Gα_o/z-biased coupling profile of the dopamine D3 receptor

Abstract

Dopamine receptors are G protein coupled receptors (GPCRs) that serve as key targets for FDA-approved drugs to treat various neuropsychiatric disorders. Notably, ~11% of all marketed GPCR-targeting drugs act on dopamine receptors. Five GPCRs mediate the effects of endogenous dopamine and compounds used to treat Parkinson’s disease, schizophrenia, and other conditions. However, on-target side effects associated with these medications highlight the need to analyze dopamine receptor signaling to design safer, more effective therapeutics. Here, we characterize the G protein coupling of dopamine D2-like receptors and observe the striking inability of D3R to engage with Gα_i proteins while effectively activating Gα_o/z subtypes. Applying orthogonal cell-based assays with wild-type G proteins in parental and ΔGα_i/o/z cells, we conclusively establish that D3R does not activate Gα_i proteins. Further analysis of Gα_i2:Gα_oA and D2R:D3R chimeras reveals that this selective inability is driven by molecular determinants located within the α5 helix of Gα_i and ICL2 of D3R. Guided by cryo-EM structures, we model the interface between these regions to clarify the structural basis of selectivity. Finally, treating hippocampal neurons in acute slices with selective agonists and antagonists reveals marked differences in their regulation of endogenous adenylyl cyclase and cAMP, underscoring the neurophysiological significance of our findings.

D3R displays a distinct Gα_o/z-biased signaling profile, lacking Gα_i coupling. Structural and chimera analyses identify determinants of selectivity, with neuronal assays confirming functional divergence, highlighting implications for more targeted dopamine therapeutics.

Introduction

Dopamine is a critical catecholaminergic neurotransmitter that regulates a broad range of functions in both the central and the peripheral nervous systems^1–3. Dopamine actions are mediated by the activation of five G protein-coupled receptors (GPCRs) that are classified into two families based on their structural and signaling properties: D1-like receptors (D1R, D5R) and D2-like receptors (D2R, D3R, D4R). D1-like receptors generally couple to Gα_s proteins that stimulate adenylyl cyclase (AC) to produce cAMP, whereas D2-like receptors predominantly couple to inhibitory Gα_i/o/z proteins, resulting in AC inhibition and activation of additional downstream signaling pathways, including those mediated by β-arrestin recruitment^3–11. A key emerging concept in signal transduction is that a single receptor can adopt multiple conformations, each selectively engaging different intracellular transducers and triggering distinct cellular responses^12–15. This signaling bias includes preference not only between G protein and β-arrestin pathways but also among different G protein subtypes^16–19. Dysregulation of D2-like receptor signaling has been implicated in a wide range of neurological and psychiatric disorders, including Parkinson’s disease (PD), Huntington’s disease (HD), schizophrenia (SZ), major depressive disorder (MDD), attention-deficit/hyperactivity disorder (ADHD), and substance use disorders^20–24. These associations highlight the critical importance of understanding the fundamental mechanisms of dopamine receptor signaling. In particular, biased signaling offers a promising framework for the development of targeted and effective therapeutics with fewer side effects, by selectively modulating disease-relevant signaling cascades²⁵. Achieving this goal requires a deeper understanding of the molecular details related to how each D2-like receptor engages with different intracellular transducers.

The D2R is the most extensively studied subtype, with detailed characterization of its coupling to Gα_i/o/z proteins¹⁶, including differences between its two long and short splicing isoforms¹⁹. In contrast, D3R and D4R, despite sharing structural and signaling features with D2R, are less understood in terms of their selective coupling to Gα protein subtypes. In particular, the Gα protein coupling profile of D3R remains controversial: while some studies reported that it does not engage Gα_i subtypes^26–28, others described broader coupling across the Gα_i/o/z family^27,29,30, or a more selective activation of Gα_i1/i3 with no engagement of Gα_i2³¹. In this study, we conclusively defined the signaling properties of D3R and D4R, demonstrating that D3R displays a distinct and consistent inability to couple with Gα_i proteins, despite robust activation of Gα_o and Gα_z. This biased coupling pattern was further corroborated by exploring structural elements and defining the responsible molecular determinants. Finally, we demonstrated the physiological relevance of this discovery in neurons. Offering physiologically relevant evidence that D3R selectively couples to Gα_o and Gα_z, we resolved previous inconsistencies and paved the way for the development of targeted D3R-based therapies.

