Functional asymmetry of G‐protein‐coupled receptor (GPCR) dimers has been reported for an increasing number of cases, but the molecular architecture of signalling units associated to these dimers remains unclear. Here, we characterized the molecular complex of the melatonin MT1 receptor, which directly and constitutively couples to Gi proteins and the regulator of G‐protein signalling (RGS) 20. The molecular organization of the ternary MT1/Gi/RGS20 complex was monitored in its basal and activated state by bioluminescence resonance energy transfer between probes inserted at multiple sites of the complex. On the basis of the reported crystal structures of Gi and the RGS domain, we propose a model wherein one Gi and one RGS20 protein bind to separate protomers of MT1 dimers in a pre‐associated complex that rearranges upon agonist activation. This model was further validated with MT1/MT2 heterodimers. Collectively, our data extend the concept of asymmetry within GPCR dimers, reinforce the notion of receptor specificity for RGS proteins and highlight the advantage of GPCRs organized as dimers in which each protomer fulfils its specific task by binding to different GPCR‐interacting proteins.
Heptahelical G‐protein‐coupled receptors (GPCRs) represent the largest family of membrane receptors with approximately 800 members in human beings responding to a wide range of extracellular stimuli. This family of receptors controls numerous processes including neuro‐transmission, cellular metabolism, inflammatory and immune responses. GPCRs are of primary therapeutic importance as they are the targets of 30–50% of currently prescribed drugs. Although GPCRs have been classically described as monomeric receptors that form a ternary complex with its ligand and heterotrimeric G protein, cumulative evidence indicates that GPCRs dimerize or oligomerize (Bouvier, 2001; George et al, 2002; Milligan, 2009). Despite the availability of high‐resolution crystal structures for several GPCRs, little is known about the stoichiometry of receptor oligomers and receptor‐associated proteins. It is now well established that monomeric GPCRs are capable of activating G proteins (Bayburt et al, 2007; Ernst et al, 2007; Whorton et al, 2007), implying a 1:1 stoichiometry between receptor and G protein in this minimal signalling unit. By extrapolation, one might anticipate that GPCR dimers, composed of two protomers, are binding to two G‐protein units. However, no experimental evidence for such an arrangement exists. In contrast, recent research suggests that the leukotriene B4 receptor BLT1 dimer, the 5‐HT2C receptor and the dopamine D2 receptor dimer interact with a single G protein (Baneres et al, 2003; Herrick‐Davis et al, 2005; Han et al, 2009). Allosteric regulation between protomers of GPCR dimers has been documented for class A and class C members (Pin et al, 2005; Sohy et al, 2007; Vilardaga et al, 2008; Han et al, 2009). The transactivation model in which the ligand binds to one protomer, whereas the second receptor protomer binds to the G protein is fully compatible with the one GPCR dimer/one G‐protein stoichiometry.
A functional significance for GPCR dimerization has been proposed in several cases. This is particularly evident for the case of GPCR heterodimers in which new pharmacological entities are formed. The benefit of homodimer formation is more difficult to appreciate. A function of GPCR dimerization was proposed in receptor export to the plasma membrane (Milligan, 2010) and various, positive and negative, allosteric transactivation modes have been proposed to occur between the two protomers of the dimer (Rovira et al, 2010). Recently, the study of Rivero‐Muller et al has provided compelling in vivo evidence for the physiological relevance of GPCR dimerization by restoring the normal luteinizing hormone (LH) actions in transgenic mice co‐expressing a binding deficient and a signalling deficient form of LH receptor through functional complementation in the absence of functional wild‐type receptors (Rivero‐Müller et al, 2010).
The classical collision‐based model predicting the recruitment of heterotrimeric G proteins to agonist‐activated receptors followed by the rapid dissociation of the Gα and Gβγ into free subunits, was recently challenged by the observation of stable pre‐associated receptor–G‐protein complexes that persist during the activation process (Bunemann et al, 2003; Galés et al, 2005). Galés et al (2006) proposed a model wherein agonist binding induces conformational rearrangements of a pre‐existing receptor–G‐protein complex, allowing the Gα–Gβγ interface to open and to allow GDP exit from the Gα subunit.
Application of proteomic approaches to the GPCR field showed that GPCRs, in addition to G proteins, can interact with multiple intracellular regulatory proteins further extending the questions of the stoichiometry and architecture of these signalling complexes (Bockaert et al, 2004; Daulat et al, 2009; Ritter and Hall, 2009). Regulators of G‐protein signalling (RGS) are GTPase‐activating proteins (GAPs) that bind to the activated form of Gα and accelerate its GTPase activity, thereby modulating G‐protein signalling (Neitzel and Hepler, 2006; Xie and Palmer, 2007). RGS proteins are a family of highly diverse, multifunctional signalling proteins that share a conserved 120–130 amino‐acid core domain (RGS domain), which is responsible for the GAP activity. Several observations indicate that RGS proteins regulate G‐protein‐dependent signalling not only in a G‐protein‐specific manner, but also in a receptor‐specific manner (Zeng et al, 1998; Xu et al, 1999; Wang et al, 2002; Hague et al, 2005). There is increasing evidence for the existence of GPCR protein complexes containing RGS and G proteins (Bernstein et al, 2004; Benians et al, 2005; Hague et al, 2005; Wang et al, 2005; Abramow‐Newerly et al, 2006; Neitzel and Hepler, 2006). However, the molecular architecture and stoichiometry of these complexes are poorly understood.
