Differential dissociation of G protein heterotrimers

Abstract

Signalling by heterotrimeric G proteins is often isoform-specific, meaning certain effectors are regulated exclusively by one family of heterotrimers. For example, in excitable cells inwardly rectifying potassium (GIRK) channels are activated by Gβγ dimers derived specifically from G_i/o heterotrimers. Since all active heterotrimers are thought to dissociate and release free Gβγ dimers, it is unclear why these channels respond primarily to dimers released by G_i/o heterotrimers. We reconstituted GIRK channel activation in cells where we could quantify heterotrimer expression at the plasma membrane, GIRK channel activation, and heterotrimer dissociation. We find that G_oA heterotrimers are more effective activators of GIRK channels than G_s heterotrimers when comparable amounts of each are available. We also find that active G_oA heterotrimers dissociate more readily than active G_s heterotrimers. Differential dissociation may thus provide a simple explanation for Gα-specific activation of GIRK channels and other Gβγ-sensitive effectors.

Heterotrimeric GTP-binding regulatory proteins (G proteins) are highly conserved mediators of transmembrane signals (Gilman, 1987). Each heterotrimer (Gαβγ) is made up of a nucleotide-binding Gα subunit and a Gβγ dimer, either of which can interact with downstream effector molecules. The first effector shown to be directly activated by Gβγ dimers was the inwardly rectifying potassium (GIRK) channel responsible for slow synaptic potentials in neurons and cholinergic inhibition of atrial myocytes (Logothetis et al. 1987). In these cells GIRK channels open in response to activation of G protein-coupled receptors (GPCRs) that activate pertussis toxin (PTX)-sensitive G proteins (e.g. G_i and G_o), but not GPCRs that activate other G protein isoforms (e.g. G_s and G_q). The mechanism underlying this specificity has been the subject of extensive investigation. It is generally agreed that specificity does not lie at the level of the GPCRs themselves, as G_s- and G_q-coupled receptors can activate GIRK channels if they are supplied with chimeric Gα subunits (Leaney et al. 2000; Rusinova et al. 2007). It is also accepted that specificity does not reflect activity of a particular Gβγ dimer, as Gβγ dimers containing various Gβ and Gγ isoforms have a similar ability to activate GIRK channels (Lei et al. 2000). Thus Gα subunits in some way confer signalling specificity, even though Gβγ subunits are responsible for channel activation.

Several studies have shown that PTX-sensitive Gα subunits or Gαβγ bind directly to GIRK channels (Huang et al. 1995; Ivanina et al. 2004; Clancy et al. 2005; Rusinova et al. 2007), and thus GIRK channels and G proteins may form preassembled signalling complexes (Riven et al. 2006). Such complexes could confer Gα-dependent specificity by facilitating the interaction of preassembled Gβγ dimers and channels. Although several characteristics of G protein-GIRK channel complexes have yet to be defined, this extreme form of compartmentalization is currently the best explanation available for the specificity of GIRK channel activation.

Here we offer an alternative (or additional) mechanism, namely that the different abilities of G proteins to activate GIRK channels might reflect their different abilities to release free Gβγ dimers. Until recently it was generally expected that G protein activation was immediately followed by dissociation into free Gα–GTP and Gβγ. Since free Gβγ dimers can activate GIRK channels, this expectation implied that there must be a mechanism that allows Gβγ dimers derived from PTX-sensitive G proteins to activate GIRK channels, yet prevents Gβγ dimers derived from other G proteins from activating the same channels. The possibility that G protein isoforms may differ with respect to dissociation has not been extensively studied.

In the present study we found that G_oA proteins were much more efficient activators of GIRK channels than G_s proteins in a heterologous expression system when comparable amounts of each G protein were present at the plasma membrane. In addition, we found that G_oA heterotrimers were much more efficient donors of free Gβγ subunits than G_s heterotrimers. Our results indicate that active Gα_s subunits retain a higher affinity for Gβγ dimers than do active Gα_oA subunits in intact cells. These findings suggest that the ability of PTX-sensitive G proteins to activate GIRK channels and to modulate other Gβγ effectors could be due to their tendency to dissociate into Gα and free Gβγ.

Methods

Plasmid DNA constructs

C-TM-Gα subunits were identical to those previously described (Digby et al. 2006), and consisted of (starting at the amino terminus) a cleavable signal peptide from human growth hormone, enhanced cyan fluorescent protein, the amino terminal 103 amino acids of the rat μ-opioid receptor (including TM 1 and intracellular loop 1) and human Gα subunits. C-TM-Gα_oA and C-TM-Gα_i1 incorporated a C to G mutation at the –4 position to render it insensitive to PTX-mediated ADP-ribosylation. Venus-labelled Gβγ dimers were constructed as previously described (Digby et al. 2006) and in the text, and were fluorescent by virtue of bimolecular fluorescence complementation (Hynes et al. 2004). For metabolic biotinylation of GIRK1 (Kir3.1), a biotin acceptor peptide (GLNDIFEAQKIEWH, where the underlined K residue is biotinylated) (Beckett et al. 1999) was inserted without linkers between residues 114 and 115 of Kir3.1. These subunits were transfected with Kir3.2d and a secreted biotin ligase (Parrott & Barry, 2001) in order to produce biotinylated heteromeric channels. All constructs were made using an adaptation of the QuikChange (Stratagene) mutagenesis protocol (Geiser et al. 2001), were expressed from pcDNA3.1 (Invitrogen), and were verified by automated sequencing. YN-Gγ2 and YC-Gβ1 were generously provided by Stephen Ikeda and Huanmian Chen (National Institutes of Alcoholism and Alcohol Abuse, Rockville, MD, USA). The rat A1R was generously provided by Mark Olah. Kir3.x plasmids were generously provided by Lily Y. Jan. The secreted variant of E. coli biotin ligase (pSecBirA) was generously provided by Michael Barry.

