Antagonists of the Receptor-G Protein Interface Block Gi-coupled Signal Transduction*

The carboxyl terminus of heterotrimeric G protein α subunits plays an important role in receptor interaction. We demonstrate that peptides corresponding to the last 11 residues of Gαi1/2 or Gαo1 impair agonist binding to A1 adenosine receptors, whereas Gαs or Gαt peptides have no effect. Previously, by using a combinatorial library we identified a series of Gαtpeptide analogs that bind rhodopsin with high affinity (Martin, E. L., Rens-Domiano, S., Schatz, P. J., and Hamm, H. E. (1996)J. Biol. Chem. 271, 361–366). Native Gαi1/2 peptide as well as several analogs were tested for their ability to modulate agonist binding or antagonist-agonist competition using cells overexpressing human A1 adenosine receptors. Three peptide analogs decreased the K i , suggesting that they disrupt the high affinity receptor-G protein interaction and stabilize an intermediate affinity state. To study the ability of the peptides to compete with endogenous Gαiproteins and block signal transduction in a native setting, we measured activation of G protein-coupled K+ channels through A1 adenosine or γ-aminobutyric acid, type B, receptors in hippocampal CA1 pyramidal neurons. Native Gαi1/2, peptide, and certain analog peptides inhibited receptor-mediated K+ channel gating, dependent on which receptor was activated. This differential perturbation of receptor-G protein interaction suggests that receptors that act on the same G protein can be selectively disrupted.

Hormones and neurotransmitters that bind to G proteincoupled receptors control a myriad of physiological functions. Transduction of these extracellular signals involves receptormediated activation of specific G proteins by catalysis of GDP/ GTP exchange. These receptors are the target for many pharmaceutical products and are the focus of intense drug discovery efforts. Traditionally, the agonist binding site is the point of intervention, but in some cases receptor subtype-selective drugs have been difficult to achieve. Another possible target for inhibition is the receptor-G protein interface, which has been defined in some detail and involves the intracellular loops of the seven-transmembrane helix receptors with several regions on heterotrimeric G proteins (1)(2)(3). It is important to assess whether inhibitors of this interface can be found or designed and whether specific inhibition can be achieved.
The carboxyl-terminal region of the G␣ subunits represents an important site of interaction between heterotrimeric G proteins and their cognate receptors. Within this region mutations (4 -6), covalent modification by pertussis toxin-catalyzed ADPribosylation (7), or binding of specific antibodies (8) all uncouple G proteins from their receptors. In particular, the last 4 residues of the G␣ carboxyl terminus play an important role in determining the fidelity of receptor activation (9,10). Moreover, synthetic peptides from various portions of the G␣ s carboxyl terminus inhibit ␤-adrenergic receptor-G s coupling (11,12). Two of these peptides, G␣ s -(384 -394) and G␣ s -(354 -372), also stabilize the high affinity state of the receptor (12). A synthetic peptide corresponding to the last 11 residues of G␣ t , G␣ t -(340 -350), both inhibits rhodopsin G t coupling and mimics G t by stabilizing the active metarhodopsin II conformation (13). By screening a combinatorial peptide-on-plasmid library based on the carboxyl terminus of G␣ t , we previously identified numerous peptides that can also mimic the effects of heterotrimeric G protein with a much greater affinity than the native sequence by both binding to rhodopsin and stabilizing it in its active conformation, metarhodopsin II (14).
The similarity between the carboxyl terminus of G␣ t and G␣ i led us to test the G␣ t peptide analogs, which bound to rhodopsin, for their ability to bind other G i -coupled receptors. In this study we have investigated whether these peptides can 1) bind to G i -coupled A 1 adenosine receptors and induce the high affinity binding of the receptor; 2) block the ability of G i proteins to stabilize the high affinity state of agonists; or 3) inhibit signal transduction through G i by two different G i -coupled receptors, A 1 adenosine and GABA B 1 receptors. Whereas some receptors activate multiple G proteins (reviewed in Ref. 15), the A 1 adenosine receptor (16,17) and GABA B receptor (18,19) are preferentially coupled to G i /G o proteins in many cellular systems.
The extracellular nucleoside adenosine regulates a variety of metabolic functions through the activation of specific cell mem-brane receptors. Adenosine receptors, which exhibit the presumed seven transmembrane-spanning topography typical of almost all G protein-coupled receptors, are currently classified into four subtypes, A 1 , A 2a , A 2b and A 3 , based on the pharmacological profile for agonist and antagonist ligands and their effects on intracellular cAMP accumulation (reviewed in Refs. 20 -23). The A 1 adenosine receptor is widely distributed in several tissues such as heart, kidney, epididymal fat, and testis, and it is especially prominent in the central nervous system (24 -27). In the brain, the highest expression is observed in cortex, cerebellum, hippocampus, and thalamus (25,26). In the cortex it represents a primary signaling target for adenosine and is thought to tonically inhibit neuronal activity. In the hippocampus the highest density is in the dendritic region of the CA1 area (28 -30) where A 1 adenosine receptors are located at synaptic and extrasynaptic sites (31,32).