Here, we used the melatonin MT1 receptor, a typical Gi‐coupled receptor (Jockers et al, 2008), as a model to study the molecular architecture of the complex composed of MT1 homodimers, RGS20 and Gαiβγ proteins. By searching for specific interacting partners of the MT1 carboxyl‐terminal domain, we previously pulled‐down RGS20 from mouse brain lysates (Maurice et al, 2008). Interaction between full‐length MT1 and RGS20 was subsequently confirmed in the pituitary pars tuberalis (Maurice et al, 2008). Here, biochemical, bioluminescence resonance energy transfer (BRET) and electrophysiological approaches were applied to characterize the complex in its basal and agonist‐activated state and reveal functional asymmetry of RGS20 and Gi coupling to MT1 receptors.
RGS20 is part of the pre‐existing MT1 receptor protein complex
To characterize the interaction between MT1 and RGS20, we performed co‐immunoprecipitation experiments by co‐expressing Flag‐MT1 or Myc‐MT2 with HA‐tagged RGS20 or RGS10 in HEK293T cells (Figure 1A and B). In accordance with previous results (Maurice et al, 2008), Flag‐MT1 interacted with HA‐RGS20, but not with the related HA‐RGS10 in both resting and melatonin (MLT)‐activated cells. The second melatonin receptor subtype, MT2, did not bind to HA‐RGS10, whereas low levels of binding were detected for HA‐RGS20 (Figure 1A and B). The interaction between MT1 and RGS20 was further confirmed by BRET in intact HEK293T cells. BRET donor saturation curves were generated by co‐expressing constant amounts of the previously described MT1–Rluc fusion protein (energy donor) (Ayoub et al, 2002) and increasing quantities of C‐terminally YFP‐tagged RGS20 (RGS20‐YFP, energy acceptor) (Figure 1D). A specific interaction between these two proteins was indicated by the hyperbolic and saturable behaviour of the BRET donor saturation curve. Stimulation of cells with MLT did not alter BRET signals. Similar results were obtained with a N‐terminally YFP‐tagged RGS20 (YFP‐RGS20) construct (not shown). In contrast, co‐expression of MT1–Rluc with RGS10‐YFP expressed at similar levels to those of RGS20‐YFP showed non‐specific BRET signals (Figure 1D). Similar negative results were obtained with an RGS7‐YFP construct (not shown). Taken together, these data show that RGS20 is part of a pre‐existing MT1‐associated protein complex for which no further RGS20 recruitment is observed upon MLT stimulation.
RGS20 regulates MT1‐dependent Kir3 channel activation
To establish the functional significance of the MT1/RGS20 interaction, we studied the effect of MT1 and RGS20 on the activation of Kir3‐type K+ channels (also designated as GIRKs), which is mediated by the Gβγ subunits of the activated G proteins (Jiang et al, 1995; Nelson et al, 1996). Application of MLT to MT1/Kir3.1/3.2‐expressing CHO cells evoked Ba2+‐sensitive outward currents reverting at potentials close to K+ equilibrium potential (−101.5±1.7 mV, n=4), indicating the activation of Kir3 channels (Dascal, 1997) (Figure 2A and B). RGS20 significantly accelerated the onset of MLT‐induced K+‐current responses, similarly to RGS10 that does not bind to MT1, but activates Gαi proteins (Figure 2A and E). Similarly, the latency between MLT application and signal onset was shortened in cells expressing RGS10 or RGS20 (Supplementary data 1). These kinetic effects are expected to result from the RGS‐mediated increase in GTPase activity of Gαi and/or from RGS GAP activity‐independent enhancement of coupling between MT1 and Kir3 channels (Doupnik et al, 1997; Jeong and Ikeda, 2001). After fast removal of MLT, Kir3 channels deactivated with strikingly slow time course, most likely reflecting the slow dissociation of MLT from the receptor (Figure 2B and F). RGS10, but not RGS20, significantly decreased the half‐decay time of Kir3 channels upon MLT withdrawal (Figure 2B and F) similar to the previously reported effect of RGS4 on MT1 desensitization (Witt‐Enderby et al, 2004). In agreement with our data, differential effects of RGS proteins on activation and deactivation kinetics have been reported earlier for several RGS–GPCR couples (Benians et al, 2005). To verify that the preventive effect of RGS20 on Kir3 channel deactivation is specific to MT1 and is not an intrinsic property of RGS20, we studied the activation of Kir3 channels in CHO cells stably expressing the Gi protein‐coupled MT2 receptor. In contrast to MT1, RGS10 and RGS20 had identical effects on the onset of MLT‐induced K+‐current responses and the half‐time of channel deactivation in these cells (Figure 2C, E and F) indicating that the effect of RGS20 on channel deactivation is specific for MT1. Altogether, these results show that MT1 activates Kir3 channels and that activation kinetics are accelerated by RGS20 consolidating the functional interaction between RGS20 and MT1. Differences observed between RGS10 and RGS20 in the half‐decay time of Kir3 channel deactivation are most likely due to differences between the direct (RGS20) and indirect (RGS10), through Gαi proteins, coupling to MT1.
RGS20 binds directly to the Cter and i3 loop of MT1
RGS20 and MT1 are known to interact with Gαi subunits (Wang et al, 1998; Brydon et al, 1999). To determine whether RGS20 binds directly or indirectly (through Gαi) to MT1, we performed pull‐down experiments using a chemically synthesized His6‐tagged peptide encompassing the entire MT1‐Cter that was immobilized on Ni‐NTA agarose beads. The corresponding MT2‐Cter peptide was used as a negative control. As the third intracellular (i3) loop has been shown to be involved in RGS binding for several other GPCRs (Bernstein et al, 2004; Hague et al, 2005; Georgoussi et al, 2006), we also included the MT1‐i3 loop in our study. As shown in Figure 3A, purified HA‐RGS20 specifically binds to the MT1‐Cter and ‐i3 loop, but only marginally to the MT2‐Cter. The existence of additional binding sites for RGS20 within MT1 (i3 loop) is consistent with co‐immunoprecipitation experiments using a Cter deletion mutant of MT1, which precipitated less, but still significant amounts of HA‐RGS20 compared with full‐length MT1 (Figure 1C).
