Variable G protein determinants of GPCR coupling selectivity

Significance

G protein-coupled receptors (GPCRs) regulate a wide variety of important cellular processes and are targeted by a large fraction of approved drugs. GPCRs signal by activating heterotrimeric G proteins and must couple to a select subset of G proteins to produce appropriate intracellular responses. It is not known how GPCRs select G proteins, but it is generally accepted that the Gα subunit C terminus is the primary G protein determinant of coupling selectivity. We systematically studied coupling of GPCRs to four families of G proteins and chimeras with C terminal regions that were exchanged between families. We uncovered rules for coupling selectivity and found that different GPCRs can recognize different features of the same G protein for selective coupling.

Abstract

G protein-coupled receptors (GPCRs) activate four families of heterotrimeric G proteins, and individual receptors must select a subset of G proteins to produce appropriate cellular responses. Although the precise mechanisms of coupling selectivity are uncertain, the Gα subunit C terminus is widely believed to be the primary determinant recognized by cognate receptors. Here, we directly assess coupling between 14 representative GPCRs and 16 Gα subunits, including one wild-type Gα subunit from each of the four families and 12 chimeras with exchanged C termini. We use a sensitive bioluminescence resonance energy transfer (BRET) assay that provides control over both ligand and nucleotide binding, and allows direct comparison across G protein families. We find that the G_s- and G_q-coupled receptors we studied are relatively promiscuous and always couple to some extent to G_i1 heterotrimers. In contrast, G_i-coupled receptors are more selective. Our results with Gα subunit chimeras show that the Gα C terminus is important for coupling selectivity, but no more so than the Gα subunit core. The relative importance of the Gα subunit core and C terminus is highly variable and, for some receptors, the Gα core is more important for selective coupling than the C terminus. Our results suggest general rules for GPCR-G protein coupling and demonstrate that the critical G protein determinants of selectivity vary widely, even for different receptors that couple to the same G protein.

G protein coupled receptors (GPCRs) exert many of their physiological effects by coupling to and activating heterotrimeric G proteins. The 16 human Gα subunit genes are classified into four families (G_s, G_i/o, G_q/11, and G_12/13), and members of each family interact with different effector molecules to produce distinct cellular responses (1, 2). Individual cells generally express many different G proteins from multiple families, and, therefore, GPCRs must be able to selectively activate subsets of G proteins. Many receptors show a preference for G proteins from just one of the four families, although promiscuous coupling to G proteins from multiple families is also not uncommon (3, 4). The mechanisms of selective GPCR-G protein coupling are not completely understood, and the structural determinants of selectivity have not been determined with precision. Analysis of GPCR primary sequences has not revealed simple conserved motifs for coupling to different G protein families (5), and analysis of GPCR–G protein complex structures has only begun to uncover features that may be important for coupling selectivity (6–12). From the standpoint of the G protein, it has long been appreciated that the Gα carboxy (C) terminus is critical for activation by receptors and also a key determinant of selectivity (13–17). The distal part of the C-terminal alpha helix (helix 5; H5) is enveloped by GPCR transmembrane (TM) domains during coupling (6), and several studies have shown that mutations within this region can promote coupling of receptors to noncognate heterotrimers (17–21). Although other Gα regions have been shown to be important for recognition by individual GPCRs (22–25), the Gα C terminus is widely considered to be the most important structural determinant of coupling selectivity.

Here, we test this idea by measuring coupling of GPCRs to four representative Gα subunits (Gα_s, Gα_i1, Gα_q, and Gα₁₂) and 12 Gα chimeras with the C terminus of one representative and the Gα core region of another. We monitor coupling using a sensitive energy transfer assay that allows comparison of all GPCR–G protein combinations directly at the coupling step. Our results demonstrate unexpectedly promiscuous coupling of many GPCRs to wild-type G proteins and support the notion that the Gα C terminus is one important determinant of coupling selectivity for many receptors. However, our results also show that the Gα C terminus is a minor contributor to selective coupling of several other receptors. Taken together our findings suggest that GPCRs recognize widely distributed structural features of G proteins and are consistent with the suggestion that different GPCRs recognize different conserved selectivity determinants of a given G protein (5).