Originally, signaling through the A 1 adenosine receptor subtype was linked to inhibition of adenylyl cyclase activity in a pertussis-sensitive manner (33,34). Since then, A 1 adenosine receptors have been shown to modulate phospholipase C activity in some systems (35,36), as well as activate K ϩ currents and inhibit voltage-gated Ca 2ϩ channels (37)(38)(39)(40) through the mediation of pertussis-sensitive G proteins (G i family) (41,42). In reconstituted systems, human and bovine A 1 adenosine receptors appear to interact preferentially with recombinant G␣ i rather than G␣ o (43,44). The bovine A 1 adenosine receptor couples selectively to the G␣ i3 subunit, whereas the human receptor is able to activate each G␣ i subtype with similar potency (45). Other researchers (46), using purified bovine brain G proteins, have shown that G i2 is more potent than G i1 or G o at restoring high affinity agonist binding to bovine brain A 1 adenosine receptors. However, the ability of A 1 adenosine receptors to preferentially interact with a specific G␣ subunit does not preclude their ability to interact with other G protein subunits in an intact system.
Here, we study the effect of synthetic peptides corresponding to the last 11 residues of G␣ i1/2 , G␣ o , G␣ t , and G␣ s or the G␣ t peptide analogs on agonist binding to the A 1 adenosine receptor in rat cortical membranes or CHO-K 1 cell membranes overexpressing the human receptor. Our findings indicate that in contrast to other receptors (12,13,47), the carboxyl-terminal region of the G␣ i1/2 subunit was not capable of stabilizing the high affinity state of A 1 adenosine receptors either in rat cortex membranes or CHO-K 1 cell membranes overexpressing the human receptor. However, the native peptide G␣ i1/2 -(344 -354) as well as some G␣ t peptide analogs can negatively modulate agonist binding and compete with heterotrimeric G protein for binding to the A 1 adenosine receptor. Moreover, G␣ i carboxylterminal peptides blocked signal transduction to activation of K ϩ channels. Depending on whether the activation was through the A 1 adenosine or GABA B receptor, different G␣ t peptide analogs were most effective. Thus, it appears that certain G␣ t peptide analogs can selectively disturb the molecular interface that occurs between G i proteins and G i proteincoupled receptors.  (48). Adenosine deaminase (ADA) and (R)-N 6 -(phenylisopropyl)adenosine (R-PIA) were purchased from Sigma, and GTP␥S was obtained from Boehringer Mannheim. Other reagents were from standard commercial sources.

Materials-[
Synthetic Peptides and MBP Fusion Proteins-Peptides were purchased from Peptidogenics (Livermore, CA). The native G␣ peptides tested for their effect on agonist binding to rat cortical membranes had a free amino terminus, whereas all other peptides were synthesized with an acetyl group at the amino terminus. All peptides were purified by reversed-phase high performance liquid chromatography (HPLC), and their purity was checked by fast atom bombardment-mass spectrometry, analytical HPLC, and amino acid analysis. Plasmid pELM3 (provided by P. J. Schatz, Affymax Research Institute, Palo Alto, CA) was used for expression of maltose-binding protein (MBP) and fusion proteins as described by Martin et al. (14). The controls were the vector alone, which expressed the TGGG linker only fused to MBP (pELM 6), and pELM 17, which expressed the peptide G ␣t -(340 -350K341R) with the TGGG linker fused to MBP. All frozen cell stocks were kept in 25% glycerol at Ϫ80°C.
Preparation of Rat Cortical Membranes-The brain cortices from young male Sprague-Dawley rats (150 -200 g) that had been subjected to cervical dislocation were rapidly removed and homogenized in 10 volumes of 0.25 M sucrose prepared in 10 mM Tris-HCl, pH 7.5, 5 mM EDTA, containing protease inhibitors (0.1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, and 100 g/ml bacitracin). The membrane homogenate was centrifuged at 1,000 ϫ g for 10 min at 4°C. The supernatant was centrifuged at 46,000 ϫ g for 20 min at 4°C. The resulting pellet was resuspended in 10 volumes of 50 mM Tris-HCl, pH 7.5, 1 mM EDTA, 5 mM MgCl 2 (TEM1 buffer) containing protease inhibitors (0.1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, and 100 g/ml bacitracin), and the homogenate was centrifuged again. The pellet was resuspended in 5 volumes of TEM1 buffer containing protease inhibitors and 2 units/ml ADA and incubated at 37°C for 30 min to remove endogenous adenosine. The membrane homogenate was centrifuged, and the final pellet was stored as aliquots at Ϫ80°C until needed. Protein concentration was determined by the method of Lowry et al. (49) using bovine serum albumin as the standard.