To identify the minimal‐binding motif within the MT1‐Cter, we first designed three truncated MT1‐Cter peptides based on its predicted secondary structure (Figure 3B). Only the full‐length peptide was able to efficiently pull‐down purified HA‐RGS20, suggesting that the membrane‐proximal 18 amino acids of the MT1‐Cter, corresponding to the predicted helix 8 (H8), are necessary for HA‐RGS20 binding. This hypothesis was confirmed with a synthetic peptide corresponding to H8, which retained similar amounts of purified HA‐RGS20 as the full‐length peptide (Figure 3B). To determine RGS domains involved in MT1 binding, we tested the binding capacity of the C‐terminal part, encompassing the RGS box (RGS20‐box), and the N‐terminal part (RGS20‐Nter), which was shown to be important for GPCR binding of other RGS proteins (Zeng et al, 1998; Bernstein et al, 2004; Hague et al, 2005; Leontiadis et al, 2009). Whereas the RGS20‐Nter interacted in pull‐down experiments with immobilized His6‐tagged MT1‐Cter and the MT1‐i3 loop peptides, the RGS20‐box bound only to the MT1‐Cter (Figure 3C). Neither the RGS20‐Nter nor the RGS20‐box bound to the truncated MT1‐Cter peptides (Supplementary data 2). Taken together, these results show that RGS20 directly interacts by its Nter domain with the MT1‐i3 loop and by its Nter and Cter domains with the membrane‐proximal H8 of the MT1‐Cter.
Molecular dynamics of the MT1/RGS20/Gi ternary complex monitored by BRET
These results show that RGS20 directly and constitutively interacts with at least two intracellular domains of MT1. Our previous studies have shown that Gi proteins are also constitutively pre‐coupled to MT1 (Roka et al, 1999; Guillaume et al, 2008; Maurice et al, 2008), raising the question of the molecular organization of RGS20 and Gi in the MT1 complex. We performed BRET experiments to study the molecular proximity of RGS20 and Gi proteins in the presence of MT1 in intact cells. HEK293 cells stably expressing MT1 were co‐transfected with previously described Gαi1–Rluc fusion proteins (Galés et al, 2006) for which the BRET energy donor Rluc was inserted at two different positions within Gαi1 (Gαi1–91‐Rluc, Gαi1–122‐Rluc, see Figure 6 for structural details of insertion sites) and with N‐ or C‐terminally YFP‐tagged RGS20 fusion proteins (Figure 4A). High BRET signals were detected for all Gαi1/RGS20 combinations, whereas low BRET signals were obtained with non‐fused YFP, expressed at similar levels, defining the background signal of the assay. Differences in BRET signals at equivalent expression levels of different Gαi1–Rluc and YFP‐tagged RGS20 fusion proteins are consistent with the different insertion positions of the Rluc and YFP molecular probes, representing different degrees of molecular proximity. To determine the fraction of the BRET signal that is receptor dependent, we pre‐treated cells with pertussis toxin (PTX) at 10 ng/ml, a concentration that abolished the inhibitory effect of MLT on the adenylyl cyclase pathway (Supplementary data 3). ADP‐ribosylation of Gαi subunits by PTX uncouples Gαi proteins from GPCRs and thus is expected to move Gαi away from MT1‐bound RGS20 proteins. PTX treatment indeed decreased BRET signals for all Gαi/RGS20 combinations, thus revealing the component of the RGS20/Gαi interaction that is receptor dependent.
To study the effect of MLT stimulation of MT1 on the molecular proximity of RGS20 and Gαi1 within the complex, BRET donor saturation curves were performed. Cells stably expressing, or not, MT1 were co‐transfected with increasing quantities of RGS20‐YFP and a fixed amount of Gαi1–122‐Rluc, a fusion protein previously shown to be sensitive to receptor stimulation (Galés et al, 2006). A BRET signal was detected in the absence of MT1 (BRETmax=105±8, BRET50=0.88±0.20) (Figure 4B). This signal was insensitive to MLT and PTX incubation (data not shown). In cells expressing MT1, and the same amount of Gαi1–122‐Rluc, the BRET signal was significantly increased (BRETmax=140±6) and the BRET50 decreased (BRET50= 0.11+0.02), indicating an increased propensity of interaction between RGS20 and Gαi in the presence of MT1 (Figure 4B and C). Data presented in Figure 4C are expressed as percentage of BRETmax to better illustrate the shift in BRET50. Stimulation with MLT further increased the BRETmax (BRETmax=192±7) between Gαi1–122‐Rluc and RGS20‐YFP without affecting the BRET50 (BRET50=0.09±0.02) (Figure 4B and C). As previously shown (Mercier et al, 2002; Couturier and Jockers, 2003), a change in the BRETmax without a shift in BRET50 upon agonist stimulation is indicative of conformational rearrangements between interacting proteins, Gαi1 and RGS20 in our case, and not of additional recruitment. Importantly, in cells expressing similar levels of the Gi‐coupled MT2 receptor, the BRET signal between Gαi1–122‐Rluc and RGS20‐YFP was not increased upon MLT stimulation (Figure 4D), showing specificity towards the receptor. The MLT‐promoted BRET observed in MT1‐expressing cells was dose dependent (Figure 4E) with an EC50 in the subnanomolar range (0.28 nM) that is in agreement with the IC50 value of MLT for the inhibition of the adenylyl cyclase pathway (Petit et al, 1999). As expected, the MT1 antagonist luzindole did not increase the basal BRET signal, showing that the MLT‐induced BRET correlates with receptor activation (Supplementary data 4). The BRET signal was specific for the Gαi protein as no MLT‐induced BRET was observed between RGS20‐YFP and a Gαq–Rluc fusion protein in MT1‐expressing cells (Supplementary data 5A). The functionality of this Gq probe was assessed in an assay in which angiotensin II induced a BRET decrease in cells co‐expressing Gαq–97‐Rluc and GFP10‐Gγ2 in the presence of the angiotensin II (AT1) receptor (Supplementary data 5B).
