preserved the long C-terminal tail of GR in the design of the fusion, even though mutant receptors lacking most of the C-terminal extension still coupled efficiently to G␣ s (20). Since signaling requires G␣ s to contact determinants within the i2 and i3 loops of GR simultaneously, the G␣ s portion of the fusion requires enough flexibility and mobility to align its C terminus close to the cytosolic loops of GR, to dissociate after GDP/GTP exchange, and to interact with adenylyl cyclase. The long C-terminal extension presumably serves as a flexible tether to minimize steric hindrance during intraprotein contact.
Transfected non-fused wild-type GR bound glucagon with the expected average dissociation constant (IC 50 value) of 46 nM (18,20,26) (Table I). The IC 50 was significantly higher than that of GRs in liver membranes, but was typical of recombinant receptors expressed transiently in COS cells or stably in HEK-293 cells (18,41). This low affinity binding is presumably caused by inefficient coupling to G proteins, or a lack of G proteins relative to overexpressed receptors (41,42). Fig. 4 shows that the GR in GR-G␣ s -L bound glucagon with a higher affinity than non-fused GR. The IC 50 for glucagon binding shifted more than 10-fold closer to values observed for high affinity native receptors in hepatocytes (Table I). Consistent with this observation, co-expression of GR and G␣ s -L also resulted in improved ligand-binding affinity (Table I). While these results supported the assumption that an increase in the proportion of G␣ s -L to GR shifted the binding affinity, the IC 50 value measured for GR-G␣ s -L was still 5-fold lower than that obtained for co-expressed GR and G␣ s -L, suggesting that the physical proximity of the signaling partners in the fusion also contributed to the improved ligand-binding affinity. In marked contrast to GR-G␣ s -L, GR-G␣ s -S had an affinity for glucagon that was essentially the same as non-fused GR, with IC 50 values at least 10-fold higher than that measured for GR-G␣ s -L (Fig. 4, Table I).
In the classical view, the GR exists in the "off" state, R, until glucagon docks into its binding site and induces a change of conformation in R to the active form R*, which initiates the signaling cascade by interacting with G proteins. The extended ternary model of receptor activation considers that receptors are in close contact with G proteins to form a complex with some basal activity even in the absence of agonist (43)(44)(45). In the fusion, the signaling partners were constrained to precouple, and should be in the activated form R*G even prior to ligand binding. It is reasonable to assume that the conformational change in the receptor induced by interaction with G␣ s -S in GR-G␣ s -S, must be different from that induced by association with G␣ s -L in GR-G␣ s -L. Our results show that glucagon was able to discriminate between GR-G␣ s -S and GR-G␣ s -L and suggest that glucagon stabilized the active state of G␣ s -Lcoupled GRs more than G␣ s -S-coupled GRs.
The high affinity state of GPCRs is known to be sensitive to guanine nucleotides, and only agonist binding to high affinity receptors are expected to be GTP-sensitive. Upon GDP/GTP exchange, GTP-occupied G␣ s dissociates from GR, and the affinity of receptor for hormone is decreased. The binding experiments were carried out using membranes isolated from cells expressing the fusion proteins and all G␣ s domains would be expected to be either GDP-bound or nucleotide-free and primed for activation by glucagon binding to GR. As shown in Fig. 5, in the presence of 100 M GTP␥S, the displacement curve for glucagon binding to GR-G␣ s -L shifted to the right to a higher IC 50 value. In contrast, the "GTP shift" was not observed with glucagon binding to membranes expressing GR-G␣ s -S or nonfused GR (Fig. 5).
Following transient transfection in COS-1 cells, the GR in both fusion proteins GR-G␣ s -S and GR-G␣ s -L activated adenylyl cyclase in response to glucagon (Fig. 6). In addition, both fusion proteins displayed 3-fold elevations in the basal levels of cAMP over that of expressed GR in the absence of glucagon. Glucagon treatment caused a further increase in adenylyl cyclase stimulation. Certain mutations in GPCRs have been shown to increase the likelihood of agonist-independent signal transduction and are associated with specific familial diseases (46,47). A His 178 3 Arg mutation in the i1 loop of GR was also reported to cause constitutive activity (48). Constitutive activity of the fusion proteins suggests that pre-coupling of the signaling partners stabilizes the active conformation to cause GDP/GTP exchange and alter the basal activity even in the absence of agonist. In contrast, G23D1R-G␣ s -L was not constitutively active even though it displayed high affinity glucagon binding. In addition, it did not support additional cAMP production in response to glucagon.
Interestingly, glucagon bound to the fusion G23D1R-G␣ s -L with an IC 50 value similar to that obtained for GR-G␣ s -L, a 10-fold increase in affinity from that obtained for G23D1R (26). This result shows that the high affinity glucagon-binding conformation in the G23D1R-G␣ s -L complex does not require the involvement of the i2 and i3 loops of GR. This observation is striking because the G23D1R receptor itself does not display high affinity glucagon binding and does not couple to G s . The result with G23D1R-G␣ s -L suggests that the GR mutant G23D1R is analogous to rhodopsin mutants that bind, but fail to activate, the retinal G protein, transducin (G t ) (49). Certain rhodopsin mutants with alterations of their i2 and i3 loops bound G t in the GDP form, but failed to induce nucleotide exchange or G t release in the presence of GTP (49,50). Even though these mutants did not activate G t , they were stabilized in the metarhodopsin II conformation, which is analogous to the high affinity agonist-binding state of GPCRs with diffusible ligands. To our knowledge, the GR mutant G23D1R in the context of the G23D1R-G␣ s -L fusion is the first example of such a phenotype in a group II GPCR.
Extensive structure-activity analysis allowed us to identify residues in glucagon that contribute a specific structural determinant to either binding or transduction (51)(52)(53). Residues at positions 1, 9, and 16 of glucagon are essential for receptor activation, but less important for binding. Replacement of these residues led to the analogue desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide, which has been shown to be a potent competitive glucagon antagonist (22,54). In the present study, desHis 1 -[Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide displayed similar binding affinity for GR and the fusion proteins. This property is characteristic of antagonists that are unable to distinguish between different receptor agonist-affinity conformations. The fact that both fusion proteins exhibited elevated basal levels of adenylyl cyclase activation even in the presence of increasing concentrations of desHis 1 [Nle 9 ,Ala 11 ,Ala 16 ]glucagon amide confirms that