Results

Unique coupling profile of the dopamine D3 receptor

We initially assessed the G protein coupling specificity of the dopamine D2-like receptors (D2R, D3R, and D4R) to confirm their preferential interaction with heterotrimeric G proteins of the Gα_i/o/z subfamily, as previously reported⁴. To this goal, we transfected HEK293T cells with a representative Gα subunit from the four major G protein families, Gα_i/o/z, Gα_s, Gα_q, and Gα_12/13, and we applied a kinetic G protein nanoBRET assay (Supplementary Fig. 1A) to assess each G protein activation by D2R, D3R, and D4R in real time. Upon stimulation with either quinpirole (Fig. 1A), a selective D2-like receptor agonist, or the non-selective endogenous agonist dopamine (Supplementary Fig. 1B), all three receptors displayed clear and exclusive activation of the Gα_o subunit. No activation of Gα_s, Gα_q, or Gα₁₃ proteins was observed under any condition, confirming the well-established Gα_i/o/z-selective signaling profile of D2-like receptors. To further investigate the receptor-specific coupling preferences, we assessed the ability of D2R, D3R, and D4R to activate each of the six main Gα_i/o/z subunits (Gα_oA, Gα_oB, Gα_i1, Gα_i2, Gα_i3, and Gα_z). Applying the same G protein nanoBRET assay, we obtained concentration–response curves to quantitatively compare the coupling efficiency of the three dopamine receptors in response to both quinpirole and dopamine (Fig. 1B, Supplementary Data 1, and Supplementary Fig. 1C). As expected, D2R and D4R displayed robust activation of all six G proteins with a preference for Gα_o over Gα_i and Gα_z. In contrast, D3R showed a distinct and restricted coupling profile, with no detectable activation of any Gα_i subtype (Gα_i1, Gα_i2, and Gα_i3), while retaining efficient coupling to Gα_oA, Gα_oB, and Gα_z (Fig. 1B, Supplementary Data 1, and Supplementary Fig. 1C). This allowed the calculation of EC₅₀ and E_max values for individual G proteins, which are visualized as radar plots to facilitate comparison of both potency and efficacy (Fig. 1C). These analyses further reinforced the functional similarity between D2R and D4R, while once again confirming the markedly different and selective profile of D3R. Taken together, these results suggest that D3R exhibits signaling selectivity within the Gα_i/o/z family, favoring Gα_o/z over Gα_i subtypes.

Fig. 1. D3R shows a Gα_o/z-selective coupling profile.

A Representative kinetic activation profiles measured using the G protein nanoBRET assay for dopamine receptors D2R, D3R, and D4R in response to application of 10 µM quinpirole applied at 10 s. Co-transfected Gα subunits from the four major G protein families are indicated. B Concentration-response curves for D2R, D3R, and D4R co-transfected with indicated Gα_i/o/z subunits and stimulated with increasing concentrations of quinpirole. Data were normalized to the signal obtained with GαoA and shown as mean ± SEM. N = 5 independent replicates. C Radar plots summarizing pEC50 and Emax for D2R, D3R, and D4R in response to quinpirole across all six Gα_i/o/z subunits.

The observation that D3R fails to couple to Gα_i subtypes prompted us to investigate whether this reflects true molecular selectivity or results from suboptimal expression of either the receptor or the G protein in our system. We first tested whether varying the expression of D3R would rescue the coupling to Gα_i. Accordingly, we transfected HEK293T cells with a fixed amount of either Gα_oA or Gα_i2 plasmids and co-transfected increasing amounts of the D3R plasmid. We found that D3R robustly activated Gα_oA at all receptor expression levels tested, demonstrating the functionality of the assay across different transfection ratios (Supplementary Fig. 2A). In stark contrast, no activation of Gα_i2 was observed with any amount of transfected D3R-encoding plasmid (Supplementary Fig. 2B). This suggests that receptor expression levels do not impact the assay ability to detect G protein activation effectively. To test whether the amount of Gα_i2 was a limiting factor, we obtained concentration-response curves by treating cells transfected with a fixed amount of D3R and systematically increasing the levels of Gα_oA, Gα_i2, and Gα_z. As a crucial negative control, no signal was generated in cells not transfected with Gα proteins. As expected, increasing the expression of Gα_oA and Gα_z resulted in a concentration-dependent decrease in the maximal response (E_max) due to a shift of the equilibrium in the formation of the Gβγ-venus: GRK3-Nluc complex towards the G protein heterotrimer (Supplementary Fig. 2C). In contrast, Gα_i2 failed to produce any measurable activation signal, demonstrating that the lack of coupling is an intrinsic property and not an assay artifact. Next, we assessed the effect of the cellular background on the interpretation of the D3R coupling results by obtaining concentration-response curves to quinpirole applied to Chinese Hamster Ovary (CHO) cells transfected with D2R or D3R and each Gα_i/o/z protein. Similar to our observations in HEK293 cells, we detected activation of all six Gα_i/o/z proteins by D2R, while D3R only activated Gα_oA, Gα_oB, and Gα_z (Supplementary Fig. 3). Taken together, these control experiments further substantiate that the observed Gα_o/z-selective coupling profile is an intrinsic property of D3R.

