Bioluminescence resonance energy transfer (BRET) and co-immunoprecipitation experiments revealed that heterotrimeric G proteins and their effectors were found in stable complexes that persisted during signal transduction. Adenylyl cyclase, Kir3.1 channel subunits and several G-protein subunits (Gαs, Gαi, Gβ1 and Gγ2) were tagged with luciferase (RLuc) or GFP, or the complementary fragments of YFP (specifically Gβ1-YFP1-158 and Gγ2-YFP159-238, which heterodimerize to produce fluorescent YFP-Gβ1γ2). BRET was observed between adenylyl-cyclase-RLuc or Kir3.1-RLuc and GFP-Gγ2, GFP-Gβ1 or YFP-Gβ1γ2. Gα subunits were also stably associated with both effectors regardless of whether or not signal transduction was initiated by a receptor agonist. Although BRET between effectors and Gβγ was increased by receptor stimulation, our data indicate that these changes are likely to be conformational in nature. Furthermore, receptor-sensitive G-protein-effector complexes could be detected before being transported to the plasma membrane, providing the first direct evidence for an intracellular site of assembly.
Heptahelical receptors, heterotrimeric G proteins and the downstream effectors that they regulate constitute the principal components of G-protein-mediated signal transduction systems. How these systems are organized within membranes remains contentious. The prevailing hypothesis has been that signaling proteins move about independently within the membrane and interact as a consequence of random collisions. However, most cells have multiple G-protein-mediated signal transduction pathways that require many different signaling components with the potential to work at cross-purposes if allowed to interact randomly. An alternative organizational paradigm in which receptors, G proteins and effectors are assembled into signaling complexes might explain the specificity and efficacy that is often observed in response to different environmental stimuli in vivo (Rebois and Hébert, 2003).
There is substantial evidence that G-protein-mediated signal transduction systems are organized as macromolecular complexes in yeast (Dohlman and Thorner, 2001) and Drosophila (Montell, 1999). Most of the evidence for the existence of stable protein-protein interactions between G proteins and effectors in mammalian cells comes from in vitro biochemical studies (e.g. Arad et al., 1984) (reviewed by Rebois et al., 1997). The first proposal that G proteins and effectors existed as a complex in cell membranes was based on data indicating that signal transduction displays first order kinetics (Levitzki, 1981; Levitzki, 1984; Levitzki, 1986; Levitzki, 1988a; Levitzki, 1988b; Marbach et al., 1990; Tolkovsky et al., 1982; Tolkovsky and Levitzky, 1981). The difficulty of studying protein interaction in the cell membrane meant that these studies stood alone for many years as evidence for these complexes in situ.
Recent studies using molecular imaging techniques in living cells such as fluorescence resonance energy transfer (FRET) or bioluminescence resonance energy transfer (BRET) have demonstrated that G-protein subunits remain associated with each other (Bünemann et al., 2003; Frank et al., 2005), and with their receptors (Galés et al., 2005; Nobles et al., 2005) during signal transduction, and that receptors are also stably associated with effectors (Lavine et al., 2002). The purpose of the present study was to use BRET to determine if G proteins and their effectors form stable complexes in living mammalian cells. Two different effectors, adenylyl cyclase (AC) and an inwardly rectifying K+ channel subunit (Kir3.1), and several different G-protein subunits were tagged with RLuc, GFP or one of two complementary fragments of YFP. The tagged signaling proteins retained biological activity, and they could be co-immunoprecipitated when co-expressed in mammalian cells, demonstrating that they could form detergent-resistant assemblies. Furthermore, BRET experiments showed that these proteins formed complexes of G-protein subunits and effectors in living cells, and these were stable regardless of whether or not signal transduction was activated.
GFP- and RLuc-tagged signaling proteins are functional
Both G-protein subunits and effectors retain their ability to function in signal transduction when tagged for BRET experiments. Previously, both AC and the K+ channel subunit Kir3.1, were shown to be functional when modified by the addition of RLuc to their C-termini (Lavine et al., 2002), as was Kir3.1-GFP where GFP is attached to the C-terminus of Kir3.1 (Mirshahi and Logothetis, 2004). Both Gβ1 and Gγ2 were tagged for use in BRET experiments. GFP-tagged Gβ1 can modulate voltage-gated Ca2+ channels (Doering et al., 2004). Further evaluation of the tagged subunits was carried out by comparing their ability to regulate K+ currents relative to their untagged counterparts. K+ currents were measured in Xenopus oocytes expressing Kir3.2, which form functional homotetrameric channels that open in response to Gβγ (Schoots et al., 1999). Although wild-type Gβ1 and Gγ2, individually, were unable to activate Kir3.2, together they produced a 2.5-fold increase in the basal K+ current (Fig. 1A). Tagged G-protein subunits could also increase Kir3.2 current, but only if co-expressed with their complementary untagged counterparts. GFP-Gβ1 or RLuc-Gβ1, resulted in an approximately 1.5-fold increase in basal current if oocytes co-expressed Gγ2. GFP-Gγ2 was able to increase the basal K+ current by approximately 1.6-fold in the presence of Gβ1 (Fig. 1A). As a more rigorous functional test, we determined whether Gβγ subunits tagged for BRET experiments could transduce the signal between the receptor and Kir3.2. K+ currents were measured in oocytes expressing various combinations of tagged or wild-type Gβ1 and Gγ2 subunits together with the β2AR, Gαs and Kir3.2. In the absence of co-expressed Gβγ, very little stimulation is seen (Fig. 1B compare top trace with second trace). Tagging Gβ1 or Gγ2 for BRET (traces 3-5) did not affect the kinetics of agonist-mediated channel activation, though channel deactivation following removal of the agonist, occurred more rapidly (Fig. 1B,C).