Preparation of CHO-K 1 Cell Membranes-Confluent CHO-K 1 cells were lysed with hypotonic buffer (10 mM HEPES, pH 7.4, 10 mM EDTA) and scraped from the plate. The lysates were homogenized, centrifuged, and washed twice in washing buffer (10 mM HEPES, pH 7.4, 5 mM EDTA). The homogenized lysates were then stored at Ϫ80°C in storage buffer (10 mM HEPES, pH 7.4, 1 mM EDTA, 10% (w/v) sucrose) until needed. Protein concentration was measured using the Coomassie Blue binding method (50) (Bio-Rad) with bovine serum albumin as the standard.
CHO-K 1 Binding Assays-Membranes (10 g/ml proteins) from CHO-K 1 cells overexpressing the human A 1 adenosine receptor were added to tubes containing HEM buffer (50 mM HEPES, pH 7.5, 1 mM EDTA, 5 mM MgCl 2 ), 2.5 units/ml ADA, [ 125 I]ABA (0.5 nM) and either the MBP fusion proteins (50 M) or synthetic peptide (100 M). The G␣ t peptide analogs were dissolved in HEM buffer, and their solubility was checked as described by Rarick et al. (51). The reaction was allowed to proceed at 30°C for 2 h. The bound and free radioligands were separated by filtration through Whatman GF/C filter paper (soaked in 0.3% polyethyleneimine) using a Brandel tissue harvester. Filters were washed twice with ice-cold TEM2 buffer (20 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 , 0.5 mM EDTA). Binding assays were performed in duplicate, and nonspecific binding was determined by adding 15 M R-PIA at the same time as the radioligand to some samples. Specific binding was 85-90% of total binding for all experiments.
For the competition assays, CHO-K 1 cell membranes (10 g/ml proteins) were added to tubes containing HEM buffer, 2.5 units/ml ADA, 0.8 nM [ 3 H]DPCPX, and increasing concentrations of R-PIA (0 to 10 M) in the presence and absence of peptide analogs (100 M). The reaction was allowed to proceed at 25°C for 2 h. Then the reaction mixtures were filtered through Whatman GF/C filter paper using a Brandel tissue harvester. Filters were washed twice with ice-cold TEM2 buffer. Binding assays were performed in duplicate, and nonspecific binding was determined by adding 15 M R-PIA at the same time as the radioligand. Specific binding was 85-90% of total binding for all experiments.
Dilution and saturation experiments using either the agonist, Whole cell electrophysiological recordings were made using the "blind" patch recording technique (52). Briefly, the patch pipette was attached to a hydraulic microdrive and advanced into the stratum pyramidal layer of the CA1 region of the hippocampus in 2-m steps. Positive pressure was applied while the electrode was advanced through the slice, and once contact between the electrode and a cell body was achieved (indicated by a small increase in the electrode resistance), suction was then applied to the electrode to form a gigaohm seal with the membrane. Further suction was used to rupture the membrane patch, providing low resistance access between the electrode filling solution and the cytoplasm of the cell. Neurons were voltageclamped at Ϫ65 mV immediately upon rupturing the membrane using the continuous single electrode voltage clamp mode of an Axoclamp-2A amplifier (Axon Instruments, Burlingame, CA). The average membrane potential of the cells was approximately Ϫ70 mV after correction for the liquid junction potential.
To ensure that the peptides had adequate time to diffuse into the cells, experiments were begun a minimum of 15 min following patch rupture. During the experiment, the holding current necessary to maintain the voltage clamp was sampled every 15 s, and membrane resistance was determined every 30 s by the current response to a Ϫ4-mV voltage command step. The access resistance between the electrode and the cytoplasm of the cell was continually monitored by observation of the cell membrane capacitive currents in response to a brief voltage step and was below 30 M⍀ in all experiments. All responses were digitized with an R.C. Electronics ISC-16 analog-to-digital card and analyzed by computer with software developed in our laboratory. Bath-applied drugs were delivered into the superfusion line by a syringe pump (Razel Scientific Instruments, Inc., Stamford, CN) from stock solutions that were 100 ϫ the final concentration of drug. Nearly all neurons were tested with both adenosine and baclofen, and when possible, cells were retested with either one or both drugs (e.g. Fig. 3C).
Data Analysis-A nonlinear multipurpose curve-fitting computer program (GraphPad Prism, version 2.0, GraphPad Software, San Diego, CA) was used for analysis of saturation and competition binding data.