Cell culture and transfection

HEK 293 cells (ATCC) were propagated in plastic flasks and on polylysine coated glass coverslips according to the supplier's protocol. Cells were transfected using polyethyleneimine and were used for experiments 12–48 h later. For experiments involving receptor activation PTX (100 ng ml⁻¹; List Biological Laboratories Inc., Campbell, CA, USA) was added to the culture medium immediately after transfection.

Cell surface staining and flow cytometry

Cells grown in T75 flasks were washed three times with phosphate-buffered saline (PBS), then stained for 5 min with a 1: 500 dilution of Alexa 647-conjugated rabbit anti-green fluorescent protein (GFP) (Invitrogen A31852) or 0.02 mg ml⁻¹ Alexa 647-conjugated streptavidin (Invitrogen S32357). Cells were washed extensively with PBS, detached with Versene (Dow Chemical Co.) and collected in 4% paraformaldehyde in PBS. Flow cytometry was carried out using a Becton Dickinson FACSCalibur flow cytometer equipped with argon (488 nm) and red diode (635 nm) lasers. Venus and Alexa 647 emissions were monitored on FL1 (530 ± 15 nm) and FL4 (661 ± 8 nm), respectively. Each experiment was performed on 10⁴ cells, and in each experiment < 1% of all data points fell above the dynamic range of the detector.

Electrophysiology

Whole-cell voltage-clamp recordings were made using standard procedures from transfected, PTX-treated cells on the stage of an Olympus IX70 inverted fluorescence microscope. Cells were held at a membrane potential of −60 mV, and each second were stepped to −100 mV for 0.2 s, and then ramped from −100 to 0 mV at a rate of 0.18 mV ms⁻¹. Patch electrodes (4–5 MΩ) were filled with a solution containing 140 mm potassium gluconate, 5 mm KCl, 0.2 mm EGTA, 10 mm Hepes, 3 mm MgATP, 0.3 mm Na₂GTP (pH 7.2, ∼295 mosmol (kg H₂O)⁻¹). Cells were perfused with a solution containing 122.5 or 150 mm NaCl, 30 mm or 5 mm KCl, 10 mm Hepes, 10 mm glucose, 1.5 mm CaCl₂, and 2.5 mm MgCl₂ (pH 7.2, ∼320 mosmol (kg H₂O)⁻¹). Solution changes were made using a multiport attachment and perfusion capillary positioned directly in front of the cell under study.

Permeabilization and antibody-mediated crosslinking

Cells were rinsed 3 times in buffer containing 140 mm potassium gluconate, 5 mm KCl, 10 mm Hepes, 1 mm EGTA, 0.3 mm CaCl₂, and 1 mm MgCl₂ (pH 7.2), and incubated at room temperature for 5 min in a 1: 200 dilution of unlabelled polyclonal rabbit anti-GFP IgG (Invitrogen A11122). Cells were washed and incubated for 5 min in a 1: 1000 dilution of goat anti-rabbit antibody (Invitrogen B2770). Cells were permeabilized with 1000 U ml⁻¹ α-haemolysin (Sigma H9395) in the presence of 0.1 mm GTPγS (or BODIPY FL GTPγS for Fig. 4A) plus 10 μm adenosine or 1 mm GDPβS (without adenosine) and incubated for at least 10 min before imaging.

Figure 4. A2AR receptor expression does not limit Gβ₁γ_2/11-V translocation.

Plots of the average normalized ratios of membrane-to-intracellular (M/I) Gβ₁γ_2/11-V fluorescence intensity from cells transfected with C-TM-Gα_s and either 0.1 μg (n = 8) or 0.8 μg (n = 7) of A2AR plasmid DNA; thinner lines represent the average ± s.e.m. The agonist sensitivity was enhanced by transfecting more plasmid DNA, suggesting more A2ARs were expressed. The maximal Gβ₁γ_2/11-V translocation was the same in both cases, suggesting that A2AR expression was not a limiting factor.

Immunoblotting

Transfected cells were washed, then lysed and harvested in RIPA buffer containing protease inhibitors. Samples were boiled and separated by SDS-PAGE on 10% gels. Proteins were transferred to nitrocellulose membranes and detected using primary antibodies raised against GFP (see above), an HRP-conjugated secondary antibody, and enhanced chemiluminescence. As a loading control blots were stripped and reprobed using a primary antibody raised against β-actin.

Imaging

Coverslips were transferred to the stage of a Leica SP2 scanning confocal microscope and imaged using a 63×, 1.4 NA objective. Cells were excited using 458 nm (for cyan fluorescent protein; CFP), 514 nm (for venus), or 633 nm (for Alexa 647) laser lines. For translocation experiments cells were perfused and exposed to adenosine as described above. More than half of these experiments for each group were performed and analysed with the experimenter blind to the transfection condition. No statistical difference was observed between experiments that were performed blind and those that were not, and thus the results were combined. For flourescence recovery after photobleaching (FRAP) experiments low intensity illumination was used during a control (prebleach) period, after which a 4 μm segment of the plasma membrane edge was irreversibly photobleached by increasing the laser intensity to 100%. Recovery of fluorescence into the bleached segment of plasma membrane was monitored for 3–5 min using low intensity illumination. Average pixel intensity in the bleached region was corrected for photobleaching during low intensity illumination, normalized and plotted versus time.