The abilities of Gγ2 and GFP-Gγ2 to mediate agonist-dependent stimulation of AC-RLuc were also compared. Agonist-induced cAMP accumulation was measured in cells co-expressing GFP-Gγ2, AC-Rluc, Gαs and Gβ1 or Gγ2, AC, Gαs and Gβ1. The expression of either wild-type or tagged signaling proteins resulted in a much higher maximum agonist-induced cAMP accumulation compared with sham-transfected cells, indicating that the exogenous proteins were functional (see Fig. 2 legend). These experiments also revealed that the EC50 for isoproterenol in cells expressing GFP-Gγ2 and AC-RLuc was similar in cells transfected with wild-type signaling proteins or in sham-transfected cells (Fig. 2).
Gα subunits are crucial for agonist-mediated regulation of AC and Kir3 channels. The stimulatory G-protein α-subunit (Gαs) retains its ability to mediate hormone-induced activation of AC when GFP is properly positioned within its α-helical domain to create a Gαs-GFP fusion protein (Yu and Rasenick, 2002; data not shown). We prepared an analogous construct in which RLuc was inserted into Gαi and found that it was able to mediate receptor stimulation of Kir3.2 or inhibition of AC expressed in HEK 293 cells (data not shown).
Tagged G-protein subunits and effectors are co-immunoprecipitated
Studies have demonstrated that Gβγ is co-immunoprecipitated with complexes containing AC (Davare et al., 2001), and that Gβγ can interact directly with the cytosolic domains of AC (Weitmann et al., 2001). In our study, detergent extracts from cells co-expressing AC-RLuc and GFP-Gβ1 were immunoprecipitated with anti-GFP antibodies. Probing western blots with anti-RLuc antibodies revealed the presence of AC-RLuc indicating co-immunoprecipitation with GFP-Gβ1 (Fig. 3A, left panel). No co-immunoprecipitation occurred when wild-type Gβ1 was used to compete for the interaction with AC-RLuc or when RLuc was substituted for AC-RLuc (Fig. 3A middle and right panels).
In an effort to quantify these interactions, the proportion of RLuc activity that was co-immunoprecipitated by anti-GFP antibodies was determined. More than 85% of an expressed RLuc-GFP fusion protein could be immunoprecipitated with anti-GFP, demonstrating the efficacy of the procedure. Gβ1 and Gγ2 form a detergent-resistant protein complex, but anti-GFP antibodies were able to precipitate only about 50% of the RLuc activity from cells co-expressing RLuc-Gβ1 and GFP-Gγ2 (Table 1). Other combinations of RLuc and GFP-tagged signaling proteins were also co-immunoprecipitated by anti-GFP antibodies. AC-RLuc was consistently immunoprecipitated with GFP-Gγ2 when co-expressed with β2AR, Gαs, and Gβ1 but the quantity of RLuc activity recovered was less than 10% of total luciferase activity. This result was not due to AC-RLuc being expressed in molar excess of the GFP-Gγ2. Conceivably, the efficiency with which AC-RLuc and GFP-Gγ2 form a complex in vivo may be compromised (e.g. because some crucial component is not being expressed in stoichiometric amounts or because endogenous proteins compete with the tagged proteins in forming a complex). Alternatively, the interaction between AC-RLuc and GFP-Gγ2 may not be stable enough to survive the conditions required for immunoprecipitation.
|Expressed proteins .||% RLuc precipitated .|
|GFP-Gγ2 + Kir3.1-RLuc||3.4±1.2|
|M2R + Gαi2 + Gβ1 + GFP-Gγ2 + Kir3.1-RLuc + Kir3.4||8.0±4.1|
|β2AR + Gαs + Gβ1 + GFP-Gγ2 + AC-RLuc||4.5±0.6|
|β2AR + Gαs + RLuc-Gβ1 + GFP-Gγ2 + AC||54±18|
|Expressed proteins .||% RLuc precipitated .|
|GFP-Gγ2 + Kir3.1-RLuc||3.4±1.2|
|M2R + Gαi2 + Gβ1 + GFP-Gγ2 + Kir3.1-RLuc + Kir3.4||8.0±4.1|
|β2AR + Gαs + Gβ1 + GFP-Gγ2 + AC-RLuc||4.5±0.6|
|β2AR + Gαs + RLuc-Gβ1 + GFP-Gγ2 + AC||54±18|
HEK 293 cells were transfected as indicated. GFP in the detergent soluble material from these cells was precipitated with anti-GFP antibodies. To control for non-specific immunoprecipitation, RLuc- and GFP-tagged proteins were expressed in separate cell populations that were combined before detergent solubilization and immunoprecipitation. The wild-type protein was substituted for the missing fusion protein in each of these populations. The amount of precipitated RLuc activity was compared with the RLuc activity remaining in solution to determine the percentage of RLuc activity that was precipitated, and values obtained for non-specific immunoprecipitation were subtracted from the appropriate experimental sample to determine what percentage of the RLuc activity was specifically precipitated. The data represent the mean ± s.d. for four (RLuc-Gβ1) or five (Kir3.1-RLuc and AC-RLuc) independent experiments. Similar results (1.9±0.7% of total RLuc precipitated) were obtained when GFP-Gαs was immunoprecipitated and associated AC-RLuc was detected by measuring luciferase activity in cells expressing β2AR, GFP-Gαs, Gβ1, Gγ2 and AC-RLuc. This was above the background from cells expressing β2AR, soluble GFP, Gβ1, Gγ2 and AC-RLuc.