Results

Promiscuous Coupling to Wild-Type G Proteins.

Coupling of a GPCR to a heterotrimeric G protein can be defined as the allosteric interaction between the ligand binding site of the receptor and the nucleotide binding site of the associated Gα subunit, such that the agonist-bound state of the receptor promotes the nucleotide-free (empty) state of the G protein and vice versa (26, 27) (Fig. 1A). To quantify coupling we monitored agonist-dependent and nucleotide-sensitive association of GPCRs with different heterotrimers using bioluminescence resonance energy transfer (BRET) (28). To interfere as little as possible with receptor–G protein interactions, we positioned BRET donor and acceptor labels far from regions identified in functional and structural studies as important for coupling. A luciferase was fused to the GPCR C terminus, and complementary fragments of the fluorescent protein Venus were fused to Gβ₁ and Gγ₂ subunits. These components were transfected together with unlabeled Gα subunits into CRISPR/Cas9-edited HEK293 cells deficient in either Gα_s, Gα_q, and Gα₁₂ subunit families (three-family knockouts; 3GKO) or all four Gα families (four-family knockouts; 4GKO) (29). When 3GKO cells were used, the S1 subunit of pertussis toxin (PTX) was coexpressed to prevent coupling to endogenous Gα_i/o subunits, except when the overexpressed Gα subunit was itself sensitive to PTX. Cells were permeabilized with digitonin and either supplemented with GDP or treated with apyrase to remove residual nucleotides.

Fig. 1.

Coupling to wild-type heterotrimers in permeabilized, nucleotide-depleted cells. (A) Receptors were fused to Renilla luciferase (Rluc8), and Gβγ dimers were fused to Venus. (B) Time course of BRET between H₁R-Rluc8 and Gα_i1βγ-Venus in response to 100 μM histamine (HA) in permeabilized cells in the presence (GDP) and absence (apyrase) of nucleotides. (C and D) H₁R couples to G_s, G_i1, and G_q heterotrimers; **P = 0.0011, ****P < 0.0001, n.s. not significant; one-way ANOVA, Sidak’s multiple comparisons. Pertussis toxin S1 subunit (PTX) was coexpressed in experiments with Gα_s, Gα_q, and Gα₁₂. (E) Similar potency of histamine-induced coupling to G_i1 (EC₅₀ = 258 nM; n = 4) and G_q (EC₅₀ = 245 nM; n = 7) heterotrimers.

Complexes between agonist-bound GPCRs and G proteins are short-lived when guanine nucleotides are present at concentrations similar to those found in the cytosol of intact cells and are greatly stabilized when guanine nucleotides are absent (30). Accordingly, nucleotide depletion significantly enhanced the magnitude of agonist-induced BRET between receptors and G proteins (Fig. 1B), and maximal BRET was observed when agonist was present and GDP was absent (Fig. 1C). As a simple index of coupling, we defined ΔBRET_max as the difference between the BRET observed when receptor–G protein complexes are least stable (agonist absent, GDP present) and that observed when receptor–G protein complexes are most stable (agonist present, GDP absent). This experimental system allows direct comparison of receptor coupling to heterotrimers with Gα subunits from different families and is sufficiently sensitive to detect secondary coupling to nonpreferred G proteins. Using this assay, we found that receptors generally coupled to the same G proteins as indicated by other methods and as annotated in the IUPHAR/BPS Guide to Pharmacology (3, 31). However, many receptors coupled to wild-type G proteins more promiscuously than expected. For example, we found that H₁ histamine receptors (H₁R), which are generally classified as G_q-coupled (3), couple strongly to both G_q and G_i heterotrimers, couple weakly to G_s heterotrimers, and do not couple significantly to G₁₂ heterotrimers (Fig. 1D). The EC₅₀ values for H₁R coupling to G_q and G_i were comparable (Fig. 1E), suggesting that activation of G_i heterotrimers by this receptor may have physiological significance, as some studies have indicated (32). We carried out similar experiments on a panel of 19 receptors (SI Appendix, Table S1), 17 of which are classified in the IUPHAR/BPS Guide to Pharmacology (3) as coupling primarily to G_s, G_i, or G_q proteins, and two of which are unclassified orphan receptors that transduce signals through G_12/13 proteins (33, 34). We found that promiscuous coupling to more than one wild-type G protein was a universal feature of the G_s-, G_q- and G₁₂-coupled receptors that we studied. In contrast, the G_i-coupled receptors that we studied were much more selective for G_i heterotrimers (Fig. 2 and SI Appendix, Fig. S1). From the perspective of the G protein, we found that G_i heterotrimers coupled to all of the receptors that we studied. It should be noted that the assay we used for these studies reflects allosteric coupling between agonist and nucleotide binding sites, but this is only a minimum requirement for agonist-dependent G protein signaling. Additional factors may be required for efficient G protein activation and signaling under physiological conditions in cells (see below).