Ligand-bias does not explain the lack of Gα_i coupling of D3R

The concept of biased agonism proposes that ligands can stabilize distinct active receptor conformations, promoting selective engagement with specific transducers and activation of particular signaling pathways¹². To determine whether quinpirole might act as a biased ligand by stabilizing D3R conformations incapable of coupling to Gα_i proteins, we compared its activity with multiple other dopamine receptor agonists. In addition to quinpirole and the endogenous ligand dopamine, we also included rotigotine, a potent D1R, D2R, and D3R agonist, which was recently used to induce an active conformation of D3R subsequently determined by cryo-EM structure^32,33, sumanirole, a reported D2R-selective agonist, and ML417, a D3R-selective agonist³⁴. Using the G protein nanoBRET assay, we generated concentration-response curves for each ligand based on their ability to activate Gα_oA, Gα_i2, or Gα_z via D2R (Fig. 2A) and D3R (Fig. 2B). Notably, none of the tested agonists induced dissociation of the G_i heterotrimer in D3R-expressing cells, although all were capable of activating Gα_oA and Gα_z (Fig. 2B and Supplementary Data 2). We further confirmed that ML417 preferentially activates D3R over D2R. While a partial activation of Gα_oA via D2R was observed at high ML417 concentrations, the ligand was 63-fold more potent at D3R and showed no activation of Gα_i2 or Gα_z by D2R. Additionally, we found that sumanirole triggered Gα_z heterotrimer dissociation with similar potency at both receptors but was five times more effective at activating Gα_oA via D2R compared to D3R. These findings exclude ligand bias as the cause of the inability of D3R to activate Gα_i-dependent signaling.

Fig. 2. The unique D3R activation profile is not a result of biased agonism.

A Concentration-response curves obtained using the G protein nanoBRET assay in wild-type HEK293 cells expressing D2R with indicated G proteins and treated with indicated agonists. Data are shown as means ± SEM. N = 5 independent replicates. B Concentration-response curves for D3R-transfected cells with indicated G proteins and treated with indicated agonists. Data were normalized to the signal obtained by applying dopamine to cells expressing Gα_oA and shown as mean ± SEM. N = 5 independent replicates.

Orthogonal assay confirms lack of D3R-Gα_i1/i2/i3 coupling, while β-arrestin recruitment is not affected

To further validate our G protein nanoBRET data, we employed the GloSensor cAMP inhibition assay in ΔGα_i/o/z cells to eliminate background interference from endogenous Gα_i/o/z proteins. Consistent with the nanoBRET data, stimulation with 1µM quinpirole led to a marked decrease in cAMP levels across all the receptor-G protein combinations, except when D3R was co-expressed with Gα_i1, Gα_i2, or Gα_i3 (Fig. 3A–C). The absence of cAMP inhibition in this condition further confirms that D3R does not functionally couple to Gα_i proteins, supporting the conclusion that D3R preferentially signals through Gα_o and Gα_z, rather than Gα_i subunits.

Fig. 3. Orthogonal assay in ΔGα_i/o/z cells confirms the lack of Gα_i1/i2/i3 coupling by D3R.