Kir3 channels can also be co-immunoprecipitated with Gβγ from rat atrial tissue (Nikolov and Ivanova-Nikolova, 2004), and several studies with purified proteins demonstrate direct interactions between Gβγ and Kir3 channels (Huang et al., 1997; Huang et al., 1995; Robillard et al., 2000). Kir3.1-RLuc was precipitated by anti-GFP antibodies from cells co-expressing GFP-Gγ2 in CHAPS-containing buffers (Table 1). There was an increase in the amount of Kir3.1-RLuc that could be precipitated with GFP-Gγ2 if the cells also expressed other components of M2 muscarinic receptor signaling complexes, including the receptor, complementary G-protein subunits, and the Kir3.4 channel subunit. This effect is presumably a consequence of a stoichiometry that favors complex formation and stability. Experiments also demonstrated that HA-tagged Kir3.1 could be precipitated with anti-FLAG antibodies from cells co-expressing FLAG-tagged Gβ1 (Fig. 3B). However, Kir3.1-RLuc was not immunoprecipitated with anti-GFP antibodies from cells co-expressing GFP-Gβ1, even in the presence of exogenously expressed Gγ2. These particular tags may reduce the affinity of Kir3.1 for Gβ1 so that they can no longer be co-immunoprecipitated when dissolved in a mixture of NP-40 and sodium deoxycholate. A salient point here is that detergent may be crucial, as Kir3.1-RLuc was co-immunoprecipitated with GFP-Gγ2 from solutions containing CHAPS (Table 1).
Non-specific protein interactions were not detected by BRET
To gauge whether BRET occurs as the result of non-specific interactions between tagged proteins, we used a number of negative controls. In some cases, cells were transfected with the cardiac voltage-gated K+ channel, KvLQT1, tagged with RLuc (KvLQT1-RLuc) and β2AR-GFP. This combination of proteins was chosen because KvLQT1 and β2AR have previously been tagged with fluorescent donor and acceptor proteins for fluorescent resonance energy transfer (FRET) experiments. Although they co-localize in cell membranes they do not normally associate, but nevertheless, at high expression levels FRET was observed in cardiomyocytes (Dilly et al., 2004). However, in our studies, BRET did not occur between either CD4-Rluc (subtracted to obtain net BRET values) or KvLQT1-RLuc and β2AR-GFP when they were expressed at the same levels as comparably tagged G-protein subunits and effectors (data not shown). Although KvLQT1 does not interact physiologically with the β2AR, it does interact with the HERG K+ channel subunit (Ehrlich et al., 2004). To confirm that the RLuc tag did not interfere with the ability of KvLQT1 to form a complex with HERG, we demonstrated that BRET occurs when KvLQT1 tagged with RLuc is co-expressed with GFP-tagged HERG (data not shown). Throughout this report, additional control fusion proteins CD8- and CD4-Rluc were used to confirm that the conditions we used did not produce non-specific BRET signals.
Interactions between Gβγ subunits and AC
BRET between GFP-Gγ2 and AC-RLuc was examined in the presence of various combinations of other signal transduction proteins (Fig. 4A). The expression levels of AC-RLuc and GFP-Gγ2, as determined by measuring RLuc activity and GFP fluorescence, respectively, were not markedly affected by co-expression of any or all of the other proteins (data not shown). Co-expression of AC-RLuc and GFP-Gγ2, led to BRET that was increased when both exogenous Gαs and Gβ1 were co-expressed (Fig. 4A).
Peptide fragments of YFP corresponding to amino acids 1-158 (YFP1-158) and 159-238 (YFP159-238) are not fluorescent when expressed alone, or when co-expressed. However, in experiments originally described by Kerppola and colleagues (Hu et al., 2002; Hu and Kerppola, 2003), it was demonstrated that if the two fragments can be brought together by fusing them to proteins that do associate, a functional YFP can be reconstituted in a process known as bimolecular fluorescence complementation (BiFC). This occurs when Gβ1 with an N-terminal YFP1-158 tag (YFP1-158-Gβ1) is co-expressed with Gγ2 having an N-terminal YFP159-238 tag (YFP159-238-Gγ2) (Hynes et al., 2004). The presence of AC, Gβ1 and Gγ2 in a single complex was demonstrated by showing that co-expressed YFP1-158-Gβ1 and YFP159-238-Gγ2 produced a fluorescent YFP-tagged Gβ1γ2 heterodimer that served as BRET partner for AC-RLuc (Fig. 4B).
Isoproterenol, acting through endogenous HEK 293 β-adrenergic receptors, tended to increase BRET between AC-RLuc and GFP-Gγ2, but the increase was only significant (134±7%, n=13) when exogenous β2AR was also co-expressed (see Fig. 4A,C, Fig. 7A and Fig. 8). By contrast, BRET between AC-RLuc and GFP-Gγ2 was not affected by the β-adrenergic antagonist propranolol (data not shown). BRET between AC-RLuc and the YFP-tagged Gβ1γ2 heterodimer produced by co-expressing YFP1-158-Gβ1 and YFP159-238-Gγ2 also increased in response to isoproterenol (Fig. 4B).
When expression levels of β2AR, AC-RLuc, Gαs and Gβ1 were kept constant while increasing GFP-Gγ2, the BRET signal increased approaching a maximum value asymptotically (Fig. 4C). The resulting dose-response curves approached higher maximal BRET values when the cells were treated with isoproterenol, but the amount of expressed GFP-Gγ2 needed to produce half-maximal BRET values (BRET50) was the same in both cases (i.e. the affinity of the interaction has not changed). The fact that the BRET between AC-RLuc and GFP-Gγ2 reached a plateau value as the amounts of expressed GFP-Gγ2 increased indicates a specific rather than non-specific protein-protein interaction. Additional evidence that the procedures used for these experiments did not produce non-specific BRET was obtained by expressing equivalent amounts of CD8-RLuc in place of AC-RLuc. CD8-RLuc is a fusion protein with a peptide sequence that makes it an integral membrane protein, and it exhibited a similar pattern of distribution to that of AC-RLuc (data not shown). Like the RLuc-tagged effectors used in our studies, the luciferase moiety of CD8-RLuc is located on the cytosolic side of the plasma membrane (Galés et al., 2005). Nevertheless, there was no detectable BRET between CD8-RLuc and GFP-Gγ2 at the expression levels used in our experiments (Fig. 4C).