Fig. 2.

G_i-coupled receptors are relatively selective. Normalized ΔBRET_max for a panel of G_s-, G_i-, G_q-, and G₁₂-coupled receptors. Each value represents the mean ΔBRET_max (normalized to the value obtained using the presumed cognate G protein) from n = 3–6 independent experiments. Individual data points for each receptor are shown in SI Appendix, Fig. S1.

Coupling to G Protein C-Terminal Chimeras.

To assess the importance of the Gα C terminus for coupling selectivity, we constructed 12 Gα chimeras consisting of the main Gα subunit core [HN.1-H5.16 using the common G protein numbering (CGN) system] of Gα_s (long isoform), Gα_i1, Gα_q, or Gα₁₂ and the final 10 amino acids (H5.17–26) of each of the others. This portion of the Gα C terminus is deeply embedded in the TM domains of GPCRs in all receptor–G protein complex structures and has been called the interface module of the C-terminal alpha helix (helix 5 or H5) (35). Previous mutagenesis studies have primarily targeted residues within this region to alter coupling selectivity (36). This region is also unstructured in G protein crystal structures (35), suggesting that it is not involved in folding or stability of the Gα subunit core and, therefore, can be exchanged without compromising the structural integrity of heterotrimers. We confirmed the ability of Gα chimeras to form heterotrimers by monitoring sequestration of overexpressed Gβγ dimers (37), and all expressed at comparable levels as indicated by immunoprecipitation (SI Appendix, Fig. S2). Conditions were optimized to provide roughly equivalent stoichiometry for each receptor–G protein pair, with the latter in excess. For simplicity, we refer to heterotrimers that incorporate these chimeras as G_si, G_sq, G_s12. We generated full coupling profiles (4 wild-type Gα subunits and 12 chimeras) for 14 GPCRs and plotted the results as ΔBRET_max heat maps (Fig. 3). Similar heat maps were constructed for GDP-sensitive BRET in the absence of agonist (ΔBRET_GDP; SI Appendix, Fig. S3) and agonist-induced BRET in the presence of GDP (ΔBRET_ag; SI Appendix, Fig. S4).

Fig. 3.

Coupling to heterotrimers with C-terminal Gα subunit chimeras. Heat maps of ΔBRET_max (normalized to the largest value) for G_s-coupled (Top), G_i-coupled (second row), G_q-coupled (third row), and G₁₂-coupled (Bottom) receptors. Maps are arranged from left to right depending on coupling that is determined more (e.g., H₁R) or less (e.g., D₁R) by the Gα C terminus (H5.17–H5.26). Each value represents the mean normalized ΔBRET_max from n = 3–6 independent experiments.

In agreement with our results using wild-type Gα subunits, we found that G_s-, G_q- and G₁₂-coupled receptors coupled more promiscuously to G protein chimeras than G_i-coupled receptors. The G_s-, G_q- and G₁₂-coupled receptors that we studied always coupled well to several chimeras in addition to wild-type G proteins, sometimes with ΔBRET_max values that approached or even exceeded that of the cognate wild-type G protein (Fig. 3). The G_i-coupled receptors that we studied were more selective. For example, the D₂ dopamine receptor (D₂R) failed to couple to any Gα chimera with a ΔBRET_max value that reached half of that observed with wild-type Gα_i1. We also observed that the Gα₁₂ core was often nonpermissive for coupling, particularly for G_i- and G_q-coupled receptors. This was sometimes partially overcome when a cognate C terminus was present, e.g., for A₁R and H₁R. By comparison, the Gα₁₂ C terminus was more accommodating to most receptors (Fig. 3). This suggests that many receptors reject Gα₁₂ core regions rather than the C terminus to select against G₁₂ heterotrimers. The most promiscuous receptor that we studied was the endothelin A (ET_AR) receptor, which coupled well to Gα_i1, Gα_q, Gα₁₂, and eight of the nine chimeras bearing the C termini of these subunits. This receptor coupled poorly to Gα_s and chimeras bearing the Gα_s C terminus, suggesting that selection against G_s coupling for this receptor is determined primarily by the C terminus.