Statistical analysis

Statistical significance was evaluated using one-way ANOVA followed by Bonferroni comparisons or Student's t test; statistical significance was defined as P < 0.01.

Results

Differential activation of GIRK channels by G_oA and G_s heterotrimers

In order to compare the ability of G_oA and G_s heterotrimers to activate GIRK channels and to liberate Gβγ subunits it was necessary to ensure that comparable amounts of these proteins were present at the plasma membrane. It was also necessary to make Gα subunits that were accessible to membrane impermeant crosslinking agents and common antibodies. Therefore, we extended Gα_oA and Gα_s with an n-terminal transmembrane domain (TM) and CFP (C) to produce C-TM-Gα_oA and C-TM-Gα_s. These subunits were transiently expressed in HEK 293 cells together with venus-labelled Gβ₁γ_2/11 dimers (Gβ₁γ_2/11-V) (Hynes et al. 2004), which were constructed so that they would be attached to the plasma membrane by a farnesyl group (see below). We have previously shown that G protein subunits labelled in this manner form functional heterotrimers at the plasma membrane (Digby et al. 2006). Confocal microscopy showed that C-TM-Gα subunits and Gβ₁γ_2/11-V dimers were localized at the plasma membrane as expected, but substantial fluorescence was also present in intracellular compartments (Fig. 1A). Staining of live cells with an Alexa 647-conjugated anti-GFP antibody selectively labelled C-TM-Gα subunits at the cell surface (Fig. 1A), and therefore antibody staining was used to compare heterotrimer abundance at the plasma membrane. Total Gβ₁γ_2/11-V intensity and surface Alexa 647 intensity were measured using flow cytometry, and were found to be positively correlated (Fig. 1B). Comparison of cells transfected with either C-TM-Gα_oA or C-TM-Gα_s indicated that both were expressed at roughly comparable levels, although C-TM-Gα_oA was consistently expressed slightly better (Fig. 1B).

Figure 1. Comparable expression of C-TM-Gα_oA and C-TM-Gα_s.

A, confocal images of live cells expressing C-TM-Gα_oA and Gβ₁γ_2/11-V stained using Alexa 647-conjugated anti-GFP. Alexa 647 signal (633 nm excitation) is present only on the cell surface, whereas C-TM-Gα_oA signal (458 nm excitation) is present at the plasma membrane and in intracellular compartments. B, scatter plots of Alexa 647 and Gβ₁γ_2/11-V fluorescence intensity from samples of 10⁴ cells. Untransfected cells showed no fluorescence in either channel, unstained cells showed Gβ₁γ_2/11-V fluorescence only, and cells transfected with C-TM-Gα_oA and C-TM-Gα_s showed comparable levels of Alexa 647 fluorescence. The quadrants indicate threshold (above background) fluorescence levels. C, an immunoblot probed with a polyclonal anti-GFP antibody (top) and reprobed with an anti-β-actin antibody (bottom). Protein samples were prepared from cells transfected with C-TM-Gα_oA or C-TM-Gα_s (together with unlabelled Gβ and Gγ subunits) and EGFP alone as well as untransfected controls (UT). Both C-TM-Gα subunits migrate as a single predominant species; the larger size of C-TM-Gα_s is predicted by the greater length (40 amino acids) of this Gα isoform.

In five separate experiments the C-TM-Gα_oA intensity (1156 ± 186 arbitrary units, a.u.; mean of means ± s.e.m.) was significantly greater than the C-TM-Gα_s intensity (972 ± 157 a.u.; P < 0.01); the mean ratio of C-TM-Gα_oA to C-TM-Gα_s staining was 1.23 ± 0.19. Control experiments indicated that Alexa 647 staining intensity was sensitive to the amount of C-TM-Gα_oA or C-TM-Gα_s plasmid DNA used for transfection, suggesting that flow cytometry was capable of detecting changes in C-TM-Gα expression (data not shown). Finally, immunoblots probed with a polyclonal antibody raised against GFP detected a single predominant band for both C-TM-Gα_oA and C-TM-Gα_s, suggesting that the majority of each protein was expressed intact (Fig. 1C). Both C-TM-Gα subunits migrated more slowly than expected based on their calculated molecular weights; however, the difference between the C-TM-Gα_oA (701 amino acids) and C-TM-Gα_s (741 amino acids) bands was consistent with the predicted size difference of the mature proteins.