Percentage fold change of cAMP accumulation in response to stimulation with 1 µM quinpirole of D2R (A), D3R (B), D4R (C) co-transfected with Gα_oA, Gα_i2, and Gα_z in ΔGα_i/o/z HEK293 cells. Fold change was calculated as the ratio of maximal amplitude over baseline. cAMP levels were measured using the GloSensor assay in HEK293 cells lacking endogenous Gα_i/o/z proteins (ΔGα_i/o/z cells). Data are shown as mean ± SEM. N = 5 independent replicates. ****P < 0.00001, determined by 2-way ANOVA and Šídák’s multiple comparisons post hoc test. D Concentration-response curves illustrating β-arrestin2 recruitment to D2R, D3R, and D4R following stimulation with quinpirole and dopamine, normalized as fold change, with response observed at the lowest concentration of ligand treated as 1. E Basal luminescence observed following co-transfection of receptor-SmBiT and βarr2-LgBiT. Data are shown as mean ± SEM. N = 3 independent replicates. **P < 0.01, determined by one-way ANOVA and Tukey post hoc test.

To assess if dopamine receptors D2R, D3R, and D4R also demonstrated a unique signaling profile in terms of β-arrestin recruitment, we employed a NanoBiT-based protein complementation assay. We generated concentration-response curves following stimulation with either dopamine or quinpirole (Fig. 3D and Supplementary Data 3). In response to both agonists, D2R showed a significantly stronger β-arrestin recruitment profile compared to D3R and D4R. It is worthwhile to note that while quinpirole was equally potent and efficacious in inducing β-arrestin recruitment to D3R and D4R, the endogenous ligand dopamine was slightly more potent and efficacious on D3R compared to D4R. Interestingly, D2R exhibited a higher basal level of β-arrestin recruitment relative to the other two receptors (Fig. 3E).

Structural determinants of D3R G protein coupling selectivity

Building on our previous findings, we sought to investigate the molecular determinants of the D3R-G protein coupling selectivity. We focused on the C-terminal α5 helix of the Gα subunit, a region known to be a primary interface for GPCR interaction^35–37. An alignment of the Gα_oA and Gα_i2 C-termini revealed key differences within the final 10 amino acids (Fig. 4A). To test the importance of this region, we engineered two chimeric proteins by replacing the last 6 or 10 amino acids of the C-terminus of Gα_i2 with the corresponding sequence from Gα_oA, creating Gα_i2Gα_oA(6) and Gα_i2Gα_oA(10) (Fig. 4A, B). To confirm the functional integrity of these chimeras, we assessed their activation by D2R in response to quinpirole (Fig. 4C). As anticipated, both chimeras produced a clear activation signal, confirming productive coupling to the receptor. When tested with D3R, both chimeras exhibited a partial yet significant rescue of the coupling (Fig. 4D). Specifically, Gα_i2Gα_oA(6) restored 31% of the activation signal observed with wild-type Gα_oA, while Gα_i2Gα_oA(10) restored 37% of the signal. These results demonstrated that the Gα C-terminus is a critical determinant of coupling specificity for D3R. However, the incomplete nature of the rescue indicates that while these residues are essential, they are not the sole factor governing the interaction, suggesting that other regions of the G protein and possibly other contributors also influence the selective recognition.

Fig. 4. The C-terminal α5 helix (H5) is important for G protein coupling to dopamine receptors, but it is not the sole driver.

A Alignment of the C-terminal α helix of Gα_i2, Gα_oA, and two chimeric constructs in which the last 6 or 10 residues of Gαi2 were replaced with the corresponding residues from Gα_oA, respectively indicated as G_i2G_oA(6) and G_i2G_oA(10). Gα_i2-specific residues are highlighted in red; GαoA-specific residues in blue. B Models of chimeric G protein constructs based on Gα_i2 (PDB: 7WVY; yellow) and Gα_oA(PDB: 8TZQ; brick red). C Concentration–response curves measured using the G protein nanoBRET assay for D2R (left) and D3R (right) co-transfected with Gα_oA, Gα_i2, G_i2G_oA(6), or G_i2G_oA(10). All experiments were conducted in wild-type HEK293T cells stimulated with increasing concentrations of quinpirole. Data were normalized to the signal obtained with GαoA and shown as mean ± SEM. N = 5 independent replicates.