We anticipated that GPCRs would couple poorly to chimeras bearing noncognate C termini, i.e., adding a noncognate C terminus to a cognate Gα core would impair coupling. Indeed, this was observed for all 42 receptor-chimera pairs that we studied. However, the extent of impairment was highly variable, ranging from a near complete loss of coupling to almost no loss (SI Appendix, Fig. S5A). G_i-coupled receptors, in particular the M₄ acetylcholine receptor (M₄R) and D₂R, were the most sensitive to noncognate C termini. Conversely, we expected that GPCRs would generally couple well to chimeras bearing cognate C termini, i.e., noncognate Gα core regions would only modestly impair coupling. However, we found that a noncognate Gα core was often as detrimental to coupling as a noncognate C terminus (SI Appendix, Fig. S5B). Although adding a cognate C terminus to a noncognate Gα core usually enhanced coupling, the gain of coupling was again highly variable and rarely a complete “switch.” For most of the receptors that we studied, coupling selectivity was determined both by the Gα subunit core and the C terminus. A clear C terminus-dominant coupling pattern was observed for only a few receptors, the most obvious examples being H₁R, P2RY10, and GPR35 (Fig. 3). Other receptors, such as H₂R and A₁R, showed a more balanced influence of the cognate C terminus and G protein core. These receptors coupled equally well to chimeras bearing either the cognate C terminus or the cognate G protein core, although coupling to these chimeras was never as efficient as coupling to cognate wild-type G proteins. Unexpectedly, D₁R and M₃R showed coupling that was determined almost entirely by the G protein core and was virtually insensitive to the Gα C terminus (Fig. 3). Full concentration–response curves for these receptors confirmed the patterns observed with single saturating concentrations of agonists. Specifically, for the D₁R and M₃R agonist potency was greater for chimeras with cognate G protein cores, whereas for the H₁R agonist potency was greater for chimeras with cognate C termini (SI Appendix, Fig. S6 and Table S2).

Efficient coupling to G proteins with noncognate Gα C termini was unexpected, and suggested either that the C terminus contributes little to the stability of some receptor–G protein complexes or, alternatively, that some receptors can interact strongly with both cognate and noncognate C termini. To better understand the basis of efficient coupling to Gα subunits with noncognate C termini, we performed atomistic molecular dynamics (MD) simulations of D₁R complexes with G_s, G_si, G_sq, and G_s12. These simulations revealed that in all cases, the C terminus contributed a substantial portion (at least one-third) of the total G_s protein interaction energy with D₁R (SI Appendix, Fig. S7B) and, therefore, was critical for the overall stability of the complex. However, this receptor accommodated noncognate C termini with small energetic penalties (SI Appendix, Fig. S7), which corresponds to the observation that the C terminus was not critical for D₁R selectivity. These simulations also identified several regions of the Gα_s core that contribute to the interaction and may also contribute to coupling selectivity for this receptor, including HN, the hns1, s2s3, and h4s6 linkers and N-terminal residues of H5 (SI Appendix, Fig. S7A).

