CODA-RET reveals functional selectivity as a result of GPCR heteromerization

Measuring receptor activation by receptor-G protein BRET

We first studied dopamine D1 and D2 receptor (D1R and D2R) conformational changes in a BRET-based receptor-G protein assay upon stimulation with the endogenous agonist dopamine (see Supplementary Results section; Supplementary Figure 1, Supplementary Table 1). The BRET¹ assay requires an energy donor, Renilla Luciferase 8²⁶ (RLuc8), fused to the C-terminus of the receptor of interest and an acceptor, mVenus²⁷, inserted into the Gα subunit at the same position where luciferase was introduced previously²¹. These sensors have been previously characterized and do not significantly alter the function^19-21 of the wild type (wt) proteins. This technology has been used to monitor conformational changes at the level of the GPCR-G protein interaction for other receptors²⁰. Consistent with the known coupling of D1R to Gαs, dopamine caused a significant BRET change in cells expressing D1R-RLuc8 with Gαs-mVenus but not with Gαi-mVenus (Supplementary Figure 1a). This G protein specificity was also observed with another D1R agonist (Supplementary Figure 2a-b, Supplementary Table 1).

Similarly, consistent with the known coupling of D2R to Gαi/o, in cells coexpressing D2R-RLuc8 with Gαi₁-mVenus we detected a significant dopamine-induced BRET change, whereas no change in BRET was detected with Gαs-mVenus (Supplementary Figure 1b). Other D2R agonists also led to enhancement of the D2R-G protein BRET signal only with Gαi-mVenus and not Gαs-mVenus (Supplementary Figure 2c-d). We evaluated three different mVenus insertion positions in Gαi, and we detected the greatest dynamic range with the insertion at position 91 (Supplementary Figure 2e). Therefore, in all subsequent experiments we used the position 91 Gαi-mVenus fusion. Changes in BRET observed with receptor-G protein based biosensors are indicative of conformational change and not necessarily of G protein activation. Importantly, when the sensor is inserted at position 91 of the Gαi subunit, the G protein conformational changes have been shown to correlate closely with G protein activation²¹. In order to confirm that the conformational change induced by dopamine also parallels G protein activation, we performed a series of experiments using a G protein biosensor in a configuration that monitors conformational changes at the level of the G protein heterotrimer itself. We expressed a non-tagged receptor (D1R or D2R) together with a Gα-RLuc, mVenus-Gγ₂ and untagged Gβ₁ and we quantified dopamine-induced BRET (Supplementary Figure 1c-d, Supplementary Table 1). As with the receptor-Gα based biosensor, we only detected BRET changes when the cognate Gα protein was coexpressed along with D1R or D2R (Supplementary Figure 1c-d). Agonist-induced BRET changes were inhibited by specific antagonists (Supplementary Figure 3), consistent with the biosensor reporting on receptor activation.

Thus, as described previously for a BRET² assay with adrenergic receptors and related Gα biosensors with luciferase or GFP10 inserted at the equivalent of position 91²¹, we were able to use agonist-induced BRET¹ changes to monitor conformational changes that are associated with the activation process. Maximal agonist-induced BRET changes were observed upon coexpression of the Gα biosensor together with untagged Gβγ (Supplementary Figure 2f), as previously observed for other receptors²⁰.

We obtained similar results with a biosensor that monitored receptor interaction with Gβγ. Here the energy acceptor was a Gβγ-mVenus biosensor, formed by complementation of split mVenus fusion constructs of Gβ₁ and Gγ₂^28,29 (Supplementary Figure 4), which allows tight control of Gβγ isoforms. Although these biosensors proved robust when coexpressed with receptor fused to RLuc8 and unlabeled Gα, a small component of the signal resulted from endogenous Gα (Supplementary Figure 4), and therefore this orientation of the assay does not give the same precision when studying conformational changes in a defined Gα subunit. Nonetheless, this novel biosensor is of great potential utility as it allows monitoring conformational changes in defined combinations of Gβ and Gγ subunits, an aspect of GPCR signaling that is not well understood.