These data show that RGS20 and Gαi are part of a constitutive protein complex organized around MT1 and subjected to conformational rearrangements after MT1 stimulation. To further characterize these molecular rearrangements, we inserted molecular Rluc and YFP probes at different sites of the MT1/RGS20/Gi ternary complex and monitored MLT‐induced BRET signals. There were no detectable rearrangement between MT1–Rluc and YFP‐RGS20 or RGS20‐YFP (Figure 5A) nor intra‐molecular rearrangements within RGS20 as monitored by Rluc–RGS20‐YFP and YFP‐RGS20–Rluc fusion proteins in cells expressing MT1 (Figure 5B and C). The relative movement between RGS20 and Gαi1 upon agonist stimulation was detected in cells co‐expressing either RGS20‐YFP or YFP‐RGS20 and Gαi1–122‐Rluc or Gαi1–91‐Rluc constructs (Figure 5D and E). The interaction between RGS20 and Gαi1–Rluc fusion proteins was confirmed by co‐immunoprecipitation experiments in cells expressing MT1 (Figure 5F). MLT stimulation did not increase the amount of precipitated Gαi1—Rluc, indicating that the MLT‐promoted BRET signal shown in Figure 5E corresponds to molecular arrangements within the complex itself rather than additional recruitment. Movement of positions 91 and 122 within the Gi protein is in agreement with the recently described activatory switch of the α‐helical domain of the GTPase domain of Gα (Galés et al, 2006). Furthermore, MLT stimulation of cells co‐expressing MT1, Gαi1–122‐Rluc and the previously described YFP‐Gγ2 fusion protein led to the expected decrease of the BRET signal most likely reflecting the movement of the α‐helical domain of Gαi away from the N‐ter of Gγ (Figure 5G). A similar decrease was observed in the presence of RGS20, indicating that binding of RGS20 to MT1 does not interfere with this activatory switch of Gαi.
RGS20 and Gi bind to separate receptor protomers in the MT1/RGS20/Gi complex
Our biochemical and BRET experiments suggest the formation of a constitutive ternary MT1/RGS20/Gi complex in the basal state with direct interactions of RGS20 and Gi with the MT1 receptor. The high‐resolution crystal structure of several GPCRs and Gi proteins have been solved and a molecular model of the relative position of these two proteins has been proposed based on extensive biochemical evidence (Lambright et al, 1996; Liang et al, 2003; Fotiadis et al, 2004; Preininger and Hamm, 2004; Rosenbaum et al, 2009). Although the crystal structure of RGS20 is currently unknown, the structure of the RGS domain has been solved for several other RGS proteins and appeared to be highly conserved (Tesmer et al, 1997; deAlba et al, 1999; Soundararajan et al, 2008). In addition to the RGS domain, RGS20 contains a large N‐terminal domain of similar size. Considering the spatial constrains imposed by the simultaneous presence of Gαi, Gβ and Gγ subunits in proximity to one receptor protomer, direct binding of RGS20 to the same receptor protomer is difficult to reconcile because of the limited space of the intracellular interface of GPCRs (approximately 45 Å diameter). We, therefore, propose that Gi and RGS20 are binding to two different protomers of a receptor dimer (Figure 6). Agonist activation of MT1 decreased BRET values between Gαi1–122‐Rluc and YFP‐Gγ2 (Figure 5G) in agreement with the previous proposed movement driving Gα and Gβγ apart to allow GDP release from the Gα subunit after G‐protein activation (Galés et al, 2006). At the same time, MLT‐induced BRET signals are increased between Gαi1–122‐Rluc (and Gαi1–91‐Rluc) and YFP‐tagged RGS20 fusion proteins (Figure 5D and E). The simplest explanation for this is a movement of the α‐helical domain of Gα towards the N‐ter of RGS20. Similar results were obtained with both N‐terminally and C‐terminally YFP‐tagged RGS20, suggesting that both extremities are most likely oriented in the same direction. Figure 6 summarizes our model, based on a GPCR dimer that positions RGS20 close to the α‐helical domain of Gαi and opposite to Gβγ to explain the observed MLT‐induced BRET changes.
Asymmetric interaction of RGS20 and Gi proteins with MT1/MT2 heterodimers
To consolidate our prediction that RGS20 and Gi are binding to two separate receptor protomers, we studied RGS20 and Gi binding to the previously described MT1/MT2 heterodimer (Ayoub et al, 2004). These heterodimers constitute an interesting model as both MLT receptors are Gi‐coupled, but only MT1 strongly binds to RGS20 as shown in Figure 1A. Accordingly, RGS20 was only weakly co‐immunoprecipitated when expressed with MT2 alone, but readily co‐immunoprecipitated by MT2 in the presence of MT1, indicating that RGS20 is part of the MT1/MT2 heterodimer‐associated protein complex (Figure 7A).