Having established conditions where C-TM-Gα_oA and C-TM-Gα_s abundance was similar, we then asked if GIRK channels would be activated selectively or preferentially by C-TM-Gα_oA. Cells were treated overnight with PTX to disable native Gα_i/o subunits and transfected with GIRK1 and GIRK2d channel subunits, Gβ₁γ_2/11-V, and either C-TM-Gα_s or C-TM-Gα_oA. C-TM-Gα_oA incorporated a mutation (C351G) that rendered it insensitive to PTX, and was activated by coexpressed A₁ adenosine receptors (A₁Rs). C-TM-Gα_s was activated by coexpressed A_2A adenosine receptors (A2ARs). Whole-cell voltage-clamp recordings revealed basal inwardly rectifying potassium currents in transfected cells exposed to 30 mm external K⁺ (Fig. 2A and B). Basal GIRK current was not significantly different in cells expressing C-TM-Gα_oA or C-TM-Gα_s together with Gβ₁γ_2/11-V dimers (P = 0.50), suggesting that similar numbers of GIRK channels were expressed in the two populations of cells. Activation of either A1Rs or A2ARs with 10 μm adenosine evoked additional GIRK current in cells transfected with C-TM-Gα subunits. Adenosine failed to evoke current in cells that were not transfected with exogenous G protein subunits, demonstrating that agonist-activated current was entirely dependent on expressed C-TM-Gα subunits (Fig. 2C). Agonist-induced currents were approximately 5-fold larger in cells expressing C-TM-Gα_oA than in cells expressing C-TM-Gα_s (P < 0.001). In order to confirm that this difference was not due to a difference in GIRK channel expression, we performed a subset of these experiments with GIRK1 subunits that incorporated a 14 amino acid biotin acceptor peptide in the extracellular loop between residues K114 and A115 (Beckett et al. 1999). GIRK channels containing these subunits were metabolically biotinylated by a coexpressed biotin ligase (Parrott & Barry, 2001), and could be specifically labelled by Alexa 647-conjugated streptavidin (Fig. 2D). Flow cytometry of Alexa 647 streptavidin-stained cells confirmed that equal numbers of GIRK channels were present in cells expressing C-TM-Gα_oA and C-TM-Gα_s (Fig. 2E). These results indicated that heterotrimers incorporating either C-TM-Gα_oA or C-TM-Gα_s were capable of activating GIRK channels in these cells. However, the difference in GIRK channel activation by these heterotrimers was too large to be accounted for by the small difference in C-TM-Gα expression, suggesting that G_oA heterotrimers were more efficient activators of these channels. These results are consistent with several previous reports of GIRK channel activation by G_s and G_i/o heterotrimers (Lim et al. 1995; Ruiz-Velasco & Ikeda, 1998; Sorota et al. 1999; Wellner-Kienitz et al. 2001).

Figure 2. Differential activation of GIRK channels by C-TM-G_oA and C-TM-G_s heterotrimers.

A, exemplary traces from cells expressing GIRK1/2 heteromers, Gβ₁γ_2/11-V, and either C-TM-Gα_oA and A1Rs or C-TM-Gα_s and A2ARs. High (30 mm) K⁺ solution reveals basal inward current in both cases, and additional current in the presence of adenosine (10 μm); Ba²⁺ refers to 0.2 mm barium. The slow reversal of adenosine-activated current in the cell expressing C-TM-Gα_s was characteristic of these cells. B, current responses to voltage ramps from –90 mV to 0 mV recorded from the same cells shown in panel A. C, summary of basal and evoked GIRK currents in cells expressing the indicated combinations of adenosine receptors, C-TM-Gα subunits and Gβ₁γ_2/11-V dimers or adenosine receptors only. C-TM-Gα_oA subunits harboured a mutation (C351G) rendering them insensitive to PTX, and all cells were treated overnight with PTX. D, a confocal image showing plasma membrane fluorescence in cells expressing metabolically biotinylated GIRK1/2 channels stained with Alexa 647-conjugated streptavidin. E, scatter plots of Alexa 647 and Gβ₁γ_2/11-V fluorescence intensity from samples of 10⁴ cells expressing metabolically biotinylated GIRK1/2 channels and either C-TM-Gα_oA or C-TM-Gα_s. Live cells were stained with Alexa 647-conjugated streptavidin (SA). Comparable levels of Alexa 647 fluorescence are evident above background levels (indicated by the quadrants; representative data from 3 similar experiments).

Differential dissociation of G_oA and G_s heterotrimers

One possible reason why G_oA heterotrimers are more efficient activators of GIRK channels than G_s heterotrimers is that the former are more likely to dissociate, and thus to liberate free Gβγ dimers. To compare dissociation of Gα_oA and Gα_s heterotrimers we took advantage of the fact that farnesylated Gβγ dimers dissociate from the plasma membrane after dissociating from Gα subunits (Akgoz et al. 2004; Kassai et al. 2005; Rosenzweig et al. 2007). The Gγ_2/11 subunit constructed for these experiments was essentially Gγ₂, with the exception that the six amino acids at the extreme C-terminus (FFCAIL) were replaced with the corresponding amino acids from Gγ₁₁ (GSCVIS). These amino acids specify lipid modification (at the –4 cysteine) with a 15-carbon farnesyl group as opposed to the usual 20-carbon geranylgeranyl group found on Gγ₂. Previous studies have shown that fluorescently labelled Gβγ dimers with Gγ₁₁ or chimeras similar to Gγ_2/11 translocate from the plasma membrane into the cell interior (possibly binding to the Golgi apparatus) after dissociation from active Gα subunits (Akgoz et al. 2004, 2006), providing a sensitive assay for heterotrimer dissociation in intact cells. Gβ₁γ_2/11-V dimers localized to the plasma membrane in unstimulated cells when coexpressed with C-TM-Gα subunits and adenosine receptors (Fig. 1A and 3A). Agonist activation caused a portion of the membrane-localized Gβ₁γ_2/11-V fluorescence to rapidly leave the plasma membrane and enter the cell interior (Fig. 3A). In the same cells membrane-localized C-TM-Gα fluorescence was not changed by agonist application (data not shown). We compared translocation of Gβ₁γ_2/11-V dimers liberated from PTX-treated cells expressing C-TM-Gα_oA or C-TM-Gα_s as described above. The normalized membrane/intracellular (M/I) venus intensity ratio decreased by 28.0 ± 1.7% (n = 50) in cells expressing C-TM-Gα_oA, but only 8.8 ± 1.0% (n = 52) in cells expressing C-TM-Gα_s (Fig. 3B; P < 0.001). Since expression levels varied widely from cell to cell after transient transfection (Fig. 1B), we wanted to ensure that we had not selected cells that expressed different amounts of C-TM-Gα or Gβ₁γ_2/11-V in these experiments. Therefore, we quantified CFP and venus fluorescence intensities in a narrow region of interest centred on the plasma membrane. The mean C-TM-Gα_oA intensity in these experiments was 1.21 times that of C-TM-Gα_s, in close agreement with the value obtained with flow cytometry. However, neither CFP nor venus fluorescence intensity was significantly different in cells expressing the two C-TM-Gα subunits (P > 0.1 for both comparisons; Fig. 3C). Gβ₁γ_2/11-V translocation after stimulation of C-TM-Gα_s was not limited by inadequate A2AR expression, as increasing the amount of A2AR plasmid DNA used for transfection increased agonist sensitivity (indicating the presence of spare receptors) but did not increase the maximal extent of translocation (Fig. 4). These results suggest that C-TM-G_oA heterotrimers liberate roughly 2–3 times as many free Gβγ dimers as a comparable number of C-TM-G_s heterotrimers.