Interactions between Gβγ subunits and Kir3.1
BRET was next used to investigate the interactions of Kir3.1 with G-protein subunits in living cells. BRET occurred in cells expressing Kir3.1-RLuc and either GFP-Gβ1 or GFP-Gγ2, and was increased by co-expression of Gγ2 or Gβ1, respectively (Fig. 5A). The ability of complementary untagged G-protein subunits to increase BRET between its GFP-tagged counterpart and Kir3.1-RLuc is an indication that stoichiometry is important in the formation of these complexes. The specificity of the interaction between GFP-Gβ1 and Kir3.1-RLuc was also investigated. Experiments showed that co-expression of wild-type Kir3.1 or Kir3.4 caused a dose-dependent inhibition of BRET between Kir3.1-RLuc and GFP-Gβ1 (Fig. 5B). Measurements of RLuc activity and GFP fluorescence revealed that the exogenously expressed Kir3.1 did not suppress the expression of the tagged proteins (data not shown). Thus, it is likely that the inhibition is due to Kir3.1 and/or Kir3.4 competing with Kir3.1-RLuc for GFP-Gβ1, again suggesting that BRET resulted from specific protein-protein interactions. The simultaneous presence of three proteins in a complex was again demonstrated by showing that co-expressed YFP1-158-Gβ1 and YFP159-238-Gγ2 served as an acceptor for Kir3.1-RLuc in BRET experiments (Fig. 5C). The occurrence of BiFC-BRET depended upon both YFP1-158-Gβ1 and YFP159-238-Gγ2 being co-expressed with Kir3.1-RLuc.
BRET saturation experiments performed in cells expressing increasing levels of GFP-Gγ2 and fixed amounts of β2AR, Gαs, Gβ1 and Kir3.1-RLuc showed that BRET also increased hyperbolically (Fig. 6A), characteristic of a specific protein-protein interaction. When the experiments were repeated using comparable amounts of CD8-RLuc in place of Kir3.1-RLuc there was a small increase in BRET indicating that these experimental conditions produced a weak, but nevertheless detectable non-specific interaction between the tagged proteins (Fig. 6A). The data in Fig. 6A were generated using cells expressing RLuc-tagged proteins at levels ∼tenfold higher than those used to generate the experimental data shown in Fig. 4C (hence the tenfold difference in the values for fluorescence/luminescence on the x axis). This higher level of expression led to a modest bystander BRET (Mercier et al., 2002) in cells co-expressing CD8-RLuc and GFP-Gγ2.
Previous studies have demonstrated that Kir-channel activity can be regulated by β-adrenergic receptors (Robillard et al., 2000). Although the Kir3.1-channel subunit forms a complex with G proteins, it is not transported to the cell surface unless co-expressed with Kir3.4 (Fig. 6B,C). Therefore, it is unlikely that BRET between tagged Kir3.1 and G-protein subunits would be affected by the membrane-impermeable β-adrenergic agonist isoproterenol in the same way that a cell surface complex containing AC and Gβγ would (Fig. 4). In fact, isoproterenol had no affect on BRET between Kir3.1-RLuc and either GFP-Gβ1 (data not shown) or GFP-Gγ2 (Fig. 6A,D) in the absence of the Kir3.4 subunit. However, a membrane-permeable β-adrenergic agonist, cimaterol, did cause an increase in BRET between Kir3.1-RLuc and GFP-Gβ1, which was blocked by propranolol (Fig. 6B). Data in this set of experiments were presented as agonist-stimulated BRET, over and above the constitutive BRET signals seen in Fig. 5 and Fig. 6A. Thus, in addition to Kir3.1 becoming associated with its cognate G protein before reaching the cell surface, the interaction also becomes sensitive to an agonist-occupied receptor. Co-expressing Kir3.4 allows Kir3.1 transportation to the plasma membrane (Fig. 6C). Since Kir3.4 (which also contains binding sites for Gβγ) can compete with Kir3.1-RLuc for binding to Gβγ, it is able to inhibit BRET between Kir3.1-RLuc and GFP-Gγ2 (Fig. 5B). However, if Kir3.4 was expressed at a level sufficient to allow Kir3.1-RLuc to be transported to the plasma membrane while only moderately attenuating BRET between Kir3.1-RLuc and GFP-Gγ2, then isoproterenol was able to increase BRET over constitutive levels between these tagged signaling proteins (Fig. 6D).
Effects of agonist-mediated signal transduction on BRET between G-protein α-subunits and effectors
BRET was also used to determine if Gα could be detected in complexes with effectors. These experiments revealed that co-expressed AC-RLuc and Gαs-GFP are found constitutively as complexes in HEK 293 cells (Fig. 7A) and that they can be co-immunoprecipitated (see legend to Table 1). BRET experiments also showed that Kir3.1-GFP was associated with Gαi-RLuc (Fig. 7B). In contrast to results obtained between Gβγ and AC, BRET between AC-RLuc and Gαs-GFP was not significantly affected by isoproterenol (Fig. 7A). Similarly, unlike BRET between Kir3.1-RLuc and GFP-Gγ2, there was again no effect of isoproterenol on the BRET between Kir3.1-GFP and Gαi-RLuc even though Kir3.4 was co-expressed so that the BRET pair was transported to the cell surface (Fig. 7B).
Specificity of G-protein-effector interactions
Finally, in an effort to thoroughly scrutinize the specificity of the interaction between AC-RLuc and GFP-Gγ2, the expression levels of these proteins were varied while keeping the ratio of donor- to acceptor-tagged proteins fixed (Kenworthy and Edidin, 1998). The amounts of co-expressed β2AR, Gαs and Gβ1 were also varied as described in the Materials and Methods. Expression levels for AC-RLuc and GFP-Gγ2 ranged over nearly two orders of magnitude (Fig. 8, inset) revealing that BRET between AC-RLuc and GFP-Gγ2 persisted even at expression levels near the limits of detection by the instrument (Fig. 8).