Having established the importance of the Gα C-terminus, we applied a similar approach to investigate the molecular determinants of the coupling selectivity of D3R. The intracellular loop 2 (ICL2) and ICL3 are known to be critical G protein interaction sites^38–41. We therefore aligned the amino acid sequences of the ICL2 and ICL3 regions of D2R and D3R to identify differences and similarities (Fig. 5A). Based on this analysis, we engineered six chimeric receptors by swapping ICL2 and/or ICL3 between D2R and D3R (Fig. 5B). To confirm that these receptor chimeras retained the ability to respond to quinpirole and activate G proteins, we adopted the G protein nanoBRET and used Gα_oA as a positive control. All six chimeric receptors produced a clear BRET signal upon stimulation, confirming they were properly expressed and functional (Fig. 5B). The magnitude of activation was largely consistent with that of the wild-type receptor, except for the D3R^D2-ICL3 chimera that showed significantly reduced coupling to Gα_oA (Fig. 5B). Although further structural modeling is needed, we speculate that the reduced signaling may result from steric clashes, as the D3R^D2-ICL3 chimera combines the longer ICL2 of D3R with the longer ICL3 of D2R. Nonetheless, this control confirmed that all chimeric receptors were functionally competent. We then co-transfected the wild-type and chimeric receptors with Gα_i2 or Gα_z and analyzed their response to quinpirole (Fig. 5B). The results revealed a striking pattern that depended exclusively on the origin of the ICL2. Every transfected receptor containing the ICL2 of D2R, such as wild-type D2R, D2R^D3-ICL3, D3R^D2-ICL2, and D3R^D2-ICL2/ICL3, showed robust coupling to Gα_i2. Conversely, every receptor containing the ICL2 of D3R, such as wild-type D3R, D3R^D2-ICL3, D2R^D3-ICL2, and D2R^D3-ICL2/ICL3, showed no detectable Gα_i2 activation. As expected, activation of Gα_z was consistent between D2R- and D3R-derived receptor chimeras (Fig. 5B). These findings pinpoint ICL2 as the primary molecular switch preventing the interaction between D3R and Gα_i, thereby enforcing the selective signaling profile observed for D3R throughout this study.

Fig. 5. Involvement of dopamine receptor intracellular loops in defining G protein selectivity.

A Sequence alignment of intracellular loop 2 (ICL2) and intracellular loop 3 (ICL3) of human D2R and D3R, highlighting the degree of conservation across the aligned sequences: *=full conservation,.=strong conservation,:=weak conservation. B Concentration–response curves obtained using the G protein nanoBRET assay for Gα_oA and Gα_i2 activation by wild-type and loop-swapped D2R and D3R constructs. Schematic representations of the used chimeric receptor constructs in which ICL2 and/or ICL3 of D2R and D3R were swapped are inserted in each graph. All experiments were conducted in wild-type HEK293T cells stimulated with increasing concentrations of quinpirole. Reference curves for wild-type D2R and D3R with Gα_oA, Gα_i2, and Gα_z are reproduced multiple times to facilitate comparisons. Data were normalized to the signal obtained with Gα_oA and shown as mean ± SEM. N = 5 independent replicates.

An investigation of the recently published structure of D3R bound to the Gα_oA-subtype of G protein (PDB:9F33) revealed that G350 is positioned in close proximity to a negatively charged cavity on D3R (Fig. 6A). The presence of an Asp residue in this location in Gα_i might lead to electrostatic repulsion, destabilizing the interaction. In contrast, in D2R bound to Gα_i1 (PDB: 7JVR), the corresponding Asp residue (D350) is stabilized by a surrounding positively charged cavity (Fig. 6A). In addition to this, T142^ICL2 on D3R is reported to get phosphorylated⁴², and the bulky phosphate group can be accommodated by a Gly residue. However, the presence of an Asp residue in this position would not only lead to electrostatic repulsion but also a steric clash. Interestingly, in the reported structures of D2R bound to either Gα_oA (PDB:8TZQ) or Gα_i1 (PDB:7JVR), the corresponding T144^4.34 exhibits a differential rotamer and points away from the Gα subunit (Fig. 6B). Superimposing the structures reveals that A345, R349, and Y354 occupy a similar position in both D2R and D3R (Fig. 6C) as well as the corresponding residues in Gα_i1 in D3R-bound Gα_i1 (Fig. 6C). While Y354 could potentially facilitate additional interactions with the surrounding positively charged pocket as compared to F354, the electron cloud over the benzene ring in both helps stabilize the G proteins in their corresponding receptor cavities. It is worth noting that several structures of D3R in complex with Gα_i1 have been recently published^30,33. A purified system may allow protein-protein interactions that would otherwise not be observed in the presence of other interacting partners. Additionally, the reported complexes have been purified from insect cells and stabilized using antibody fragments, which might facilitate the generation of stable complexes that may be less prevalent in mammalian systems. Indeed, G protein recruitment assays carried out in Sf9 insect cells report the interaction of Gα_i1 with D3R³⁰.