Figure 3. Differential dissociation of C-TM-G_oA and C-TM-G_s heterotrimers.

A, images of venus fluorescence in a cell expressing C-TM-Gα_oA, Gβ₁γ_2/11-V and A1Rs before (control) and after (adenosine) application of 10 μm adenosine; the lookup table is inverted for clarity. Gβ₁γ_2/11-V fluorescence translocates from the plasma membrane to the cell interior. This is shown more clearly in the difference image (right), where regions that lost fluorescence after adenosine are pseudocoloured blue, and regions that gained fluorescence are pseudocoloured red. Regions of interest (ROI) typical of membrane and intracellular regions are show in blue and red, respectively. B, plots of the average normalized ratios of membrane-to-intracellular (M/I) Gβ₁γ_2/11-V fluorescence intensity from cells expressing C-TM-Gα_oA and A1Rs (n = 50) or C-TM-Gα_s and A2ARs (n = 52); thinner lines represent the average ± s.e.m. C, CFP (left) and venus (right) fluorescence intensity in a narrow region of interest including the plasma membrane in cells expressing C-TM-Gα_oA (n = 15) or C-TM-Gα_s (n = 18) together with Gβ₁γ_2/11-V and cognate adenosine receptors.

Differential affinity of active Gα_oA and Gα_s for Gβγ

We next considered the possible mechanisms responsible for differential dissociation of G_oA and G_s heterotrimers after agonist activation of GPCRs. These can be divided into two broad categories: either a smaller fraction of G_s heterotrimers is activated by ligand-bound receptors, or Gβγ has a higher affinity for Gα_s–GTP than for Gα_oA–GTP. To discriminate between these possibilities we eliminated the potential contribution of the activation step by loading cells with the slowly hydrolysable GTP analogue GTPγS, and then compared the affinities of C-TM-Gα_s–GTPγS and C-TM-Gα_oA–GTPγS for Gβγ using fluorescence recovery after photobleaching (FRAP). In this method C-TM-Gα subunits are immobilized by cell surface crosslinking, and the effect of immobile C-TM-Gα on Gβ₁γ₂-V mobility is used as an indicator of binding (Digby et al. 2006). C-TM-Gα_oA and C-TM-Gα_s were selectively crosslinked with an anti-GFP antibody followed by a secondary antibody. Cells were then permeabilized in potassium gluconate buffer with staphylococcal α-haemolysin (α-toxin) in the presence of 0.1 mm GTPγS or 1 mm GDPβS, which served as a control. Nucleotide loading was directly confirmed by imaging uptake of BODIPY FL GTPγS (Fig. 5A). Cells expressed either C-TM-Gα_oA or C-TM-Gα_s, their cognate adenosine receptors and Gβ₁γ₂-V. Geranylgeranylated Gγ₂ subunits were used in these experiments so that Gβ₁γ₂-V would remain attached to the plasma membrane after activation. In experiments with GTPγS adenosine (10 μm) was added to accelerate nucleotide exchange. FRAP experiments showed that under control conditions C-TM-Gαβ₁γ₂-V heterotrimers were highly mobile, as indicated by complete recovery of fluorescence 180 s after photobleaching (Fig. 5C and D). Crosslinking immobilized C-TM-Gα subunits, and decreased the mobility of Gβ₁γ₂-V, confirming the formation of heterotrimers (Digby et al. 2006). For both C-TM-Gα_oA- and C-TM-Gα_s-expressing cells, Gβ₁γ₂-V mobility was significantly greater in cells loaded with GTPγS than in cells loaded with GDPβS (P < 0.001; Fig. 5B, C and D). This result suggests that heterotrimers were activated by GTPγS loading under these conditions, as shown previously for photoreceptor cells (Rosenzweig et al. 2007). However, Gβ₁γ₂-V mobility was significantly greater in cells expressing C-TM-Gα_oA than in cells expressing C-TM-Gα_s, irrespective of the loaded nucleotide (P < 0.0005; Fig. 5B, C and D). This result suggests that GDPβS- and GTPγS-bound C-TM-Gα_s subunits retain a higher affinity for Gβ₁γ₂-V than their C-TM-Gα_oA counterparts (Sarvazyan et al. 2002). This difference could account for differential release of free Gβγ dimers by active Gα_oA and Gα_s.

Figure 5. The apparent affinity of active C-TM-Gα_s for Gβ₁γ₂-V is higher than that of active C-TM-Gα_oA.