Expression of D1R alters NPA-induced D2R-G protein BRET

Next we evaluated the impact of receptor coexpression on the agonist induced BRET change. While coexpression of untagged D2R did not alter D1R-dependent Gαs conformational change (Figure 1a), the presence of D1R modified D2R-mediated agonist-induced BRET changes (Figure 1b). Whereas the quinpirole-induced BRET change was not altered in the presence of D1R, we observed a significant gain of potency for R(–)-propylnorapomorphine (NPA) (p=0.0001, Figure 1b, Table 1), without any change in coupling specificity (Supplementary Figure 2g-h). These data are compatible with the possibility that D1R-D2R heteromerization may be responsible for at least some of the functional selectivity manifested in the ability of different D2R drugs to promote different conformational rearrangements. Although we cannot completely rule out signaling integration resulting from crosstalk of parallel effector pathways, our readout is upstream at the level of the receptor-G protein interaction and not downstream effectors where integration might be more likely.

Controlling the protomers in a signaling unit

To dissect functional selectivity of D2R drugs at the dimer level, we sought to devise a method that would allow dissection of G protein conformational changes by a defined receptor dimer using receptor sensors recently validated for D2R¹⁹ that exploit complementation of split luciferase as a reporter of molecular proximity³⁰. In these constructs the C-terminus of the receptor is fused to an N-terminal fragment (L1) or C-terminal fragment (L2) of RLuc8. We also constructed the equivalent D1R fusion constructs. No differences between D2R-D2R and D2R-D1R were observed when molecular proximity was evaluated by luciferase complementation (Supplementary Figure 5a) or by BRET titration experiments (Supplementary Table 2), suggesting a similar propensity for homo- and heteromerization. Similarly, none of the ligands tested altered the propensity for interaction based on the results of D1R-D2R BRET titration experiments (Supplementary Table 3). Furthermore, none of the drugs targeting one or both receptors altered RLuc8 complementation of the split D2R-D1R or D2R-D2R constructs (Supplementary Figure 5b-c).

Receptor-G protein BRET experiments similar to those described above (Figure 1) were performed in cells that coexpressed D1R-L1 and D1R-L2 or D2R-L1 and D2R-L2 with the Gαs or Gαi mVenus acceptors (Figure 2). Dopamine enhanced the BRET signal in cells that coexpressed D1R-L1, D1R-L2 and Gαs-mVenus (Figure 2a-b, Supplementary Figure 6a-b), but not when the D1R-splits were coexpressed with Gαi-mVenus (Figure 2b), consistent with the expected G protein coupling specificity. G protein specificity also was observed with other D1R agonists (Figure 2c-d). In contrast, Gαi specificity was observed for agonist-induced BRET changes for the complemented D2R-L1 and D2R-L2 splits (Figure 2e-h). Moreover, no significant differences in the potency of dopamine-induced BRET changes were observed between the full length receptor-luciferase fusions and the split luciferase receptor constructs (Table 1, Supplementary Table 1). This indicates not only that the splits express and function normally, but also that the orientation of the complemented donor is similar and has not introduced an observable alteration of the receptor-G protein interface. As for the full length RLuc8 donor assays, Gαi with mVenus inserted at position 91 resulted in the best dynamic range (Supplementary Figure 7a-b) and was used for all subsequent experiments. Importantly, selective drugs did not trigger any BRET change if the target receptor was not expressed (quinpirole in Supplementary Figure 6b). D1R agonists caused concentration-dependent conformational changes in the D1R homomer indicative of Gαs activation that are in agreement with the relative potencies and efficacies reported previously using other assays³¹ (Figure 2a-d, Supplementary Figure 6b, see Table 1 for quantified potencies). The same was observed for D2R homomer activation of Gαi, which identified NPA as the most potent agonist studied (Figure 2e-h, Table 1). Comparable results were observed when we used the Gβγ-mVenus or Gγ-mVenus biosensors for D2R homomers (Supplementary Figure 8, Supplementary Table 1).

As expected, selective antagonists dramatically decreased agonist potency at the cognate receptor. In cells coexpressing the D1R splits and Gαs-mVenus, the D1R antagonist SCH23390 decreased the potency of dopamine, DAR100 and SKF83822 for activating Gαs (Figure 2b, Supplementary Figure 6c-d); and the D2R antagonist sulpiride decreased the potency of dopamine, NPA and quinpirole for activating Gαi (Figure 2f, Supplementary Figure 7c-d). In contrast, SCH23390 did not inhibit NPA- or quinpirole-induced BRET increase in D2R Gαi-mVenus expressing cells (Supplementary Figure 7c-f). Notably the D2R-selective antagonist sulpiride has a fixed negative charge and at 1 μM is essentially membrane impermeant³². Therefore, the dramatic inhibition of the agonist-induced BRET change by this concentration of sulpiride in cells coexpressing the D2R splits and Gαi-mVenus, (Figure 2f, Supplementary Figure 7c-f), indicates that essentially all of the agonist-induced BRET changes take place at the plasma membrane.