A partial F test evaluated whether the data were best fit by a one-or two-site model. The IC 50 values calculated from the competition curves were converted to K i values by the Cheng and Prusoff equation (53). The dose-response curves for the G␣ peptides were fit using nonlinear regression analysis and IC 50 values were derived (GraphPad Prism, version 2.0). Data are presented as mean Ϯ S.E. of at least three experiments, unless otherwise noted. The statistical differences were determined using the unpaired t test (GraphPad Prism, version 2.0). Table I shows the amino acid sequences of all peptides used for the experiments and indicates the EC 50 values of their ability to stabilize rhodopsin in its active conformation, metarhodopsin II. Synthetic or MBP-fused peptides were tested for their ability to modulate agonist binding to A 1 adenosine receptors from two different species, rat and human. The human receptor was expressed in stably transfected cells, whereas the rat receptor was in its native cellular background.  (Fig. 1). In contrast, neither G␣ s -(384 -394) nor G␣ t -(340 -350) inhibited agonist binding to A 1 adenosine receptors at concentrations of 200 M (Fig. 1). Therefore, under these conditions, both G␣ i1/2 and G␣ o1 peptides appear to compete with heterotrimeric G proteins for interaction with the A 1 TABLE I Amino acid sequences of native G␣ peptides and analogs and their ability to stabilize metarhodopsin II Peptides were synthesized and purified as described under "Experimental Procedures." Their purity was checked by fast atom bombardment-mass spectrometry, analytical HPLC, and amino acid analysis. Native G␣ peptides tested for their effect on agonist binding to rat cortical membranes had a free amino terminus, whereas all other peptides were synthesized with an acetyl group at the amino terminus. Metarhodopsin II stabilization was measured as described by Martin et al. (14), and the EC 50 values were derived by fitting the dose-response curves with a nonlinear regression program (GraphPad Prism, version 2.0). ND, not determined. Bold letters indicate those amino acid residues that differ with respect to those in the G␣ i1/2 carboxyl-terminal sequence. Underlined letters indicate amino acid residues that are conserved change with respect to those in the G␣ i1/2 carboxyl-terminal sequence. adenosine receptors in rat cortical membranes, but they are unable to stabilize high affinity agonist binding.

Effects of Native G␣ Carboxyl-terminal Peptides on Agonist Binding to Rat A 1 Adenosine Receptors-The
Effects of G␣ t Peptide Analogs from the Combinatorial Library-By using a combinatorial approach, we previously identified a series of G␣ t peptide analogs which can both bind to rhodopsin and stabilize the receptor in its active conformation, metarhodopsin II, with higher affinity than the native G␣ t carboxyl-terminal peptide (14). There is a high degree of homology between G␣ t and G␣ i carboxyl-terminal regions, with only one amino acid difference between G␣ t and G␣ i1/2 (Table I) and two amino acid differences between G␣ t and G␣ i3 . To determine whether these peptides could stabilize the high affinity agonist binding state of G i -coupled receptors, we tested MBP-fused peptides or synthetic peptide analogs in agonist binding assays of human A 1 adenosine receptors overexpressed in CHO-K 1 cells. As reported under "Experimental Procedures," approximately 50% of A 1 adenosine receptors in CHO-K 1 cell membranes are coupled to G proteins. Therefore, if any of the MBP fusion proteins or peptide analogs were able to mimic the heterotrimeric G protein, we should detect an increase of agonist binding. None of the MBP fusion proteins or peptide analogs tested resulted in a significant increase of specific [ 125 I]ABA binding to CHO-K 1 cell membranes expressing human A 1 adenosine receptors (data not shown). However, a few of the analogs either as MBP fusion proteins or synthetic peptides inhibited agonist binding to CHO-K 1 cell membranes. MBP fusion proteins 19 and 24 (50 M) inhibited [ 125 I]ABAspecific binding by 24 and 23%, respectively. The corresponding synthetic peptides (100 M) resulted in 21 and 33% decrease of specific binding, respectively.