BRET was observed between heterotrimeric G proteins and two of the effectors that they regulate, AC and the Kir3.1 subunit. BRET between effectors and either tagged Gβ or Gγ subunits was increased by co-expression of untagged Gγ2 or Gβ1, respectively. These results corroborate data from functional and co-immunoprecipitation studies highlighting the importance of co-expressing stoichiometric amounts of multiple signaling proteins even when studying individual interactions between two partners. Furthermore, the ability of complementary wild-type G-protein subunits to augment BRET between their GFP-tagged counterparts and AC-RLuc or Kir3.1-RLuc indicates that there is an interdependence, and consequently a specificity associated with these interactions. Similarly, the fact that untagged Kir3.1 or Kir3.4 attenuated BRET between Kir3.1-RLuc and GFP-Gβ1 provides another indication that the interaction of the K+ channel with G-protein subunits is specific. For comparison, β2AR-GFP was co-expressed with KvLQT1-RLuc and CD8-RLuc was co-expressed with GFP-Gγ2. Although these proteins co-localize in the plasma membrane they do not interact, and there was little or no BRET between them, indicating that non-specific interactions were not a significant problem in our studies. The specificity of these interactions was also indicated by the finding that BRET between AC-RLuc and GFP-Gγ2 persisted at low protein expression levels (Kenworthy and Edidin, 1998). In addition, co-expression of both Gαs and Gβ1 significantly increased BRET between AC-RLuc and GFP-Gγ2. The presence of exogenously expressed Gα, Gγ2 or Gβ1 probably increases the cellular complement of GFP-tagged heterotrimeric G proteins that is available to interact with AC-RLuc, and provides indirect evidence that all three G-protein subunits are associated with AC even in the absence of signal transduction. BRET experiments provided direct evidence that Gα subunits were associated with both AC and Kir3.1, and combining BiFC with BRET provided a way of showing that AC and Kir3.1 formed complexes with the Gβ1γ2 heterodimer. BRET between GFP-tagged Gβ1 or Gγ2 or YFP-tagged Gβ1γ2 and Kir3.1-RLuc or AC-RLuc increased during signal transduction whereas BRET between the tagged Gα subunits and effectors was unaffected. These observations are consistent with the heterotrimeric G protein being persistently associated with effectors whereas the agonist-induced changes in BRET indicate conformational perturbations that accompany activation of the G protein.
The increase in BRET between AC-RLuc and GFP-Gγ2 caused by treating cells with the agonist isoproterenol may be attributable to (1) a change in the number of protein complexes due to a change in affinity, and/or (2) a change in the relative orientation of the donor and acceptor owing to a change in conformation. To identify the basis of isoproterenol-induced changes in BRET between AC-RLuc and GFP-Gγ2, experiments were conducted in which the expression of the donor RLuc-tagged protein was maintained at a constant level while that of the acceptor GFP-tagged protein was varied. If the BRET is saturable (i.e. approaches a maximum value asymptotically as the concentration of acceptor-tagged protein increases) a BRET50 can be determined. Since BRET50 is proportional to KD (Mercier et al., 2002), some conclusion can be made regarding the affinity of the two partners in the presence and absence of isoproterenol. The BRET50 for cells co-expressing AC-RLuc and GFP-Gγ2 was the same whether or not the cells were treated with isoproterenol, suggesting that activation of the G protein did not change the affinity of these proteins for one another. Although the agonist did not shift the BRET50, there was an increase in the maximum BRET. These curves were well fitted by single exponential functions suggesting that they report on interactions with one and the same equilibrium constant. Thus, a change in the plateau value, which represents the acceptor-saturated state, with no change in the affinity provides additional evidence that there is an agonist-induced conformational change within the complex rather than a change in the number of complexes.
Data presented here and elsewhere indicate that whether or not signaling is activated, receptors are associated with effectors (Davare et al., 2001; Kitano et al., 2003; Lavine et al., 2002; Liu et al., 2004), and that G-protein subunits remain associated with each other (Bünemann et al., 2003; Evanko et al., 2005; Frank et al., 2005; Ganpat et al., 2000), with their receptors (Galés et al., 2005; Lachance et al., 1999) and with their effectors. These data suggest that assemblies containing these components, and perhaps many additional proteins needed for G-protein-mediated signaling (Rebois and Hébert, 2003), may represent a common organizational theme needed to ensure the specificity and efficacy of signal transduction. This raises the question of where these complexes are assembled. Kir3.1 is not transported to the plasma membrane unless a heterologous Kir3 subunit (e.g. Kir3.2 or Kir3.4) is also expressed (Ma et al., 2002). Nevertheless, BRET was observed between Kir3.1-RLuc and GFP-Gγ2, Kir3.1-RLuc and GFP-Gβ1, Kir3.1-RLuc and YFP-Gβ1γ2, and Gαi-RLuc and Kir3.1-GFP in the absence of a co-expressed Kir3 targeting subunit, suggesting that Kir3.1 and Gαiβ1γ2 form a complex before reaching the plasma membrane. This, together with evidence that β2AR and Kir3.1 channel subunits also form complexes before reaching the plasma membrane (Lavine et al., 2002) suggest that receptors, G-protein heterotrimers and effectors initially come together inside the cell. Furthermore, the hypothesis that these intracellular assemblies are at least partially functional, and contain heterotrimeric G proteins that can be activated is supported by the finding that the membrane-permeable agonist, cimaterol, invoked a change in the BRET signal between Kir3.1-RLuc and GFP-Gγ2. The membrane-impermeable agonist isoproterenol caused a similar change in the BRET, but only if the cells co-expressed Kir3.4 targeting the complex to the cell surface.