We then asked if receptors that interacted well with nucleotide-free G proteins bearing noncognate Gα C termini in our direct BRET coupling assay and MD simulations could also efficiently activate these G proteins and produce downstream signals. We found that wild-type G_s, G_si, G_sq, and G_s12 all effectively restored potent activation of adenylate cyclase in response to D₁R activation in cells lacking endogenous Gα_s subunits (Fig. 4A and SI Appendix, Fig. S8), whereas our assays did not detect signals mediated by wild-type G_i1, G_q, or G₁₂ in these cells (SI Appendix, Fig. S9). Likewise, we found that wild-type G_q, G_qi, G_qs, and G_q12 all fully restored calcium release from intracellular stores in response to M₃R activation in cells lacking endogenous Gα_q subunits (Fig. 4B and SI Appendix, Fig. S8), whereas our assays did not detect signals mediated by wild-type G_s, G_i1, or G₁₂ in these cells (SI Appendix, Fig. S9). Notably, both D₁R and M₃R coupled significantly to G_i1 heterotrimers in our direct BRET assay, with ΔBRET_max values that were approximately half that of the respective cognate heterotrimers (Fig. 2) but failed to inhibit cAMP accumulation through G_i1 (SI Appendix, Fig. S9). This suggests that receptors may couple to heterotrimers, as defined by formation of an agonist- and nucleotide-sensitive complex, and yet fail to activate a given signaling pathway even when overexpressed. In contrast to what we observed with D₁R and M₃R, G protein chimeras with noncognate C termini only partially restored calcium release from intracellular stores in response to H₁R activation. The rank order of restoration by chimeras was G_qi > G_qs > G_q12 (Fig. 4C), which was consistent with the rank order of H₁R coupling to wild-type Gα subunits (Fig. 1). These results confirm that some GPCRs can efficiently couple to and activate heterotrimers with a variety of noncognate Gα C termini. Taken together with our direct coupling BRET results, these findings indicate a highly variable influence of the Gα subunit C terminus on coupling selectivity for different GPCRs.

Fig. 4.

Activation of Gα subunit chimeras with noncognate C termini. (A) G_si, G_sq, and G_s12 chimeras fully restored cAMP accumulation in response to D₁R stimulation (10 μM SKF 81297) in cells lacking endogenous Gα_s family subunits (n = 4). (B) G_qs, G_qi, and G_q12 chimeras fully restored calcium release in response to M₃R stimulation (100 μM acetylcholine) in cells lacking endogenous Gα_q family subunits (n = 6). (C) G_qi but not G_qs and G_q12 chimeras fully restored calcium release in response to H₁R stimulation (100 μM histamine) in cells lacking endogenous Gα_q subunits (n = 4). Traces are averages of several experiments, and data points represent independent experiments. Pertussis toxin S1 subunit (PTX) was coexpressed as indicated.

Discussion

In this study, we used energy transfer between GPCRs and heterotrimeric G proteins to monitor allosteric coupling between agonist and nucleotide binding. This approach allowed us to directly compare coupling of several receptors to multiple G protein families without relying on downstream signals such as second messenger production. The use of G protein-deficient cell lines restricted coupling to defined wild-type and chimeric Gα subunits, and permeabilization allowed us to control occupancy of both agonist and nucleotide binding sites.

Our results revealed that many receptors couple more promiscuously than expected. For example, every GPCR that we tested coupled to G_i1 to some extent, and every G_s-coupled receptor that we tested also coupled somewhat to G_q. As a general rule, it appears that receptors that couple primarily to G_i heterotrimers are more selective than receptors that couple primarily to G_s and G_q heterotrimers. These trends are not readily apparent in annotated databases of coupling selectivity (3, 4). We suspect that the sensitivity of our assay, which detects stable complexes between agonist-occupied receptors and nucleotide-free G proteins, allowed us to detect secondary coupling (particularly to G_i1) that might easily be overlooked in functional studies. Indeed, our results with D₁R and M₃R coupling to G_i1 suggest that this method can detect interactions that are too inefficient to lead to physiologically significant G protein activation in cells. However, it is also possible that weak secondary coupling interactions have physiological significance under certain circumstances, as previously demonstrated for β₂AR activation of G_i heterotrimers (38, 39). We also suspect that a single assay that reports coupling to G proteins directly from all four families is likely to produce a more accurate profile of subtype selectivity than comparison across families with multiple assays based on second messenger accumulation and/or gene expression.