Since we could not detect a significant effect of either MBP fusion proteins or peptide analogs on direct agonist binding to human A 1 adenosine receptors, an alternative approach that allows identification of the two affinity states of the receptor was used to detect variations of its agonist affinity states. Inhibition of radiolabeled antagonist binding, [ 3 H]DPCPX, by the full agonist R-PIA in the presence and absence of native peptide G␣ i1/2 -(344 -354) or G␣ t analogs was studied. We focused our interest on the activity of six peptide analogs (peptides 8, 9, 15, 19, 23, and 24) which demonstrated different abilities to stabilize metarhodopsin II (Table I). All competition curves showed biphasic patterns. Analysis of data using a nonlinear curve-fitting program revealed that they were better represented by a two-site rather than a one-site model, indi-cating the presence of two affinity states of the receptor. Fig. 2 shows the competition binding isotherms obtained in the absence (control) or presence of peptide G␣ i1/2 -(344 -354) as well as the G␣ t peptide analogs 15 or 19. The native peptide G␣ i1/ 2-(344 -354) did not induce any significant modification of the displacement curve (Table II). This result indicates that under these conditions the native peptide G␣ i1/2 -(344 -354) is unable to mimic the effect of heterotrimeric G i proteins and thus stabilize the high affinity state of the receptor. Both peptides 15 and 19 demonstrated significant right-shifts of the displacement curves (Fig. 2). The percentage of receptors in the high affinity state decreased significantly in the presence of peptide 15 compared with control (47% to 33%; p Ͻ 0.05). These findings indicate that peptide 15 partially disrupts the high affinity interaction between A 1 adenosine receptors and G i proteins. Interestingly, in the presence of peptide 15, the K i values for the high and low affinity states of the receptor showed a 13and 5-fold increase, respectively (Table II). Both values were significantly different from control K i values (p Ͻ 0.0001 and p Ͻ 0.01, respectively). This suggests that this peptide is able to modulate the high and low affinity states of the receptor. In the presence of peptide 19, the K i values for the high affinity and low affinity states of the receptor increased 9-(p Ͻ 0.0001) and 2-fold, respectively, whereas the percentage of the high affinity receptors decreased by approximately 10% (Table II). Thus, peptide 19 also appears to disrupt the high affinity interaction between A 1 adenosine receptors and G i proteins, but it modulates only the high affinity state of the receptor. Another G␣ t peptide analog, peptide 8, also affected the K i values for both the high and low affinity state of the A 1 adenosine receptor. In the presence of this peptide, the K i values for the high and low affinity state of the receptor increased 10-and 5-fold compared with control values (p Ͻ 0.0001 and p Ͻ 0.005, respectively), whereas the percentage of receptors in the high affinity state did not change (Table II). Therefore, peptide 8 has the ability to modulate both the high and low affinity state of the A 1 adenosine receptor, but it does not change the relative distribution of these two affinity states. All other peptide analogs that were tested had no significant effects on binding parameters (Table II).
Functional Effects of G␣ i -and G␣ t -related Synthetic Peptides on Inhibition of Signal Transduction-To study the ability of these peptides to compete with endogenous G proteins in the native setting and block signaling through G i , we introduced the peptides into hippocampal neurons through a patch pipette and subsequently determined the extent to which G proteincoupled inwardly rectifying K ϩ channels (GIRKs) could be activated by either A 1 or GABA B receptors. Superfusion of hippocampal brain slices with 100 M adenosine or 50 M baclofen elicit outward currents in CA1 pyramidal neurons, a reflection of the activation of GIRK by A 1 adenosine or GABA B receptors, respectively. This effect is mediated via a pertussis toxin-sensitive G protein (54). An example of recording under standard conditions is shown in Fig. 3A. The maximal outward current induced by adenosine was 50 Ϯ 5.5 pA (n ϭ 38 cells). Internal dialysis of these neurons with the carboxyl-terminal G␣ 1/2 peptide completely eliminated the adenosine response ( Fig. 4; p Ͻ 0.0001 versus control). Synthetic peptides 19 and 24 (Figs. 3C and 4) were also able to completely block the response to adenosine (p Ͻ 0.0001 and p Ͻ 0.002 versus control, respectively), whereas other peptides (8, 9, and 15) appeared to partially block the adenosine response, although these effects were not statistically significant. Peptide 23 had no effect on the response to adenosine (Fig. 3B). Thus, the native G␣ i1/2 peptide as well as peptides 19 and 24 are effective inhibitors of G protein-coupled signal transduction through A 1 adenosine receptors.
Specificity of Functional Blockade of G i -coupled Receptors by G␣ Peptides and Analogs-One would expect that a peptide corresponding to the carboxyl terminus of G␣ i would block signaling through all G i -coupled receptors. It is of great interest to determine whether there can be any selectivity at the receptor-G protein interface. To evaluate whether the G␣ t peptide analogs show a pattern of specificity for different G i -coupled receptors, we measured the effect of these peptides on GABA B receptor mediated activation of GIRK in CA1 pyramidal neurons, an effect which is also mediated via a pertussis toxin-sensitive G protein (55). Superfusion of hippocampal brain slices with 50 M baclofen elicited large outward currents (Fig. 3A) with an average maximal response of 83 Ϯ 8.8 pA (n ϭ 29). Internal dialysis of CA1 pyramidal neurons with carboxylterminal G␣ peptides either had no effect upon current response to baclofen (peptide 23; Fig. 3B) or reduced the baclofen response (the native G␣ i1/2 , peptides 8,9,15,19, and 24 (Fig.  4)). No synthetic peptide completely blocked baclofen-induced activation of GIRK. However, peptides 8, 15, and 24 produced significant reduction of the current response compared with control ( p Ͻ 0.005, p Ͻ 0.007, and p Ͻ 0.05).