A, images of HEK 293 cells bathed in 0.1 mm BODIPY FL GTPγS. Intact cells exclude green fluorescence, whereas cells permeabilized with α-toxin accumulate the labelled nucleotide. B, recovery of Gβ₁γ₂-V fluorescence into photobleached regions of the plasma membrane in cells expressing immobile C-TM-Gα_oA or C-TM-Gα_s. Cells were permeabilized in the presence of 0.1 mm GTPγS and 10 μm adenosine or 1 mm GDPβS. Traces are the average of all experiments summarized in panels C and D. C, summary of fluorescence recovery 180 s after photobleaching in cells expressing C-TM-Gα_oA (left) together with Gβ₁γ₂-V (right). Both were mobile in cells that were not antibody crosslinked (no Ab; n = 6), as indicated by complete fluorescence recovery. Both became less mobile in crosslinked cells loaded with GDPβS (n = 11). C-TM-Gα_oA mobility was not significantly increased in crosslinked cells loaded with GTPγS, whereas Gβ₁γ₂-V mobility was significantly increased in these cells (n = 14), indicating a GTPγS-induced decrease in the affinity of C-TM-Gα_oA for Gβ₁γ₂-V. D, summary of fluorescence recovery in cells expressing C-TM-Gα_s (left) and Gβ₁γ₂-V (right). Both were mobile without antibody crosslinking (no Ab; n = 6), and both became less mobile in crosslinked cells loaded with GDPβS (n = 10). C-TM-Gα_s mobility was not increased in cells loaded with GTPγS, whereas Gβ₁γ₂-V mobility was significantly increased in these cells (n = 14), indicating a GTPγS-induced decrease in the affinity of C-TM-Gα_s for Gβ₁γ₂-V. Comparison of cells expressing C-TM-Gα_oA and C-TM-Gα_s indicated that Gβ₁γ₂-V mobility was greater both in cells loaded with GDPβS (P < 0.0005) and in cells loaded with GTPγS (P < 0.00001), indicating that in both cases the affinity of immobile C-TM-Gα_oA for Gβ₁γ₂-V was lower.

Parallel experiments with C-TM-Gα_i1

PTX-sensitive G proteins in general are effective activators of Gβγ-sensitive effector molecules such as GIRK channels. This suggests that PTX-sensitive Gα subunits other than Gα_oA should also efficiently release free Gβγ dimers. To test this prediction we repeated several experiments with C-TM-Gα_i1. In cells expressing C-TM-Gα_i1 (and Gβ₁γ_2/11-V) adenosine-induced GIRK current averaged 702 ± 149 pA (n = 9), which was significantly greater than in cells expressing C-TM-Gα_s (260 ± 29 pA, n = 53; P < 0.01), but not significantly different from in cells expressing C-TM-Gα_oA (1267 ± 146 pA, n = 46; P = 0.10). C-TM-Gα_i1 also supported translocation of Gβ₁γ_2/11-V dimers away from the plasma membrane. The membrane/intracellular (M/I) venus intensity ratio decreased by 13.0 ± 1.4% (n = 22) cells expressing C-TM-Gα_i1. In experiments carried out in parallel, this ratio decreased by 18.0 ± 1.3% in cells expressing C-TM-Gα_oA (n = 39), but only 6.7 ± 0.7% (n = 29) in cells expressing C-TM-Gα_s. Thus C-TM-Gα_i1 was intermediate between C-TM-Gα_oA and C-TM-Gα_s with respect to liberation of free Gβ₁γ_2/11-V dimers; all three values were significantly different from each other (P < 0.01, one-way ANOVA). Finally, we assessed the ability of immobile (antibody crosslinked) C-TM-Gα_i1 to decrease the lateral mobility of Gβ₁γ₂-V dimers. In permeabilized cells loaded with GDPβS, recovery of Gβ₁γ₂-V fluorescence at 180 s was 31.8 ± 2.9% (n = 10) for C-TM-Gα_i1, compared to 48.6 ± 5.2% (n = 11; P < 0.05) for C-TM-Gα_oA and 23.7 ± 1.5% (n = 10; P > 0.05) for C-TM-Gα_s. This suggested that the affinity of inactive C-TM-Gα_i1 subunits for Gβ₁γ_2/11-V dimers was more similar to that of C-TM-Gα_s than C-TM-Gα_oA (Sarvazyan et al. 2002). In contrast, in cells loaded with GTPγS recovery of Gβ₁γ₂-V fluorescence was 67.5 ± 6.2% (n = 10) for C-TM-Gα_i1, compared to 75.6 ± 2.4% (n = 14; P > 0.05) for C-TM-Gα_oA and 43.7 ± 2.9% (n = 14; P < 0.05) for C-TM-Gα_s. This implied that the affinity of active C-TM-Gα_i1 subunits for Gβ₁γ_2/11-V dimers was more similar to that of C-TM-Gα_oA than C-TM-Gα_s. Taken together these results suggest that under these conditions C-TM-Gα_i1 is a more effective donor of free Gβγ dimers than C-TM-Gα_s, but slightly less effective than C-TM-Gα_oA.

Discussion

In the mammalian CNS activation of GPCRs that couple to PTX-sensitive G proteins inhibits neurons in part by opening GIRK channels. In contrast, activation of GPCRs that couple to PTX-insensitive G proteins often increases excitability, and does not lead to GIRK channel opening (Nicoll, 1988). Analogous regulation of GIRK channels by G proteins is observed in cardiac myocytes (Breitwieser & Szabo, 1985; Pfaffinger et al. 1985). The fact that these channels are activated by direct binding of Gβγ dimers (Logothetis et al. 1987; Reuveny et al. 1994; Wickman et al. 1994; Huang et al. 1995) raises a question of specificity that is often framed in the following terms: if GIRK channels are activated by Gβγ dimers and all active G protein heterotrimers liberate free Gβγ dimers, then why aren't these channels activated by PTX-insensitive G proteins? Here we have shown that G_oA heterotrimers are better sources of free Gβγ dimers than G_s heterotrimers. Differential dissociation provides a parsimonious mechanism for activation of Gβγ effectors by specific G protein isoforms.