Fig. 6. Structural comparison of the G protein binding cavity of D2R and D3R.

A Surface representation of the charge distribution in the pocket occupied by Gα_oA in D3R (top) and Gα_i1 in D2R (bottom). Blue represents positively charged surfaces while red represents negatively charged surfaces. B T142^ICL2 faces towards the α5-helix in D2R, while T144^4.34 in D3R exhibits a different rotamer, facing away from the α5-helix. C Superimposition of the Gα subunit in the various structures reveals the orientation of the differing residues. (D3R-Gα_oA: PDB – 9F33, D3R – cornflower blue, Gα_oA – rosy brown; D2R- Gα_oA: PDB – 8TZQ, D2R – pale violet red, Gα_oA – gray; D2R- Gα_i1: PDB – 7JVR, D2R – brick red, Gα_i1 – peach puff).

We undertook site-directed mutagenesis to generate mammalian expression clones encoding point-mutated versions of D3R (D3R^T142A) and Gα_i2 (Gα_i2^D351G). These mutants were evaluated using the G protein nanoBRET assay (Supplementary Fig. 4). Introduction of a glycine at position 351 of Gα_i2, the residue that is found at the corresponding position in Gα_o, was efficiently activated by D2R (Supplementary Fig. 4A), but did not produce a BRET signal upon D3R activation (Supplementary Fig. 4B). Likewise, the D3R^T142A mutant failed to produce D3R-mediated activation of Gα_i2 (Supplementary Fig. 4C). Together, these results indicate that the receptor: G protein interaction interface requires contributions from multiple sites to support stable complex formation and to efficiently promote GDP-to-GTP exchange on Gα_i2 and heterotrimer dissociation.

The unique coupling profiles of D2R and D3R produce distinct responses in hippocampal CA1 neurons

To examine the unusual coupling profile of D3R in a physiological context, we focused on cAMP signaling in CA1 hippocampal neurons that richly express D2R and D3R⁴³. We crossed the cAMP Encoded Reporter (CAMPER) mice, which conditionally express the ^TEPac^VV FRET-based cAMP reporter⁴⁴, with the CaMKIIα-Cre strain to achieve expression in the hippocampus⁴⁵. Indeed, 2-photon imaging in acute brain slices from CaMKIIα-Cre: CAMPER^+/– revealed robust biosensor expression in the CA1 region (Fig. 7A). We first used the G protein nanoBRET assay to test three antagonists, PG01037, L-741626, and NGB-2904, that are reported as selective towards D2R or D3R^46–48. We found that PG01037 inhibited both D2R and D3R, while L-741626 was selective towards D2R over D3R (Supplementary Fig. 5A–C). To test the role of D2R and D3R in regulating AC, we therefore used a combination of selective agonists and antagonists (Fig. 7B). Accordingly, we used bath application of agonists at a low concentration (100 nM) to target D2R (sumanirole) or D3R (ML417) activation (Fig. 7C). The Gα_s-coupled β-adrenergic receptor was stimulated with isoproterenol (100 nM) as a control (Fig. 7C). Sumanirole induced robust inhibition of cAMP, which is consistent with D2R-mediated Gα_i activation (Fig. 7D). On the other hand, ML417 application paradoxically led to an increase in cAMP production (Fig. 7D). As a control, the recording buffer alone exerted no change in cAMP dynamics. Subsequent application of isoproterenol increased cAMP in both the buffer and ML417 experiments, the effect of isoproterenol was significantly greater after ML417 priming (Fig. 7D). Pre-treatment with sumanirole completely blocked the effect of isoproterenol (Fig. 7D). Then, we repeated these experiments with the co-application of receptor antagonists. We found that both L-741626 (10 µM) and PG01037 (10 µM) blocked sumanirole-mediated cAMP inhibition, suggesting that sumanirole operates through D2R-Gα_i in CA1 neurons. Moreover, isoproterenol-mediated cAMP synthesis was not impacted by either antagonist (Fig. 7E-F). We next observed that L-741626 had no effect on the ML417-mediated sensitization of isoproterenol-induced cAMP production. On the other hand, PG01037 blocked ML417-induced sensitization of cAMP following isoproterenol application (Fig. 7E-F). We interpret the results such that PG01037 blocks ML417 on D3R. Meanwhile, L-741626 blocks D2R, which is not robustly activated by ML417 at 100 nM in CA1 neurons. Altogether, these data suggest that sumanirole strongly inhibits AC activity via D2R-Gα_i as cAMP inhibition persisted even in the presence of isoproterenol, while the ability of ML417 to enhance cAMP in a way that is blocked by PG01037 supports the notion that D3R-Gα_o leads to release of Gβγ to conditionally sensitize AC activity⁴⁹. An alternative explanation is that ML417 could act as a weak partial agonist at D2R, thereby reducing the tonic AC inhibition mediated by D2R. In contrast to this hypothesis, we found that 100 nM ML417 does not inhibit dopamine-mediated responses (Supplementary Fig. 6). Collectively, these results capture how the exquisite coupling profiles of D2R and D3R contribute toward neuronal signal transduction.