In addition to the arguments for agonist-induced conformational changes in stable G-protein-effector complexes, we have reported receptor-mediated changes in the receptor-G-protein interaction (Galés et al., 2005) but not in the receptor/effector interaction (Lavine et al., 2002) under similar experimental conditions. The simplest interpretation is that the receptor, G protein and effector remain closely associated as a complex during signal transduction and changes in BRET reflect changes in protein conformation. These observations are supported by recent studies demonstrating that G-protein heterotrimers remain associated with Kir3 channels even after channel activation (Clancy et al., 2005).
If heterotrimeric G proteins are assembled at the same time that receptors and effectors are incorporated into a signaling complex, this might explain how G proteins with specific Gβγ subtype combinations are produced. Although Gα subunits, and effectors for that matter, do not seem to discriminate with regard to the subtype composition of the Gβγ heterodimer they interact with, receptors apparently have certain requirements in this regard (for reviews, see Rebois and Hébert, 2003; Robishaw and Berlot, 2004). The finding that β-adrenergic receptor subtypes show selectivity for Gβγ heterodimers of a particular subtype composition in regulating Kir3 channels (Robillard et al., 2000) may be considered as corroborating evidence. Consequently, the presence of a receptor during assembly of signaling complexes may be responsible for dictating the subtype composition of the heterotrimeric G proteins associated with it. Put another way, the receptor acts as a scaffold for a particular signaling complex. We have recently demonstrated that RGS2 also stably interacts with Gs-AC signaling complexes (Roy et al., 2006). Given the large number of other proteins that are likely to be involved in the formation, trafficking, regulation and maintenance of G-protein-mediated signal transduction pathways (Rebois and Hébert, 2003) it seems inevitable that additional components will be identified as part of these signaling complexes as well.
Materials and Methods
The preparation of recombinant plasmids encoding rat type II adenylyl cyclase and the human inwardly rectifying K+ channel subunit Kir3.1, each with a C-terminal Renilla reniformus luciferase tag (AC-RLuc and Kir3.1-RLuc, respectively), as well as HA-tagged Kir3.1 have been described (Lavine et al., 2002). The construction of recombinant plasmids coding for the rat stimulatory heterotrimeric G-protein α-subunit (Gαs), the rat inhibitory heterotrimeric G-protein α-subunit (Gαi2), the bovine heterotrimeric G-protein β1 subunit (Gβ1), and the human inwardly rectifying potassium channel subunit Kir3.4 have also been reported (Lavine et al., 2002). The plasmid encoding Gαs with the enhanced variant of GFP (EGFP) inserted into its α-helical domain was a generous gift of M. Rasenick (University of Illinois, Chicago, IL). Plasmids encoding bovine Gγ2 with either GFP2 (Lavoie et al., 2002) or EGFP fused to its N-terminus (GFP-Gγ2) were constructed. For Gαi1-RLuc, the RLuc moiety was introduced between L91 and K92 in the loop connecting helices A and B, analogous to a YFP fusion protein previously described (Bünemann et al., 2003). To construct a plasmid coding for the C-terminus of the voltage-dependent K+ channel, KvLQT1 fused to RLuc, the stop codon was removed from the KvLQT1 cDNA (generous gift of M. Sanguinetti, University of Utah, Salt Lake City, UT) and replaced with restriction site for NheI. An existing NheI site upstream of the sequence for KvLQT1 allowed the full-length channel cDNA to be inserted in-frame with RLuc. Membrane localization of the resulting fusion protein was confirmed by confocal microscopy (data not shown). A plasmid encoding Gβ1 with RLuc fused to its N-terminus (RLuc-Gβ1) was generated by transferring the cDNA for human Gβ1 from pcDNA3.1-Gβ1 (UMR cDNA Resource, Rolla, MO) into the humanized pRluc-C1 vector cut with restriction enzymes HindIII and XbaI (PerkinElmer Life Sciences). A construct encoding Gβ1 with EGFP fused to its N-terminus (GFP-Gβ1) (Doering et al., 2004) was a generous gift from G. Zamponi (University of Calgary, Calgary, AB). Plasmids encoding YFP1-158 fused to the N-terminus of human Gβ1 (YFP1-158-Gβ1), and YFP159-238 fused to the N-terminus of human Gγ2 (YFP159-238-Gγ2), were generous gifts of C. Berlot (Geisinger Institute, Danville, PA). The HindIII-BamHI fragment of pcDNA-CD8-βARK-CT (Crespo et al., 1995) encoding the extracellular and transmembrane domain of the CD8 lymphocyte-specific receptor (from codon 1 to 209) was subcloned into the HindIII-BamHI sites of the humanized pRLuc-N1 vector (PerkinElmer Life Sciences). CD4 tagged with RLuc was obtained from J. Stankova at the Université de Sherbrooke (Sherbrooke, QC). The cDNA for the human β2-adrenergic receptor (β2AR) was obtained from the American Type Culture Collection (Manassas, VA) and subcloned into pcDNA 3.1 by Varitas. The cDNA for the human M2 muscarinic receptor and FLAG-tagged Gβ1 were obtained from the UMR cDNA resource center (Rolla, MO). The cell line stably expressing HA-tagged β2AR (5 pmol/mg membrane protein) was generated as described (Azzi et al., 2003). β2AR-GFP and β2AR-RLuc constructs, respectively, were as previously described (Angers et al., 2000).