We then studied the proximity between RGS20 and Gαi in the presence of the MT1/MT2 heterodimer using BRET. To monitor MLT‐induced rearrangements between RGS20 and Gαi exclusively in protein complexes associated with the MT1/MT2 heterodimer, we replaced wild‐type MT1 by the previously described MT1 A252C/G258T mutant (MT1**), which is devoid of MLT binding (Gubitz and Reppert, 2000). In this configuration, confounding effects of co‐expressed MT1 homodimers on ligand‐induced rearrangements between RGS20 and Gαi are excluded, as MT1 cannot be activated by MLT. Absence of MLT binding and proper cell surface expression of MT1** was confirmed by 2‐[125I]‐MLT binding and immunofluorescence microscopy, respectively (Supplementary data 6A and B). Co‐immunoprecipitation experiments showed that MT1** still heterodimerizes with MT2 and binds to RGS20 (Supplementary data 6C and D). Co‐expression of RGS20‐YFP and Gαi1–122‐Rluc in cells expressing MT2/MT1** heterodimers resulted in a BRET signal that was markedly increased in the presence of MLT, in contrast to cells expressing MT2 alone, which is consistent with the specific association of RGS20 and Gαi1 to the MT2/MT1** heterodimer (Figure 7B). In cells expressing MT1 alone, a similar pattern was observed, although the basal BRET level was higher than in cells expressing either MT2 homodimers or MT2/MT1** heterodimers. PTX pre‐treatment decreased basal and MLT‐induced BRET signals in cells expressing MT1 or MT2/MT1**, indicating that BRET signals are mainly dependent on the functional interaction of Gαi with receptor homo‐ and heterodimers. Co‐immunoprecipitation experiments showed that the amount of RGS20 bound to the MT1**/MT2 heterodimer was not modified by agonist stimulation, indicating that the agonist‐induced BRET is due to conformational changes within a pre‐existing MT1**/MT2/RGS20/Gi complex, as observed for the MT1/RGS20/Gi complex (Figure 7C). Collectively, our BRET and biochemical data are compatible with the formation of a pre‐existing MT1**/MT2/RGS20/Gi quaternary complex, wherein the MT2 protomer can cis‐activate its Gαi protein and subsequently increase the proximity of Gαi to RGS20 associated to the MT1** protomer.
To verify whether the recruitment of RGS20 to the MT1**/MT2 heterodimer is able to recapitulate the functional consequences of RGS20 expression seen on wild‐type MT1 homodimers (see Figure 2), we expressed the MT1** mutant in CHO cells stably expressing MT2. Introduction of MT1** indeed converted the profile of deactivation seen in cells expressing MT2 alone (see Figure 2C and E) into an MT1‐like profile in which only RGS10, but not RGS20, decreased the half‐decay time of Kir3 channel activation (Figure 2D and F). The effects of RGS10 and RGS20 on rise time and latency of signal onset were all similar to those seen in cells expressing MT1 and MT2 alone confirming the functional expression of RGS protein in the MT1**/MT2 context (Figure 2E; Supplementary data 1B). This shows that the majority of MT2 is engaged into MT1**/MT2 heterodimers and that the functional effect of RGS20 on Kir3 channel activation by MT1**/MT2 heterodimers is similar to that observed for MT1 homodimers, but different from MT2 homodimers.
To further explore the functional consequences of RGS20 binding to MT1, we measured [35S]GTPγS incorporation in CHO cell membranes expressing MT1, MT2 or MT1**/MT2 in the absence or presence of purified HA‐RGS20. Stimulation with MLT induced the expected two‐ to three‐fold increase in [35S]GTPγS binding in all three cell types (not shown). Addition of purified HA‐RGS20 into the assay further increased [35S]GTPγS binding in MT1‐expressing membranes, but not in MT2‐expressing membranes (Figure 7D). Importantly, co‐expression of the MT1** mutant in MT2‐expressing cells also enhanced [35S]GTPγS binding, indicating functional coupling of RGS20 proteins to MT1**/MT2 heterodimers. Altogether, these data support the functional importance of RGS20 in the MT1**/MT2/RGS20/Gi quaternary complex.
By using MT1 homodimers and MT1/MT2 heterodimers as model GPCRs, we are extending here two emerging concepts: the pre‐assembly of GPCR‐interacting complexes and the asymmetric function and organization of GPCR dimers. In addition, we are providing a new functional justification for GPCR dimerization that applies to homo‐ and heterodimers, namely the possibility of simultaneous and direct binding of GPCR‐interacting proteins (GIPs) to the same GPCR dimer composed of two asymmetric protomers. Heterotrimeric G proteins are central, although not exclusive signal transducers of GPCRs. An increasing number of reports suggests the formation of pre‐assembled receptor–G‐protein complexes, which rearrange upon agonist activation of the receptor (Bunemann et al, 2003; Galés et al, 2006; Audet et al, 2008). This central complex is surrounded by a number of other GIPs that might either compete with the G protein for receptor binding, as in the case of arrestin (Lohse et al, 1990), or simultaneously bind to the receptor as shown for the multi‐PDZ domain protein MUPP1 (Guillaume et al, 2008). Formation of such complexes raises the question of their molecular organization. Although the crystal structure of the GPCR–G‐protein complex has not been solved, molecular modelling of the receptor–G‐protein interaction based on high‐resolution crystal structures of receptors and G proteins alone and biochemical data indicate that the intracellular receptor surface is covered by the heterotrimeric G protein (Arimoto et al, 2001; Hamm, 2001; Liang et al, 2003; Fotiadis et al, 2004) (see also Figure 6). Consequently, small surface accessibility appears difficult to accommodate with simultaneous binding of other GIPs, particularly those that bind to membrane‐proximal receptor domains. Formation of receptor signalling units containing distinct GIPs is one possibility to solve this problem. Our data suggest that there might exist at least one other alternative, namely preferential binding of GIPs to the different protomers of GPCR dimers. Given the fact that most GPCRs dimerize, such architecture is likely to be of general importance for GPCR function. This asymmetric model can accommodate binding of GIPs that must be located in the same complex such as G proteins and RGS proteins. The increasing number of examples reporting functional asymmetry of class A and C GPCR dimers is fully compatible with the notion that receptor protomers interact with different GIPs. Several modes of functional cross‐talk have been proposed including positive and negative allosteric trans‐ and cis‐activation within dimers (Pin et al, 2005; Sohy et al, 2007; Vilardaga et al, 2008; Han et al, 2009). In the case of cis‐activation (ligand binding and G‐protein activation by the same protomer), the function of the second protomer has remained elusive so far (Damian et al, 2008). Our model suggests that the second protomer might function as scaffold to preferentially target GIPs to the receptor dimer. The possible extension of our model towards GPCR oligomers adds further complexity and needs to be explored in the future.