In solution active G protein heterotrimers readily dissociate into Gα–GTP subunits and Gβγ dimers (Gilman, 1987). It has long been thought that this also occurs in vivo, although some authors have emphasized that signalling may not require physical dissociation (Rebois et al. 1997; Levitzki & Klein, 2002). Physical dissociation of G proteins in intact cells has recently been demonstrated (Digby et al. 2006), but it is not clear to what extent this applies to various Gα isoforms. Studies using fluorescence resonance energy transfer (FRET) and bioluminescence resonance energy transfer (BRET) have shown both increases and decreases in energy transfer between various labelled Gα subunits and Gβγ dimers depending on the location of the label and the Gα isoform under study (Janetopoulos & Devreotes, 2002; Bunemann et al. 2003; Azpiazu & Gautam, 2004; Frank et al. 2005; Gales et al. 2006; Gibson & Gilman, 2006). These studies make it clear that G protein heterotrimers do not dissociate instantaneously after GTP binding, and that some fraction remains intact throughout an entire GTP binding–hydrolysis cycle. However, isoform-specific differences in the fraction of active heterotrimers that dissociates are difficult to address using FRET and BRET owing to possible differences in the efficiency of energy transfer between both different isoforms and different subpopulations of each isoform. Unambiguous demonstration of differential heterotrimer dissociation requires the use of methods that discriminate physical dissociation from conformational changes. In the present study we monitored the loss of Gβ₁γ_2/11-V fluorescence from the plasma membrane after activation of C-TM-Gα_oA and C-TM-Gα_s. These modified Gβγ dimers dissociate from the plasma membrane when they dissociate from Gα (Akgoz et al. 2004, 2006), and thus provide a sensitive live-cell indicator of heterotrimer dissociation. We found that activation of C-TM-Gα_oA produced roughly three times as many free Gβγ dimers as activation of C-TM-Gα_s when both Gα subunits were expressed at comparable densities. C-TM-Gα_i1 was an intermediate donor of Gβγ, although we did not measure the abundance of this subunit.

Several possible mechanisms could account for differential release of free Gβγ dimers. These mechanisms can be divided into those that involve steps prior to GTP binding (i.e. receptor activation and GDP release) and those that involve GTP binding and steps after GTP binding (i.e. the conformational change in the Gα switch regions and dissociation of Gβγ from active, GTP-bound Gα). Either type of mechanism could contribute to signalling specificity. As a first step towards determining the mechanism underlying differential release of Gβγ dimers, we tested whether the relatively modest dissociation from C-TM-Gα_s was due to fewer of these subunits being driven to the active state or a greater affinity between active C-TM-Gα_s subunits and Gβγ dimers. Our FRAP results from cells loaded with GTPγS suggest that both inactive and active C-TM-Gα_s subunits have a higher affinity for Gβγ dimers than their C-TM-Gα_oA counterparts, suggesting the difference between these two Gα subunits lies downstream of GTP binding. These results are also in agreement with a biochemical study that reported a 40-fold higher affinity between inactive Gα_s and Gβγ than between inactive Gα_o and Gβγ (Sarvazyan et al. 2002), although this comparison was between subunits that did not contain the lipid modifications found in Gα subunits in cells. Comparison of affinities between active Gα–GTP (or Gα–GTPγS) isoforms and Gβγ dimers in vitro has not been reported. Inactive Gα subunits contact Gβγ dimers at both a switch region interface, which changes conformation after GTP binding, and an amino-terminal (αN helix) interface, which does not change conformation after GTP binding (Wall et al. 1995; Lambright et al. 1996). One possibility is that structural differences between the different Gα isoforms in these regions determines differences in the likelihood that heterotrimers will physically dissociate prior to GTP hydrolysis. However, in preliminary experiments with chimeric Gα subunits we have thus far failed to identify a clear structural basis for differential heterotrimer dissociation (data not shown). Therefore, at the present time we can only conclude that the difference between the Gα subunits that we studied is likely to be due to differential dissociation of active heterotrimers rather than differential activation of heterotrimers. It is also important to emphasize that our experiments were carried out with specific modified G protein subunits in a model cell system, and our conclusions are correspondingly limited to this experimental system. Additional experiments will be required to determine if wild-type Gα subunits differentially release Gβγ after activation.

If active G_s heterotrimers are less likely to liberate free Gβγ dimers, then they should also be less able to activate Gβγ effectors. Our results verify this prediction. We found that heterotrimers containing C-TM-Gα_s subunits could activate GIRK channels, but that they did so less efficiently than heterotrimers containing C-TM-Gα_oA subunits. Our results agree well with several previous studies that demonstrated activation of GIRK channels by Gβγ dimers derived from G_s heterotrimers, although in every case this activation was less robust than that mediated by PTX-sensitive heterotrimers (Lim et al. 1995; Ruiz-Velasco & Ikeda, 1998; Sorota et al. 1999; Wellner-Kienitz et al. 2001). Taken together these results suggest that the specificity of GIRK channel activation is relative rather than absolute. An important distinction between our results and previous studies is the quantitative comparison of G protein abundance at the plasma membrane made possible by our reconstitution strategy. Although previous studies have shown that native or overexpressed G_s heterotrimers can activate GIRK channels, none of these studies assessed the relative abundance of PTX-insensitive and PTX-sensitive heterotrimers at the plasma membrane. Our present results extend previous work by showing that specificity persists even when the availability of G protein heterotrimers at the cell surface is comparable, and thus reflects the identity of the Gα subunits rather than simply their abundance. This conclusion is based on the assumption that our flow cytometry and immunoblot experiments detected fully functional C-TM-Gα subunits.