Protein expression in HEK 293 cells
HEK 293 cells were plated at a density of 105 cells/cm2 in six- or 24-well polylysine-coated plates. Approximately 24 hours later, cells were transfected using Lipofectamine Plus (Invitrogen) as described (Lavine et al., 2002). Unless otherwise indicated, 24-well plates received 0.4 μg plasmid DNA/well (0.2 μg/well for the expression of AC) and 6-well plates received 1.0 μg plasmid DNA/well (0.1 μg/well for the expression of AC). For some experiments, the expression levels of both AC-RLuc and GFP-Gγ2 were varied without changing their molar ratio. Although the amount of recombinant plasmids used for transfections varied, the total amount of DNA was kept constant by adding vector DNA. Assessing the expression of GFP-Gγ2 by fluorescence microscopy revealed that transfection efficiency was >70%. Varying the amount of plasmid used for transfections affected the level of protein expression in individual cells, but not overall transfection efficiency. Although there was a direct correlation between the amount of plasmid transfected and the amount of protein expressed, it was not possible to express AC-RLuc and GFP-Gγ2 in a fixed ratio by simply varying plasmid concentrations while keeping the ratio of the plasmids for these proteins constant (i.e. the correlation between plasmid concentration and protein expression was linear for both AC-RLuc and GFP-Gγ2, but the slopes of the lines relating the amount of plasmid used for transfection versus the amount of expressed protein differed). Consequently, plasmid concentrations for AC-RLuc and GFP-Gγ2 were adjusted independently (0.3-40 ng of plasmid for AC-RLuc and 5-300 ng for GFP-Gγ2) to achieve a fixed ratio of expressed proteins. The levels of co-expressed β2AR, Gαs and Gβ1 were also varied by maintaining the amounts of their plasmids in proportion to that for GFP-Gγ2. At 24 hours after transfection, media was replaced with DMEM containing 10% fetal bovine serum and cells were used for experiments approximately 48 hours after transfection.
Functional assessment of fusion proteins in oocytes and HEK 293 cells
cDNA constructs for Kir3.2, Gβ1, Gγ2, GFP-Gγ2, FLAG-Gβ1, GFP-Gβ1 and RLuc-Gβ1 were linearized with restriction enzymes, and purified using Geneclean (Bio 101). RNA synthesis and Xenopus oocyte isolation, injection and electrophysiological recordings were performed as described (Robillard et al., 2000). Standard recording solution was KD-98 (98 mM KCl, 1 mM MgCl2, 5 mM K-HEPES, pH 7.5). To verify the function of Gαi-RLuc, we co-expressed this cDNA or pcDNA vector control with human M2 muscarinic receptor and Kir3.2 in HEK 293 cells. At 3 days after transfection, currents were recorded using whole-cell patch clamp, as previously described, before and after stimulation with 1 μM carbachol (Sigma) (Weerapura et al., 2002). Where necessary, data was collected and analyzed with pCLAMP v.7 and Clampfit (Axon Instruments). cAMP determinations in HEK 293 cells were performed as described (Lavine et al., 2002). Briefly, cells in 24-well tissue culture plates were incubated at 37°C for 15 minutes (a convenient period of time during which cAMP production was constant) in serum-free medium containing 1 mM isobutylmethylxanthine in the absence or presence of 10 μM (-)isoproterenol, and cellular cAMP levels were measured by radioimmune assay.
Estimating GFP-Gγ2 and AC-RLuc expression
Protein expression levels for AC-RLuc and GFP-Gγ2 were quantified by measuring RLuc activity and GFP fluorescence, respectively. For some calculations, these values were converted to moles of protein as follows: specific [125I]CYP binding to HEK 293 cells expressing either β2AR-RLuc or β2AR-GFP was measured. The use of [125I]CYP for β2AR binding assured that total receptor levels would be detected and sham transfected cells were used to determine the background for [125I]CYP binding. RLuc activity or GFP fluorescence measurements were made using the same cells so that a ratio of RLuc activity/mole of β2AR-RLuc or GFP fluorescence/mole of β2AR-GFP could be calculated. These ratios were used to convert RLuc activity and GFP fluorescence in cells expressing AC-RLuc and GFP-Gγ2 to moles of AC and Gγ2 protein, respectively. At the lowest expression levels, it was estimated that the cells produced 2500 fmol of GFP-Gγ2 and 50 fmol of AC-RLuc/mg of total cell protein.
Preparation of cells for immunoprecipitation and BRET assays
The culture media was removed from transfected HEK 293 cell and replaced with a solution containing trypsin/EDTA in Hank's Balanced Salt Solution lacking Ca2+ and Mg2+ (Invitrogen). After incubating for 10 minutes at 37°C, an equal volume of culture medium containing 10% FBS and 0.1% soybean trypsin inhibitor was added, the cells were harvested, and collected by centrifugation at 110 g for 5 minutes at room temperature. Cells were resuspended in Dulbecco's phosphate-buffered saline (DPBS) containing protease inhibitors (Roche), 1 mg glucose/ml and 1 mM ascorbate. The cell suspension was assayed for protein using Bio-Rad Protein Assay Reagent. This procedure reproducibly yielded preparations where cell viability was >95% as assessed by Trypan Blue staining.