As shown here for the MT1/RGS20 and MT1/MT2/RGS20 complexes, a similar binding mode is likely to occur for several other receptor‐RGS pairs. Several studies have reported a direct interaction between RGS proteins and GPCRs (Neitzel and Hepler, 2006). Other studies, which did not address the question of the direct interaction between RGS and receptors, reported implied GPCR–RGS interactions irrespective of the activation state of the receptor (reviewed in Neitzel and Hepler, 2006). The existence of preformed scaffolding complexes containing RGS proteins, G proteins and GPCRs might explain the frequently reported receptor selectivity of RGS proteins.
Preferential asymmetric binding of GIPs to GPCR dimers is fully compatible with the recently proposed model of agonist‐induced conformational reorganization within pre‐existing receptor–G‐protein complexes, reflecting most likely the proposed movement between Gα and Gβγ required to allow GDP release from the Gα subunit (Galés et al, 2006). Indeed, BRET between YFP fusion proteins of RGS20 and Gαi1–91‐Rluc/Gαi1–122‐Rluc was sensitive to agonist stimulation (movement of α‐helical domain at positions 91 and 122 of Gαi1). Insertion of BRET probes at multiple other sites of the complex did not reveal further molecular rearrangements, suggesting a well defined and spatially limited activation mechanism initiated at the level of the ligand‐activated receptor and transmitted through the G protein to RGS20. Studies with the MT1**/MT2 heterodimer containing a ligand‐binding‐deficient MT1 mutant showed that activation of only one receptor protomer is sufficient for G protein cis activation, indicating that the function of the MT2 protomer within the heterodimer is to promote ligand and G‐protein binding, whereas the function of the MT1 protomer is to provide a preferential anchoring site for RGS20. Although we cannot completely exclude the possibility that RGS20 binds directly to MT2 in the MT1**/MT2 heterodimer, this possibility seems unlikely because of the high intrinsic affinity of RGS20 to MT1 and the MT1‐like signalling profile of MT1**/MT2 heterodimers in functional assays.
Our model predicts a binding mode between RGS20 and Gi in a pre‐existing complex with a receptor that is different from the spatial organization of the previously reported RGS/Gαi complex (Tesmer et al, 1997). The latter complex forms only between the AlF−4 activated Gαi subunit and the RGS domain through a binding site partially overlapping with the Gβγ‐binding site of Gαi. In contrast, binding in the presence of the receptor occurs in the presence of Gβγ and most likely close to the α‐helical domain of Gαi and opposite to Gβγ. This model has obviously to be confirmed in further studies, but raises the interesting possibility of a new regulatory mode of Gi by RGS proteins within receptor complexes. Whether MT1 activation promotes or inhibits RGS20 function in the complex, as might be suggested by the absence of RGS20 activity upon Kir3 channel deactivation, remains to be shown.
In conclusion, we propose a new model highlighting the advantage of GPCRs to be organized as dimers in which each protomer fulfils its specific task by preferentially binding to a specific GIP. This model is fully compatible with the emerging concept of asymmetry within GPCR dimers and may explain receptor specificity of proteins involved in GPCR signalling such as RGS proteins.
Materials and methods
cDNAs encoding human HA‐RGS20 and HA‐RGS10 were purchased from UMR cDNA Resource Center and pGEX‐4T‐1 vector from Amersham Pharmacia. The Flag‐MT1, Flag‐MT1ΔCter, Myc‐MT2, MT1–Rluc and MT2–Rluc constructs have been described elsewhere (Ayoub et al, 2002; Guillaume et al, 2008). The cDNAs encoding Gαi1 was kindly provided by Dr M Ayoub (Montpellier, France) (Ayoub et al, 2007) and Gαi1–91‐Rluc, Gαi1–122‐Rluc and YFP‐Gγ2 by Dr M Bouvier (Montreal, Canada) (Galés et al, 2006). RGS20‐YFP, YFP‐RGS20 and RGS10‐YFP constructs were obtained using the HA‐RGS20 or HA‐RGS10 cDNAs as template and the Phusion High‐Fidelity DNA Polymerase (Finnzymes). Restriction sites for EcoRV and XhoI (YFP‐RGS20) or HindIII and BamHI (RGS20‐YFP, RGS10‐YFP) were introduced by PCR. After digestion by the respective restriction enzymes, the resulting inserts were ligated into a pcDNA3.1 vector encoding the YFP. The Flag‐MT1 A252C/G258T mutant (MT1**) was obtained using Flag‐MT1 cDNA as template by PCR‐based site‐directed mutagenesis in two steps. Briefly, internal primers were used to generate Flag‐MT1 A252C, then Flag‐MT1 A252C/G258T. The YFP‐RGS20–Rluc and Rluc–RGS20‐YFP constructs were obtained using, respectively, YFP‐RGS20 and RGS20‐YFP cDNAs as template. Restriction sites for HindIII and ApaI (YFP‐RGS20–Rluc) or EcoRV and ApaI (Rluc–RGS20‐YFP) were introduced by PCR. After digestion by the respective restriction enzymes, the resulting inserts were ligated into a phRluc.N2 (YFP‐RGS20–Rluc) or phRluc.C2 (Rluc–RGS20‐YFP) plasmid. All DNA sequences were confirmed by sequencing.