Fig. 7. Agonism of D2R, but not D3R, inhibits cAMP signaling in CA1 hippocampal neurons.

A Schematic of the approach to visualize dorsal hippocampal CA1 neuronal cAMP dynamics in acute brain slices from CaMKIIα-Cre:CAMPER+/– adult mice. Created with Adobe Illustrator. Scale bar represents 100 microns. B Schematic illustration of D2R and D3R signal decoding in dorsal hippocampal CA1 neurons. C Average trace of cAMP responses to agonist stimulation with ML417 (100 nM; n = 35 neurons/6 mice), Sumanirole maleate (100 nM; n = 38 neurons/6 mice), or control buffer (n = 35 neurons/6 mice). In the same experiment, slices were then immediately stimulated with Isoproterenol (Iso; 100 nM). D left) Maximum agonist-induced cAMP amplitude calculated as peak agonist response minus basal cAMP level. One-way ANOVA was performed with Dunnett’s multiple comparisons test to buffer control. F = 99.23, R2 = 0.6331, F (DFn, DFd) = 1.365 (2, 115). right) Maximum Iso induced cAMP amplitude calculated as peak iso response minus peak agonist response. One-way ANOVA was performed with Dunnett’s multiple comparisons test to buffer control. F = 239.4, R2 = 0.8063, F (DFn, DFd) = 7.619 (2, 115). Data shown as mean ± SEM, where each dot represents an individual neuron. Exact P values are depicted on the graphs. E Average trace of cAMP responses to stimulation of: L-741626 (10 µM) with Suminarole (100 nM; n = 16 neurons/4 mice), PG01037 (10 µM) with Suminarole (100 nM; n = 20 neurons/4 mice), L-741626 (10 µM) with ML417 (100 nM; n = 24 neurons/5 mice), or PG01037 (10 µM) with ML417 (100 nM; n = 18 neurons/4 mice). Slices were then stimulated with Isoproterenol (Iso; 100 nM). F Maximum Iso induced cAMP amplitude. One-way ANOVA was performed with Dunnett’s multiple comparisons test. F = 4.781, R2 = 0.1624, F (DFn, DFd) = 1.603 (3, 74). Data shown as mean ± SEM where each dot represents an individual neuron. Exact P values are depicted on the graphs.