For some experiments, western blotting was used to determine if RLuc-tagged proteins were co-immunoprecipitated with GFP-tagged proteins. HEK 293 cells expressing Gβ1-GFP together with Gγ2 and AC-RLuc were washed with ice-cold phosphate-buffered saline (PBS) and resuspended in 5 mM Tris, pH 7.4, 2 mM EDTA, 5 μg/ml leupeptin, 10 μg/ml benzamidine, and 5 μg/ml soybean trypsin inhibitor. Cells were homogenized with a Polytron and the membranes dissolved with RIPA buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.5, 1 mM EDTA pH 8.0, 1% NP-40, 0.5% sodium deoxycholate, 5 μg/ml aprotinin, 2 μg/ml leupeptin, and 1 mM PMSF). The detergent extracts were centrifuged to remove insoluble material, and stored at –80°C until required. Immunoprecipitation was accomplished by adding 0.2 μg of anti-GFP antibody (Clontech) to 450 μl of clarified detergent extract, and incubated with shaking for 16 hours at 4°C. Antibody complexes were precipitated with 100 μl of a 50% slurry of anti-mouse IgG agarose (Sigma), washed in RIPA buffer, and resuspended in SDS-PAGE sample buffer. Following electrophoresis, proteins were transferred to nitrocellulose membranes (Bio-Rad) and co-precipitated RLuc-tagged proteins detected with anti-luciferase antibodies (1:1000; Chemicon). In some cases, FLAG-tagged Gβ1 was co-expressed with HA-tagged Kir3.1. Co-immunoprecipitations were performed as described above using beads conjugated to mouse monoclonal anti-FLAG M2 antibody (Sigma, as per manufacturer's instructions). Immunoblots were probed with either a monoclonal anti-HA antibody (Covance, 1:1000) or with a polyclonal anti-FLAG antibody (Sigma, 1:1000). HRP-conjugated secondary antibodies were from Sigma (anti-mouse or anti-rabbit, 1:20,000). For quantitative immunoprecipitation, samples of intact cells containing 50 μg of protein were collected by centrifugation at 110 g for 5 minutes at room temperature, and the pellet resuspended in 250 μl of Solution A (20 mM HEPES, 2 mM MgCl2, 1 mM EDTA and 0.1 mM DTT with protease inhibitors, pH 8.0) containing 1% CHAPS (Sigma). After incubating at 4°C for 20 minutes, an equal volume of Solution A was added and the samples were centrifuged at 100,000 g for 1 hour. The supernatants from each sample received 1 μl of polyclonal anti-GFP antibody (BD Biosciences), and were incubated for 1 hour at room temperature before adding 40 μl of a 50% slurry of Protein A-Sepharose (Zymed) in DPBS. Following an additional incubation for 1 hour at room temperature with vigorous shaking, beads were washed several times with DPBS. Luciferase activity was assayed with a Turner Design 20/20 Luminometer using 5 μM coelenterazine H as a substrate. Luciferase activity remaining in the detergent extract was also assayed to determine percentage of total RLuc activity immunoprecipitated.
For cells expressing EGFP- and YFP-tagged proteins, BRET was assessed by measuring the ratio of light passed by ≥480 nm filters to that passed by a 450/58 nm filter with coelenterazine H (Molecular Probes) as a substrate for RLuc. For cells expressing proteins fused to GFP2, the substrate for RLuc was DeepBlueC (PerkinElmer), and BRET was assayed using 515/30 and 410/80 nm filters. Samples of cell suspensions containing 25-100 μg of protein were dispensed into 96-well white OptiPlates (PerkinElmer). The assay was initiated by adding RLuc substrate followed by the addition of vehicle, isoproterenol, cimaterol and/or propranolol. For some experiments, BRET was assayed before and after the addition of ligand to cells. The sample volume in each well was 100 μl and the final concentrations of RLuc substrate and ligands were 5 μM and 1 μM, respectively. BRET background was determined under conditions where resonance energy transfer between RLuc and GFP either could not, or did not, occur. This was accomplished by expressing RLuc or RLuc-tagged proteins either alone or together with GFP or GFP-tagged proteins, none of which interact physiologically. The background was the same regardless of which of the aforementioned individual proteins or combinations of proteins were expressed, and it has been subtracted to yield net BRET.
Previous studies have shown that a GFP-tagged β2AR is both biologically active and forms a complex that gives BRET with co-expressed β2AR-RLuc (Angers et al., 2000), as well as with Kir3.1-RLuc or AC-RLuc (Lavine et al., 2002). Therefore, co-expression of β2AR-GFP together with β2AR-RLuc, Kir3.1-RLuc or AC-RLuc were used as positive controls in BRET experiments (data not shown). In some experiments, fusion proteins of YFP with RLuc also served as a positive BRET control.
At 24 hours after transfecting HEK 293 cells with Kir3.1-RLuc and/or FLAG-tagged Kir3.4, they were seeded on glass coverslips that had been treated with laminin (10 μg/ml). At 4 hours after seeding, the cells were fixed with 3% paraformaldehyde for 20 minutes, followed by three washes of 5 minutes with PBS, and incubation with 2% normal donkey serum (NDS; Jackson Laboratories) to block non-specific binding. The cells were permeabilized with 0.2% Triton X-100 for 1 hour in an incubation chamber, followed by an overnight incubation at 4°C with the primary antibodies (1:200 dilution of rabbit anti-FLAG, Sigma; and/or 1:100 dilution of mouse anti-luciferase, Chemicon International). Primary antibodies were removed by three washes of 5 minutes with PBS followed by incubation for 45 minutes at room temperature with a 1:500 dilution of a secondary antibody (goat anti-rabbit conjugated with Alexa Fluor 647 and/or goat anti-mouse conjugated with Alexa Fluor 488; Molecular Probes). Both primary and secondary antibodies were diluted in 1% NDS and 0.04% Triton X-100. Cells were examined on an inverted laser-scanning microscope (LCM 510; Zeiss).
This work was supported in part by grants from the Heart and Stroke Foundation of Quebec and the Canadian Institutes of Health Research (to T.E.H. and M.B.). This work was also supported in part with funds provided by the Intramural Research Programs of the National Institute of Neurological Disorder and Stroke (NINDS) and the National Institute of Deafness and Communicative Disorders (NIDCD). The authors thank C. Berlot, M. Sanguinetti, J. Stankova and G. Zamponi for recombinant plasmids, and L. R. Villeneuve for assistance with confocal microscopy. T.E.H. is a Senior Scholar of the Fonds de la Recherche en Santé du Québec (FRSQ). M.B. is holder of a Canada Research Chair in Molecular Pharmacology and Signal Transduction. C.G. is holder of an INSERM postdoctoral fellowship. D.J.D. is holder of a Heart and Stroke Foundation of Canada postdoctoral fellowship.