Targeting the D1R-D2R heteromeric signaling unit

This approach allows targeting of specific protomers through luciferase complementation and thus analysis of signaling of defined heteromers without a contribution from homomeric complexes. Thus, by coexpressing D1R-L1 and D2R-L2 (or D2R-L1 and D1R-L2), the only luminescence signal will result from complementation of D1R and D2R and any homomeric species will be silent at the level of the BRET readout. We first studied the Gαs pathway by coexpressing both D2R-L1 and D1R-L2 with Gαs-mVenus. Selective D1R stimulation resulted in a concentration-dependent BRET increase with agonist potencies similar to those observed in the D1R homomers (Figure 3a-d, Supplementary Figure 6e-f, Table 1, Supplementary Table 1). The D2R-specific agonist quinpirole did not alter the Gαs-mVenus BRET signal (Supplementary Figure 6f). Furthermore, as expected, the selective D1R inhibitor SCH23390 competed with the D1R-selective compounds, whereas the D2R-selective antagonist sulpiride did not inhibit the D1R-mediated Gαs response (Figure 3a-d). Selective antagonism by SCH23390 was also seen for DAR100 (Supplementary Figure 6g-h). Swapping the orientation of the splits led to the same results (Supplementary Figure 9a). This set of experiments demonstrates that, agonist binding to the D1R protomer in a D1R-D2R complex is sufficient to trigger a conformational change between the heteromer and Gαs consistent with activation. This establishes that one agonist is sufficient to activate a receptor heteromer. Moreover, since our approach monitors the direct physical interaction between the receptor heteromer and Gα protein, downstream signaling crosstalk is very unlikely to contribute to the observed BRET changes.

Functional selectivity of NPA at the D2R-D1R heteromer

Next, we studied the Gαi pathway by coexpressing D2R-L1 and D1R-L2 with Gαi-mVenus. Again D2R agonists led to BRET increases similar to those observed in the D2R homomers (Figure 3e-h, Table 1). Swapping the splits led to comparable results with the same potency for quinpirole (Supplementary Figure 8b, Supplementary Table 1). As expected, agonist-induced BRET was inhibited by the D2R antagonist sulpiride (Figure 3e-h), but not by the D1R antagonist SCH23390 (Figure 3e-h). Thus, agonist binding to the D2R in the heteromer triggers a conformational change in the D2R-D1R-Gαi signaling complex.

With Gαs and the complemented donors, we detected no difference in potency between the D1R homomer and D1R-D2R heteromer for any of the D1R agonists tested (Figure 4a, Supplementary Figure 9a, Table 1 and Supplementary Table 1). Likewise, with Gαi the potencies of quinpirole and dopamine to enhance BRET were the same in the complemented D2R homomer and the D1R-D2R heteromer (Figure 4b). In contrast, in agreement with the increased potency for the NPA induced BRET change in D2R-RLuc8 Gαi-mVenus in the presence of untagged D1R (Figure 1b, Table 1), NPA was approximately 10-fold more potent at the heteromer than at the D2R homomer (p<0.005, Figure 4b, Table 1). These results provide evidence for a novel molecular mechanism for functional selectivity, by which a drug has different potencies when targeting homo- or heteromeric complexes.

Since NPA also binds and activates D1R, it is conceivable that NPA binding to the D1R protomer in the heteromer might help to trans-activate Gαi and thereby enhance NPA's potency, but our results suggest otherwise. The NPA-induced BRET change with the D2R-D1R-Gαi sensor was competed by the D2R-inhibitor sulpiride but not by the D1R-inhibitor SCH23390 (Figure 3h), suggesting that NPA activates Gαi only by binding to the D2R and not to the D1R. In addition, whereas complementation of a mutant D2R that does not bind agonist³³ (D2R-D114A-L1) with D1R produced a heteromer that signaled normally to Gαs, neither NPA, quinpirole, nor dopamine signaled to Gαi (Supplementary Figure 10), indicating the binding to the D2R protomer is essential for Gαi activation.