Can differential dissociation of G protein isoforms account entirely for the degree of signalling specificity observed in native neurons and myocytes? The answer to this question depends on the relative expression of G protein isoforms in native cells, as an overabundance of Gα_s could compensate for the relatively poor ability of this subunit to provide free Gβγ. Gα_o is highly abundant in the brain, where it represents approximately 0.5% of the total membrane protein (Sternweis & Robishaw, 1984; Huff et al. 1985). It has been estimated that the abundance of Gα_s in the brain is less than one-fifth that of Gα_o (Huff et al. 1985). Therefore, it seems possible that the relative scarcity of Gα_s combined with its reluctance to release Gβγ dimers can account for the fact that G_s heterotrimers do not activate GIRK channels (or other Gβγ effectors) in neurons. The situation may be very different in tissues, cells, or subcellular compartments where the relative expression of Gα_s is higher. The suggestion that differential dissociation is important for signalling specificity predicts that PTX-sensitive G proteins in general will be better donors of Gβγ dimers than PTX-insensitive G proteins. Preliminary results assessing the ability of other heterotrimers (e.g. Gα_q) to release free Gβγ are consistent with this expectation (G. J. Digby & N. A. Lambert, unpublished observations).

Gα subunits and heterotrimers can bind to GIRK channels directly (Huang et al. 1995; Ivanina et al. 2004; Clancy et al. 2005; Rusinova et al. 2007). This finding has led to the suggestion that GIRK channels form stable multimolecular signalling complexes with G proteins and GPCRs (Lavine et al. 2002; Clancy et al. 2005; Riven et al. 2006). It has also been suggested that Gα subunits directly regulate GIRK channel function by priming them for Gβγ activation (Peleg et al. 2002). Two recent studies using mutant GIRK channels (Clancy et al. 2005) and chimeric Gα subunits (Rusinova et al. 2007) found a correlation between Gα binding to and receptor-mediated activation of these channels, suggesting that complex formation contributes to specificity. In addition, Ivanina et al. (2004) found that Gα_i1 and Gα_i3 differentially bound to and activated GIRK channels even though comparable amounts of both were present in cells. Preassembled complexes and differential dissociation are not mutually exclusive mechanisms of specific signalling provided free Gβγ dimers can still activate channels already bound by Gαβγ. Indeed, the degree of specificity that we observed with respect to GIRK channel activation (5-fold) was greater than the degree of differential subunit dissociation (3-fold), suggesting that additional mechanisms of specificity are likely to be involved. It is conceivable that differential dissociation plays no part in the specificity of GIRK channel activation if preassociated heterotrimers are the sole source of Gβγ capable of activating the channels, and if the mechanism used by heterotrimers to deliver Gβγ dimers to their associated channels differs substantially from physical dissociation. We think this possibility is unlikely for several reasons. First, overexpression of Gβγ alone activates GIRK channels in cells where there is ample native Gα to support receptor-mediated activation (and thus presumably complex formation) (Lei et al. 2000; Ruiz-Velasco & Ikeda, 2001). Second, several studies (including this one) have shown that G_s heterotrimers can activate GIRK channels even though these heterotrimers apparently do not bind to the GIRK channel C-terminus (Clancy et al. 2005). Finally, several Gβγ-binding proteins including overexpressed Gα subunits, phosducin and βARK can block receptor-mediated GIRK channel activation, presumably by intercepting receptor-mobilized Gβγ dimers before they activate GIRK channels (Fernandez-Fernandez et al. 2001; Peleg et al. 2002; Rishal et al. 2005). Provided these Gβγ ‘buffers’ do not prevent heterotrimer formation to begin with, their ability to interfere with agonist-induced channel activation would suggest that their access to the GIRK-binding surface of Gβγ is at least as good as that of the channels themselves, which seems unlikely if heterotrimers and channels are exclusively preassociated.

In summary, we have shown that G protein isoforms differ substantially with respect to their ability to liberate free Gβ₁γ_2/11 dimers. This difference may be important for Gα-specific activation of Gβγ-responsive effectors. Our results support the view that there is a dynamic equilibrium between Gα-GTPGβγ, Gα-GTP and Gβγ. Therefore, in addition to the rates of guanine nucleotide exchange and GTP hydrolysis, the affinity of active Gα subunits and Gβγ dimers for each other is an important parameter that can contribute to functional differences between G protein isoforms, including activation of Gβγ effectors. The functional consequences of subunit dissociation for trafficking have been highlighted by a recent study in photoreceptor cells. Active rod transducin subunits translocate from the rod outer segment because these subunits dissociate from each other, whereas active cone transducin subunits remain in the cone outer segment because they are less prone to dissociation (Rosenzweig et al. 2007). Interestingly, cone transducin did dissociate if cone Gα_t was expressed in rod cells, suggesting the identity of the Gβγ dimer was important for the difference in this system. This study together with the present study underscores the need to consider the identity of all three G protein subunits. Taken together, our results and those of previous studies suggest that models of G protein signalling should take into account the propensity of a particular heterotrimer to dissociate after activation in cases where physical dissociation is important for function.