Monoclonal anti‐HA and anti‐GFP antibodies were purchased from Roche Diagnostics. Monoclonal and polyclonal anti‐Flag antibodies were from Sigma. Monoclonal and polyclonal anti‐Myc and polyclonal anti‐Gαi3 antibodies were purchased from Santa Cruz Biotechnology and monoclonal anti‐Rluc from Chemicon.
Cell culture and transfections
Human embryonic kidney 293T (HEK293T) and HEK293 cells stably expressing MT1 or MT2 were cultured in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) foetal bovine serum, 100 units/ml penicillin, 0.1 mg/ml streptomycin, 0.02 M Hepes with, or not, 0.4 mg/ml geneticin, at 37°C in a humidified atmosphere at 95% air and 5% CO2. Transient transfections were performed with FuGENE6 (Roche Molecular Biochemicals) or JET‐PEI (Polyplus Transfection), according to the manufacturer's protocol. For electrophysiology experiments, CHO‐K1 cells stably expressing human MT1 or MT2 were maintained in Ham's F12 Glutamax supplemented with 10% foetal bovine serum, 100 units/ml penicillin, 0.1 mg/ml streptomycin and 0.4 mg/ml geneticin. Transient transfections of concatemers of Kir3.1/3.2 subunits, RGS20, RGS10 and Flag‐MT1** were performed with Lipofectamine 2000 (Invitrogen). For [35S]GTPγS‐binding studies, transient transfections of Flag‐MT1** were performed by electroporation using the Amaxa kit T™ (Lonza).
Expression and purification of recombinant proteins
To express HA‐RGS20 as a GST fusion protein, restriction sites for BamHI and XhoI were introduced immediately adjacent to the initiation and termination codon by PCR of the plasmid encoding HA‐RGS20 using the phusion high‐fidelity DNA polymerase (Finnzymes). For HA‐RGS20‐Nter, corresponding to amino‐acid (aa) 1–89 of RGS20 (whole sequence 216 aa), a restriction site for BamHI was introduced immediately adjacent to the initiation codon and for SalI after a stop codon by PCR of the HA‐RGS20 plasmid. For HA‐RGS20‐box, corresponding to aa 90–216 of RGS20, the RGS box was sub‐cloned in a plasmid encoding amino‐terminal HA tag then restriction sites for SalI and NotI were introduced immediately adjacent to the initiation and termination codon by PCR. After digestion by their respective restriction sites, the resulting inserts were ligated into pGEX‐4T‐1 vector. All DNA sequences were confirmed by sequencing. The resulting plasmids were transformed into BL21(DE3) cells (Invitrogen) and cells were grown in LB media containing 50 μg/ml ampicillin. Expression of GST fusion proteins was induced by adding 300 μM to 1 mM isopropylthiogalactoside (Sigma) to mid‐log cultures. Cultures were harvested after 5 to 7 h at 37°C and cells centrifuged at 6000 g for 15 min. The pellets were re‐suspended in cold PBS containing protease inhibitor cocktail EDTA free, crushed with an Ultra‐turrax T25 and sonicated. After adding 1% Triton‐X100 (Sigma), the cell suspensions were centrifuged at 10 000 g for 30 min, the supernatants collected and applied to glutathione‐agarose beads (Sigma) that had been equilibrated with 1% Triton‐X100 in PBS. After extensive washes, HA‐tagged proteins were eluted by 0.04 U/μl of thrombin (Amersham Biosciences) in PBS (2 h, 22°C). Eluates were applied to new freshly glutathione‐agarose beads to eliminate residual contaminant GST and bacterial proteins. Purity of HA‐tagged proteins was assessed by SDS–PAGE after silver staining.
Experiments with CHO‐K1 cells stably expressing MT1 or MT2 were performed at room temperature (23–24°C) 1–3 days after a transfection with Kir3.1/3.2, RGS20, RGS10 and Flag‐MT1**. Cells were continuously superfused with an extracellular solution composed of (in mM): 145 NaCl, 2.5 KCl, 1 MgCl2, 2 CaCl2, 10 HEPES, 25 glucose; pH 7.3, 323 mosm. Patch pipettes had resistances between 3–4 MΩ when filled with intracellular solution composed of (in mM) 107.5 potassium gluconate, 32.5 KCl, 10 HEPES, 5 EGTA, 4 MgATP, 0.6 NaGTP, 10 Tris phosphocreatine; pH 7.2, 297 mosm. Series resistance (<5 MΩ) was compensated by 80%. Melatonin receptor responses were evoked by fast application of MLT and recorded with an Axopatch 200B patch‐clamp amplifier; filtering and sampling frequencies were set to 1 and 5 kHz, respectively. The deactivation time course was often not exponential, but exhibited a slight hump that made exponential curve fitting imprecise. Therefore, we quantified the rate of deactivation by measuring the time for the current level at the end of MLT application to drop by 50% (half‐decay time). Recording of currents, curve fitting and further data analyses were performed with pClamp software.