Our results with wild-type Gα subunits are consistent with recent computational and structural studies, which have suggested that the outward displacement of transmembrane domain 6 (TM6) is restricted in active G_i-coupled receptors (8, 10–12, 40, 41). This produces a relatively small pocket in the cytoplasmic surface of the receptor that can only accommodate the relatively small C termini of G_i/o family Gα subunits. This mechanism predicts stringent rejection of G_s and G_q heterotrimers by G_i-coupled receptors, but no similar barrier to promiscuous G_i activation, as we observed. However, our results with Gα chimeras also show that some G_i-coupled receptors (e.g., α₂AR and A₁R) can tolerate noncognate C termini to some extent and, therefore, suggest that the Gα_i core region also partly determines selectivity for these receptors.

Overall our results support the well-established role of the Gα subunit C terminus as a key determinant of receptor-G protein coupling selectivity. However, our findings also emphasize that other selectivity determinants exist that are equally important or more important for G protein recognition by many receptors. This is consistent with several previous reports of selectivity determinants that lie outside of the distal C terminus (22–25) and suggests that recognition of several spatially distributed regions of Gα subunits is likely to be a general property of GPCRs. Remarkably, a few receptors virtually ignored the distal C terminus for the purposes of coupling selectivity. This demonstrates that the broad functional diversity of GPCRs extends to the mechanism of receptor-G protein coupling selectivity and provides direct evidence to support the prediction that different receptors will recognize different conserved features of a particular G protein family (5).

Materials and Methods

Materials.

Trypsin, Dulbecco's phospate-buffered saline (DPBS), phosphate-buffered saline (PBS), Hanks' balanced salt solution (HBSS), fetal bovine serum (FBS), Minimal Essential Medium (MEM), Dulbecco's Modified Eagle Medium (DMEM), penicillin/streptomycin, and l-glutamine were from GIBCO (ThermoFisher Scientific). Some receptor ligands, d-luciferin, and forskolin were purchased from Cayman Chemical. The remaining receptor ligands, digitonin, apyrase, protease inhibitor, and GDP were purchased from MilliporeSigma. All detergents were purchased from Anaspec. PEI MAX was purchased from Polysciences Inc.

Plasmids.

Several different GPCR-luciferase constructs were made by appending either the Renilla luciferase variant Rluc8 or NanoLuc (Nluc) directly to the receptor C terminus either by QuikChange PCR or by subcloning into pRluc8-N1 or pNluc-N1 vectors. The V2R-Rluc8 plasmid was received as a gift from Kevin Pfleger (Harry Perkins Institute of Medical Research, Nedlands, WA, Australia). In some cases, a short GGSG linker was inserted between the GPCR and luciferase. Untagged GPCR sequences were obtained either from the cDNA Resource Center, https://www.cdna.org/home.php?cat=0 (Bloomsburg University), as a gift from Jonathan Javitch (Columbia University, New York, NY) or as a gift from Bryan Roth (University of North Carolina, Chapel Hill, NC; PRESTO-Tango Kit: no. 1000000068, Addgene, Watertown, MA). Plasmids encoding Gα subunits were purchased from the cDNA Resource Center. To generate chimeric Gα subunits and Gα_{s Q227L10delct}, we used the PCR, reverse primers incorporating alternative C-terminal sequences, and wild-type Gα templates to amplify full-length Gα subunit sequences that were ligated into pcDNA3.1(+) using KpnI and XhoI. To generate p115RhoGEF-Rluc8, the sequence encoding p115RhoGEF was amplified from p115RhoGEF-GFP and ligated into a Rluc8-N1 vector with BglII and AgeI. All plasmid constructs were verified by Sanger sequencing. A plasmid encoding the S1 subunit of pertussis toxin (PTX-S1) was kindly provided by Stephen R. Ikeda (National Institute on Alcohol Abuse and Alcoholism, Rockville, MD). Plasmids encoding masGRKct-Rluc8, Venus-Kras, Venus-1–155-Gγ₂, and Venus-155–239-Gβ₁ have been described previously (37, 42). The Glosensor-22F cAMP plasmid (E2301) was obtained from Promega, p115RhoGEF-GFP was received as a gift from Phil Wedegaertner (Thomas Jefferson College, Philadelphia, PA), and the pT7-CalfluxVTN plasmid (43) was a gift from Carl Johnson (Vanderbilt University, Nashville, TN; Addgene plasmid 83926).

Cell Culture and Transfection.