Importantly, expression of excess unlabelled D1R did not alter the ability of the D2R-D2R splits to activate Gαi (Supplementary Figure 11a-b) and failed to allow the D2R-D2R splits to activate Gαs (Supplementary Figure 11e-g). Similarly, expression of unlabeled D2R did not impact the ability of the D1R-D1R splits to activate Gαs (Supplementary Figure 11c-d) and failed to produce activation of Gαi (Supplementary Figure 11h-j). These critical controls establish that the biosensors only read out on the activity of the receptor protomers to which they are directly fused.

Constructs for expression vectors and transfection

The cDNA for human D1R was obtained from www.cdna.org. D1R was tagged with signal peptide (SP) followed⁴⁶ by a Myc epitope tag using standard molecular biology procedures. The cDNAs encoding full length Renilla Luciferase 8 (RLuc8) or fragments for the L1 (residues 1-229) or L2 (residues 230-311) were fused in frame to the C terminus of D1R in the pcDNA3.1 vector. The human D2_shortR sensors were already reported¹⁹. The following human G protein constructs were used: untagged Gα_i1 and long Gαs, Gα_i1-mVenus with mVenus inserted at position 60, 91, or 122, long Gαs-mVenus with mVenus inserted at position 71 (aligned with position 60 in Gα_i1), Gα_i1-RLuc (humanized regular Renilla Luciferase) inserted at position 91, Gαs-RLuc with RLuc inserted at position 113, untagged Gβ₁, Gβ₁ fused with V2 (C terminal split of mVenus, aa 156-240) at its N-terminus, untagged Gγ₂, Gγ₂ fused with V1 (N terminal split of mVenus, aa 1-155) at its N-terminus, Gγ₂ fused to full-length mVenus at its N-terminus. All the constructs were confirmed by sequencing analysis.

A constant amount of plasmid cDNA was transfected into HEK-293T cells using polyethylenimine (PEI) (Polysciences Inc) in a 1 to 3 ratio in 10 cm dishes. Cells were maintained in culture with DMEM supplemented with 10% FBS. The transfected amount and ratio among the receptor-L1, receptor-L2, Gα, Gβ₁, Gγ₂ was optimized by testing various ratios of plasmids encoding the different sensors. Experiments were performed ∼48 hours post-transfection.

BRET

BRET¹, which differs from other forms of BRET (BRET² and BRET³) in that it uses a yellow fluorescent protein as an acceptor for energy transfer from luciferase, were performed as described^19,47. Briefly, cells were harvested, washed and resuspended in PBS. Approximately 200,000 cells/well were distributed in 96-well plates and 5 μM coelenterazine H (substrate for luciferase) was added to each well. One min after addition of coelenterazine H ligands were added to each well and immediately after the addition, the fluorescence (excitation at 500 nm and emission at 540 nm, 1s recording) and luminescence after substrate addition (no filters, 1s recording) were quantified (Polarstar and Pherastar, respectively, BMG). In parallel cells the BRET signal was determined by quantifying and calculating the ratio of the light emitted by mVenus (510–540 nm) over that emitted by RLuc8 or RLuc (485 nm) for BRET¹. The net BRET values were obtained by subtracting the background determined in cells expressing RLuc8 or RLuc alone. Results are expressed as the BRET change produced by the corresponding drug. As shown in the cartoons, bimolecular complementation of donor was incorporated into the BRET assay. In a majority of the combinations tested, luminescence resulted from bimolecular luminescence complementation between D2R-L1 and D1R-L2 and RET took place between that complex and Gα-mVenus. Data and statistical analysis were performed with Prism 5.01 (GraphPad Software).

FACS Analysis

FACS was performed to determine surface expression of each receptor construct, as described⁴⁸ with the Guava Easy Cite cytometer (Guava Techonologies) or C6 Flow Cytometer (Accuri). Briefly, a fraction of the same cells used for luminescence complementation was harvested and incubated with anti-Flag (Sigma) or anti-Myc (Hybridoma Facility, Mount Sinai, New York), washed and incubated with secondary antimouse antibody labelled with Alexa Fluor 647 (Invitrogen).

Drugs

Pharmacological reagents were purchased from following venders: dopamine (Sigma), quinpirole (Sigma), R(–)-Propylnorapomorphine (NPA) (Sigma), dihydrexidine (DAR-100) (Tocris), SKF38393 (Sigma), SKF81297 (Tocris), SKF83822 (Tocris), sulpiride (Sigma), SCH23390 (Sigma). Each ligand was solubilized in a corresponding solvent (e.g. methanol, ethanol, and DMSO) and non-ligand control contained the same amount of solvent to address potential solvent effects on the BRET assay.