FIG. 3. Effects of intracellular dialysis with G␣ t peptide analogs on current responses to adenosine and baclofen.
Each panel shows the holding current required to clamp a cell to Ϫ65 mV during superfusion with 100 M adenosine and 50 M baclofen (indicated by horizontal bars below the records). In control cells, adenosine responses were usually smaller than the responses to baclofen (A). Dialysis with peptide 23 (B) did not appear to have any significant effect on the adenosine or baclofen responses (in this cell the responses were, if anything, larger than the control responses in A, whereas dialysis with peptide 24 markedly attenuated the adenosine response relative to the baclofen response C). Each panel represents data from a different cell. In all cases, patch rupture occurred at least 15 min prior to the beginning of adenosine superfusion. In cases such as that illustrated in C, where adenosine and baclofen could be tested repeatedly, no significant differences were observed between initial drug tests and subsequent responses, suggesting that the peptide effects were near maximal within the 15-min waiting period following patch rupture. the adenosine response was virtually abolished, whereas the baclofen response was only reduced, and in each of these cases the effect of the peptides on adenosine responses was significantly greater than the effect on the GABA B response (p Ͻ 0.05 in each case). These differences were readily observed in individual cells (e.g. Fig. 3C); thus, the failure to antagonize the responses to baclofen could not be attributed to inadequate dialysis of the cells with the peptide, because the adenosine responses in many such cases were completely blocked. The overall results are shown in Fig. 4, in which the holding current for each condition is expressed as percentage of control. A similar pattern of effects on changes in the presence of peptides was observed for whole cell conductance elicited by either adenosine or baclofen, although the changes in conductance appeared to be somewhat more variable than the changes in holding current (data not shown). Clearly, the ability of some peptides to inhibit receptor-mediated K ϩ channel gating was different depending on which receptor was activated. This finding suggests that in the same rat hippocampal neurons A 1 adenosine and GABA B receptor G i protein interaction can be differentially disrupted by peptide analogs of the carboxyl terminus of G␣ i . DISCUSSION Recently, a variety of studies have focused on finding new agents that selectively uncouple receptors from G proteins (14,56,57) and thus disrupt cellular responses. The carboxyl-terminal region of G␣ subunits provides the molecular basis for receptor-mediated activation of G proteins and plays a crucial role in determining the fidelity of this activation (13,47). Synthetic peptides corresponding to the last 11 residues of the G␣ t and G␣ s subunit are able to mimic the conformational effects of heterotrimeric G proteins on their cognate receptors, rhodopsin and ␤-adrenergic receptors, by stabilizing their active conformation, although with low potency (12,13,47). By using a random "peptides-on-plasmids" library approach (14), we identified several analogs of the native peptide G␣ t -(340 -350) that bind with high affinity to rhodopsin and stabilize its active form, metarhodopsin II. As the carboxyl-terminal sequences of G␣ i1/2 and G␣ i3 diverge just by one and two amino acids from G␣ t carboxyl-terminal sequence, respectively, we evaluated the effects of native G␣ and analog peptides on receptor coupling to G i proteins.
No G␣ Carboxyl-terminal Peptides Stabilize the High Affinity State of Rat A 1 Adenosine Receptors-The effects of synthetic peptides G␣ i1/2 -(344 -354), G␣ o1 -(344 -354), G␣ t -(340 -350), and G␣ s -(384 -394) on agonist binding to A 1 adenosine receptors were studied using rat cortical membranes. None of the native peptide sequences increased agonist binding, implying that they are unable to stabilize the high affinity state of the receptor. On the contrary, peptide G␣ i1/2 -(344 -354) and G␣ o1 -(344 -354) inhibited specific binding in a dose-dependent fashion. The result suggests that synthetic peptides corresponding to G␣ i1/2 and G␣ o1 carboxyl-terminal sequence disrupt the interaction between A 1 adenosine receptors and G i proteins. Since in our assay conditions, agonist binding was also sensitive to inhibition by GTP␥S, indicating that most receptors were effectively interacting with G proteins, probably these peptides compete with G i /G o for binding to the receptor. Such an outcome implies that peptide G␣ i1/2 -(344 -354) and G␣ o1 -(344 -354) are not able to mimic heterotrimeric G protein in stabilizing the high affinity state of the rat A 1 adenosine receptor. The inability of peptide G␣ t -(340 -350) to inhibit agonist binding was also quite surprising since there is only one amino acid difference between it and peptide G␣ i1/2 -(344 -354). This difference may be due to G␣ t peptide having decreased affinity for the rat A 1 adenosine receptor, which would be critical for its competition with G i /G o proteins.