Rluc‐ and YFP‐tagged protein constructs were transiently co‐transfected into HEK293T cells or HEK293 cells stably expressing MT1 or MT2, seeding in 12‐well plates. Twenty‐four hours after transfection, cells were trypsinated and 1 × 105 were distributed in a 96‐well white Optiplate (PerkinElmer Life Sciences) pre‐coated with 10 μg/ml poly‐l‐lysine (Sigma) for another 24 h. After two washes with PBS, coelenterazine h substrate (Molecular Probes) and MLT (or PBS) were added for 15 min at a final concentration of 5 and 1 μM, respectively. Light‐emission acquisition at 485 and 530 nm was then started. The BRET ratio was calculated as the ratio of the emission at 530 over 485 nm of co‐transfected 10 ng Rluc‐tagged protein and 12.5 to 500 ng YFP‐tagged protein. BRET signal values were then corrected by subtracting the background BRET signal detected when the Rluc‐tagged protein was expressed alone from the BRET signal detected in cells co‐expressing both Rluc‐ and YFP‐tagged constructs. Luminescence and fluorescence were measured simultaneously using the lumino/fluorometer MithrasTM (Berthold) that allows the sequential integration of luminescence signals detected with two filter settings (Rluc filter, 485±10 nm; YFP filter, 530±12.5 nm). Total fluorescence was measured with the fluorometer FusionTM (Packard Instrument Company). The results were expressed in milliBRET units, 1 milliBRET corresponding to the BRET ratio values multiplied by 1000. In some experiments, PTX (Alexis® Biochemicals) was used to discriminate between receptor‐dependent and ‐independent effects. Cells were incubated with 10 ng/ml PTX in serum‐free medium (5 h, 37°C) before BRET measurements.
HEK293T cells grown in six‐well plates were co‐transfected with plasmids encoding the indicated constructs. Forty‐eight hours after transfection, cells were stimulated, or not, with 1 μM MLT for 15 min at 37°C, washed two times in PBS, and lysed in 500 μl cold lysis buffer (75 mM Tris, 2 mM EDTA, 12 mM MgCl2, 10 mM CHAPS, protease inhibitor cocktail EDTA free, 10 mM NaF, 2 mM Na3VO4, pH 7.4). After sonication and solubilization during 3–5 h at 4°C under gentle end‐over‐end mixing, lysates were centrifuged at 12 000 g during 1 h at 4°C. Immunoprecipitations were performed using 2–4 μg of the indicated antibodies pre‐adsorbed on protein G sepharose beads (Sigma) for 2 h at 4°C. Immunoprecipitated proteins were eluted with Laemmli buffer and subjected to SDS–PAGE and immunoblotting. Immunoblottings were performed using the indicated antibodies and immunoreactivity was revealed using secondary antibodies coupled to 680 or 800 nm fluorophores using the Odyssey LI‐COR infrared fluorescent scanner (ScienceTec).
Peptides encompassing the His6‐MT1‐Cter, His6‐MT1‐i3 loop, His6‐MT2‐Cter (NeoMPS) and the His6‐truncated MT1‐Cter (Proteogenix) were chemically synthesized and purified by HPLC (>90% purity). A total of 35 nmol of purified peptides were immobilized on 20 μl Ni‐NTA agarose beads (Qiagen) and quantitative immobilization was monitored by measuring the absorbance of the supernatant as previously described (Maurice et al, 2008). Beads were incubated with 0.5 μg of purified HA‐tagged full‐length, Nter or Cter domain of RGS20 in 500 μl of binding buffer (20 mM NaH2PO4, 10 mM CHAPS, 150 mM NaCl, 2 mM Na3VO4, 10 mM NaF, protease inhibitor cocktail EDTA free, 100 μM GDP, 1 mM AEBSF, 20 mM imidazole, 0.05 % BSA, pH 8) for 2 h at 4°C with gentle shaking. The beads were washed three times with binding buffer, and recruited proteins were eluted with Leammli buffer and subjected to SDS–PAGE and immunoblotting.
CHO cells (2.106) stably expressing MT1 or MT2 were electroporated using the Amaxa kit T™ with 4 μg of Flag‐MT1** or empty vector (pcDNA3) cDNAs according to manufacturer's specifications. Forty‐eight hours after electroporation, [35S]GTPγS binding was determined from crude membranes in 100 μl of reaction mixture containing 20 mM HEPES (pH 7.4), 100 mM NaCl, 3 mM MgCl2, 20 μg/ml saponin, 3 μM GDP, 0.3 nM [35S]GTPγS and purified HA‐RGS20 (0.1 μM) with or without 1 μM melatonin. The reaction was started by transferring tubes at room temperature and stopped after 60 min incubation by addition of 1 ml of ice‐cold stop buffer containing 10 mM Tris–HCl (pH 8.0), 100 mM NaCl, 20 mM MgCl2, 0.1 mM GTP. Bound and free radioactivity was separated by filtration over GF/F glass fibre filters (Whatman).
GPCR, heterotrimeric G‐protein and RGS structure representation
Ribbon diagrams were generated using coordinates from Protein Data Bank files as indicated and visualized with CHIMERA software.
Data and statistical analysis
All data represent the mean±s.e.m. of three to five independent experiments for biochemical and BRET studies and 4 to 21 for electrophysiology. The results were analysed by PRISM (GraphPad Software Inc.) and statistical significance was assessed by two‐way Anova or Student's t tests for biochemical and BRET studies or ANOVA with the Dunnett's multiple comparison test for electrophysiology (*P< 0.05; **P<0.01; ***P<0.001).
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
This work was supported by grants from SERVIER, the Fondation Recherche Médicale (‘Equipe FRM’), Fondation pour la Recherche sur le Cerveau (FRC) Neurodon, Institut National de la Santé et de la Recherche Médicale (INSERM), Centre National de la Recherche Scientifique (CNRS). RT was supported by the Wellcome Trust International Senior Research Fellowship Award and by a grant from GACR (309/06/1304). BB was supported by the Swiss Science Foundation (3100A0‐117816). AMD held an EGID fellowship. We thank Dr Michel Bouvier (University of Montreal) for kindly providing the G‐protein Rluc and YFP fusion proteins, Dr Billy Breton (University of Montreal), Dr Julie Dam (Institut Cochin) and Patty Chen (Institut Cochin) for helpful discussions and comments.
- Copyright © 2010 European Molecular Biology Organization