HEK 293 cells (American Type Culture Collection) were propagated in plastic flasks and on six-well plates according to the supplier’s protocol. HEK 293 cells with targeted deletion of GNAS and GNAL (GSKO), HEK 293 cells with additional targeted deletion of GNAS, GNAL, GNAQ, GNA11, GNA12, and GNA13 that are G protein three family knockouts (3GKO), and HEK 293 cells with additional targeted deletions to the 3GKO cells of GNAI1, GNAI2, GNAI3, GNAT1, GNAT2, GNAZ, and GNAO1 that are G protein four family knockouts (4GKO) were derived, authenticated, and propagated as previously described (29, 44). Cells were transiently transfected in growth medium using linear polyethyleneimine MAX (PEI MAX; MW 40,000) at an N/P ratio of 20 and were used for experiments 12–48 h later. Up to 3.0 μg of plasmid DNA was transfected in each well of a six-well plate.

BRET and luminescence assays.

Validation of Gα subunit ability to form heterotrimers: HEK 293 cells were transiently transfected with masGRKct-Rluc8, Venus-1–155-Gγ₂, Venus-155–239-Gβ₁, pcDNA3.1(+), and a Gα subunit in a (1:1:1:5:0) ratio or a (1:1:1:4:1) ratio or a (1:1:1:0:5) ratio. After a 24-h incubation, cells were washed twice with 1× DPBS, harvested by trituration, and transferred to opaque black 96-well plates.

Measurement of coupling between receptor and G protein in nucleotide-depleted cells.

3GKO or 4GKO cells were transiently transfected with a GPCR-Rluc8 and Gα subunit pair, Venus-1-155-Gγ₂, Venus-155-239-Gβ₁, and pcDNA3.1(+) or PTX-S1 in a (1:3:1:1:1) ratio. Experiments with Gα_i C termini were conducted in 4GKO cells for Gα_i cognate receptors and in 3GKO cells for all other receptors. Experiments with Gα_i C termini were conducted without PTX-S1, all other Gα subunits were cotransfected with PTX-S1. After a 48-h incubation, cells were washed twice with permeabilization buffer (potassium permeabilization solution; KPS) containing 140 mM KCl, 10 mM NaCl, 1 mM MgCl₂, 0.1 mM KEGTA, 20 mM NaHEPES (pH 7.2), harvested by trituration, permeabilized in KPS buffer containing 10 μg·mL⁻¹ high purity digitonin, and transferred to opaque black 96-well plate. Measurements were made from permeabilized cells supplemented either with 0.5 mM GDP or 2 U·mL⁻¹ apyrase, in both cases with or without agonist (SI Appendix, Table S1).

Data processing.

Net BRET was calculated as the raw BRET ratio minus the same ratio measured from cells expressing only the BRET donor. Heatmaps represent ΔBRET_max (Fig. 3; BRET in the presence of agonist and apyrase minus BRET in the presence of GDP alone), ΔBRET_GDP (SI Appendix, Fig. S2; BRET in the presence of apyrase alone minus BRET in the presence of GDP alone), and ΔBRET_ag (SI Appendix, Fig. S3; BRET in the presence of agonist and GDP minus BRET in the presence of GDP alone). Background ΔBRET (presumably due to endogenous Gα subunits remaining in 3GKO and 4GKO cells) measured from control cells not expressing exogenous Gα subunits was routinely subtracted. Control cells expressed PTX-S1 for experiments with non-Gα_i C termini. Normalized ΔBRET values were obtained by dividing the ΔBRET observed for each chimera in a heatmap by the highest value observed for each receptor. For the Glosensor assays with G_s, G_si, G_sq, and G_s12, vehicle-subtracted luminescence changes in response to agonist were normalized to vehicle-subtracted luminescence changes in response to forskolin. For the Glosensor assay with G_i1 the percent change in luminescence was given as agonist-induced change in luminescence over baseline luminescence. For the Calflux assay ΔBRET_HA and ΔBRET_Ach were calculated from the kinetic experiments by subtracting the average of the time points before agonist addition from the average of the final five time points after agonist addition.

Detailed methods related to second messenger assays, BRET and luminescence measurements, Gα subunit immunoprecipitation and structural modeling and molecular dynamics simulations can be found in SI Appendix.