No G␣ t Peptide Analogs Stabilize the High Affinity State of Human A 1 Adenosine Receptors-These peptide analogs were selected for their ability to bind with high affinity to metarhodopsin II. All analogs stabilized metarhodopsin II with higher affinity than the native G␣ t peptide (14). It is possible that if another receptor is used to screen the combinatorial library, high affinity peptides that selectively bind and stabilize this receptor in its active state might be found. The structural basis of this idea is that receptors have different amino acid sequences and thus perhaps some differences in structure in their G protein binding region(s). The combinatorial approach should be able to find such differences. These considerations motivated us to examine the effects of G␣ t -(340 -350) analogs on agonist binding to human A 1 adenosine receptors overexpressed in CHO-K 1 cell membranes. Under our cell culture and binding assay conditions, approximately 50% of total receptors appeared to be in the high affinity state and thus effectively coupled to G proteins. Therefore, peptides did not need to compete with the heterotrimeric G proteins for binding to the receptor and modulating agonist affinity. However, none of the G␣ t -(340 -350) peptide analogs either as MBP fusion proteins or synthetic peptides were able to increase agonist binding to human A 1 adenosine receptors, suggesting they were unable to mimic the effects of heterotrimeric G i /G o proteins.
To confirm this finding, a delineation of receptor affinity states was carried out by competing the radiolabeled antagonist bound, [ 3 H]DPCPX, with an agonist (R-PIA) in the presence and absence of the native peptide G␣ i1/2 -(344 -354) or G␣ t -(340 -350) analogs. High and low affinity states of the receptor are detectable in this assay condition. None of the synthetic peptides resulted in a significant increase of the receptor number in the high affinity state confirming that they are not able to mimic the allosteric effect of heterotrimeric G proteins on receptor conformation.
Modulation of Agonist Affinity by G␣ t Peptide Analogs-Although G␣ peptides could not mimic heterotrimeric G proteins and stabilize the high affinity state of the A 1 adenosine receptor, we found evidence that the peptides bind to the receptors and compete with heterotrimeric G proteins, as indi-  8,9,15,19,23,24), and the asterisks indicate points that are significantly different from the corresponding control responses (p Ͻ .05). Bars denoted with the symbols are those where the peptide inhibited the adenosine response to a significantly greater extent than the baclofen response (p Ͻ 0.05). cated by a decrease of agonist high affinity binding. Interestingly, three G␣ t -(340 -350) analogs, peptides 8, 15 and 19, increased the K i value of the high affinity state of the receptor, indicating that these peptides not only disrupt the high affinity interaction between agonist-activated receptors and G i proteins, but they also stabilize the receptor at an intermediate affinity state. Thus, these peptides appear to be affecting the receptor conformation in a subtle way. One might speculate that there is a continuum of conformations between R and R*, and these peptides may stabilize an intermediate conformation. The existence of multiple distinct active receptor states differing in their G protein-coupling abilities has been suggested for both rhodopsin-(58 -61) and the human thyroidstimulating hormone receptor (15).
The K i value for the receptor in the low affinity state increased significantly in the presence of peptide 8 and 15. One possibility is that the peptides bind to uncoupled receptors in the low affinity state and determine conformational changes leading to a further decrease of receptor affinity for the agonist. However, the significance of this variation must be interpreted with caution since a filtration binding assay is not very suitable for studying low affinity binding sites. That the native peptide G␣ i1/2 -(344 -354) and the other G␣ t -(340 -350) analogs do not result in any significant modification of the displacement curves may be due to their ability to bind mainly to the uncoupled receptor without changing its conformation and thus the affinity state.
Eleven amino acid peptides from the carboxyl terminus of G␣ t and G␣ s bind to rhodopsin and ␤-adrenergic receptor, respectively, and mimic G t and G s in stabilizing their active conformation (12,13,47). The carboxyl-terminal region of G␣ i subunits may have the necessary sequence for specific and efficient signal transduction from the A 1 adenosine receptor but by itself may be not be able to modulate the high affinity state of the receptor. This raises the question of whether other parts of the G␣ i subunit participate in forming the binding site for the A 1 adenosine receptor. Peptide mapping and proteolytic digestion studies have shown that both rhodopsin and ␤-adrenergic receptor contact at least one other region on G␣ t and G␣ s , respectively, which is located in the ␣4-␤6 and ␤6-␣5 loops of G␣ subunits (12,13,62). Recently, Bae et al. (63) have reported that the presence of G␣ i1 residues 299 -318 in G␣ i1/t chimeras is required both to increase agonist binding to the 5-HT 1B receptor and obtain receptor-stimulated GTP␥S binding. This study has pointed out the importance of the ␣4 helix and ␣4-␤6 loop region for specific recognition between the 5-HT 1B receptor and G␣ i1 subunit. Thus, evidence is accumulating that depicts the receptor binding site on the G␣ subunits as a mosaic with each piece playing a distinct role depending on the type of receptor and G protein. It may be that to stabilize the high affinity state of the A 1 adenosine receptor, a multiple interaction between the receptor and the G␣ subunit or even the heterotrimeric G protein is required.
Functional Effects of the Native G␣ i Carboxyl-terminal Peptide and G␣ t Analogs in Intact Cells-To test the hypothesis that some peptides that disrupt the interaction between the A 1 adenosine receptor and the carboxyl-terminal region of G i proteins are also able to impair signal transduction, we measured the effects of synthetic peptides on the activation of GIRKs by A 1 adenosine receptors in rat hippocampal CA1 pyramidal neurons. Except for peptide 23 which did not show any significant effect on receptor-mediated opening of K ϩ channels, all other peptides reduced the K ϩ current with different activity. Thus, peptides that modulated the high affinity agonist binding to the receptor also disrupted signal transduction between A 1 adenosine receptors and G i proteins. The native peptide G␣ i1/2 -(344 -354) as well as peptide analogs 19 and 24 completely blocked the adenosine-activated response. There did not appear to be a strict correspondence between the ability of the peptides to modulate agonist binding or antagonist displacement by an agonist and their potencies as inhibitors of A 1 adenosine receptor-mediated activation of K ϩ current. However, the efficacy of the native peptide G␣ i1/2 -(344 -354) was consistent with its activity as inhibitor of agonist binding to A 1 adenosine receptors in rat cortical membranes.
Specificity of the Functional Disruption of Receptor G i Protein Coupling by G␣ Peptides-One would expect that a peptide corresponding to the carboxyl terminus of G␣ i would block signaling through all G i -coupled receptors. Although most receptors show specificity for a particular class of G proteins, it is much less clear whether drugs that target the receptor-G protein interface could be highly specific for a particular receptor. Thus, it was important to evaluate the efficacy of G␣ peptides on activation of GIRKs by a different G i -coupled receptor. GABA B receptors modulate the activity of several downstream effectors (64) mainly through the activation of pertussis toxinsensitive G proteins (17,55). However, this receptor differs structurally from other G protein-coupled neurotransmitter receptors and forms a separate gene family together with the metabotropic receptor for L-glutamate (64). In this context, it was of interest that three of the peptides (the native peptide G␣ i1/2 -(344 -354) and peptide analogs 19 and 24) were more effective blockers of A 1 than GABA B receptor responses, whereas peptide 15 was more effective in blocking GABA B than A 1 receptor responses. Although both A 1 and GABA B receptors in rat hippocampal CA1 pyramidal neurons activate what appears to be a common population of GIRKs (65,66), such that outward current responses to baclofen occlude responses to adenosine, there is no definitive evidence that the same G proteins mediate these actions. Thus, the differential effects of these peptides could reflect either the specificity of the interaction of the peptides with homologous but not identical regions of the A 1 and GABA B receptors or could reflect mediation of these responses by different G proteins. In this context, it should also be noted that these peptides are all analogs of the carboxyl terminus of G␣ subunits, whereas it is the ␤␥ dimer that is thought to activate GIRKs (67,68). Thus, although it is possible that these peptides interact directly with GIRKs to inhibit their function, as has been shown for G ␣ai1 (69), this could not explain their selectivity in blocking actions mediated via A 1 receptors versus GABA B receptors. The most probable explanation for these results is that the peptides interact instead with specific elements of the A 1 and GABA B receptors in slightly different ways to disrupt their interaction with the corresponding G proteins. This is also suggested by comparing the potency of the peptides for rhodopsin and A 1 adenosine receptors. Peptide 15, which was the most potent at decreasing A 1 adenosine receptor agonist affinity (Table II), was the least potent of the analogs at stabilizing metarhodopsin II (Table I).
In conclusion, we have found that peptide analogs based on the carboxyl terminus of G␣ t -(340 -350) uncouple A 1 adenosine receptors from G i proteins, but they do not stabilize the high affinity state of the receptor. Peptide analogs with different abilities to modulate agonist binding and signal transduction activation have been also identified. Evidence for selective effects of the analog peptides on different G i -coupled receptors was also demonstrated. Thus, the receptor-G protein interface is a possible target for inhibition of G protein-coupled receptor activation. Future studies will assess whether inhibitors of this interface can be found or designed and whether specificity of inhibition can be achieved.