Discussion

Our study conclusively demonstrated that D3R is incapable of productive coupling with Gα protein subtypes Gα_i1, Gα_i2, and Gα_i3, while it effectively couples to and activates Gα_o and Gα_z. Although previous studies have partially addressed the G protein coupling selectivity of D3R, they produced inconsistent findings across different assay platforms, likely because of using modified systems that did not employ wild-type G proteins^26,27,30,33. Results from Lane and colleagues suggested that D3R preferentially couples to Gα_oA, with minimal or no activation of Gα_i1–3, while Gα_z was not assessed²⁶. These conclusions were based on two approaches. First, tetracycline-inducible, PTX-resistant Gα subunits were expressed in HEK293 cells constitutively expressing D2R or D3R, and receptor activity was measured using a [³⁵S]GTPγS binding assay after dopamine stimulation. Although D3R appeared not to couple to Gα_i2, induction of Gα_i2 also increased basal activity, complicating interpretation. Moreover, the engineered G proteins contained a mutation in the α5 helix, a key region for receptor interaction, raising concerns about the artificial system. Second, receptor–G protein fusion constructs were used to control stoichiometry, but elevated basal activity of D3R and α5 helix modifications again left the question of true Gα_i coupling unresolved. An independent study using miniG protein recruitment and G-CASE assays reported conflicting coupling profiles²⁷. MiniG assays detected miniG_o recruitment by D2R, D3R, and D4R, but miniG_i1 only by D2R, whereas G-CASE suggested broader coupling of all three receptors to the Gα_i/o/z family. However, both approaches have limitations: miniG proteins lack the GTPase domain, increasing promiscuity⁵⁰, and G-CASE relies on modified G proteins, with tagging that may affect sensor performance²⁷. Recent cryo-EM studies have resolved D3R-Gα_i complexes; however, such static structures can misrepresent physiological coupling, especially when stabilized using nanobodies, detergent micelles, or mutated receptor constructs^30,33. Notably, G protein-specific signaling studies associated with these structures were obtained either from BRET-based G-protein dissociation assays or from NanoBiT recruitment experiments. NanoBiT G protein recruitment assays were conducted in insect cells co-expressing the human DRD3 receptor fused at its C-terminus to LgBiT via a 15–amino acid flexible linker, together with Gα_i1 and SmBiT-tagged human Gβ₁ and Gγ₂³⁰. While this approach detects G protein recruitment to the receptor, it does not assess functional dissociation of the G protein heterotrimer. Additionally, the assay was performed in Sf9 insect cells, which lack mammalian post-translational modifications that may be essential for authentic receptor behavior. In a separate study, the same group used a GloSensor-cAMP assay relying on endogenous G proteins and, consistent with our results (Fig. 3B), observed inhibition of cAMP accumulation by D3R³³. Proximal G protein dissociation assays, such as the TRUPATH assay, rely on engineered Gα and Gγ subunits fused to a BRET donor–acceptor pair and have likewise been used to characterize D3R coupling^28,31. Applying this approach, one study reported that D3R fails to couple to Gα_i2, while retaining coupling to Gα_i1 and Gα_i3³¹. Nonetheless, using similar strategies, the TRUPATH assay developers confirmed our observations of D3R lacking Gα_i1/i2/i3 coupling capability²⁸. To circumvent potential artifacts arising from modified G proteins or receptor constructs, we utilized wild-type Gα subunits expressed in both parental and ΔGα_i/o/z HEK293 cells and assessed D3R function across two complementary functional assays. This experimental design, combined with the inclusion of multiple positive and negative controls, minimized artifacts associated with engineered or truncated Gα proteins and enabled a more physiologically relevant evaluation of receptor coupling selectivity. Our findings confirmed that D2R and D4R couple broadly to all Gα_i/o/z subtypes, while D3R exhibits exclusive coupling selectivity for Gα_oA, Gα_oB, and Gα_z, with no detectable engagement with Gα_i1, Gα_i2, and Gα_i3 under any tested condition.

Employing both G protein and receptor chimeras, we explored the molecular basis underlying this coupling selectivity. We confirmed the importance of the interface between the C-terminal α5 helix of the Gα proteins and the ICL2 of D3R in defining the biased coupling of D3R toward Gα_o, while we showed that the ICL3 does not contribute to this selectivity. This is in conflict with previous findings that introducing 12 amino acids from the C-terminal segment of the ICL3 of D2R into the corresponding region of D3R partially restored Gα_i coupling²⁶. Such discrepancies may arise from engineered features that bypass natural coupling constraints, highlighting the need for further molecular investigations. Models of the receptor-G protein interaction, based on available cryo-EM structures^30,33, suggest that steric clashes and electrostatic repulsion may be responsible for the inability of D3R to engage Gα_i proteins. In particular, our analysis points to the involvement of D351 in Gα_i2 and T142^ICL2 in D3R, especially if phosphorylated by GRK2 as previously reported⁴². However, experimental probing revealed that it is not a single residue responsible for the coupling preference exhibited by D3R, but rather a combinatorial effect of the different residues.

The Gα_o/z-biased coupling of D3R has significant physiological implications. Although Gα subunits within the Gα_i/o/z family share high sequence similarity, they can engage distinct effectors and mediate non-redundant functions, as shown by the divergent phenotypes observed in Gα_o, Gα_i, and Gα_z knockout models and by recent studies characterizing unique intracellular effectors^49,51–56. The restricted coupling profile of D3R suggests it may activate a distinct subset of signaling pathways limited to Gα_o/z and Gβγ-mediated responses, or it may bias signaling by excluding Gα_i subtypes, thereby enhancing the activation of alternative effectors, such as GIRK channels or Gβγ-sensitive adenylyl cyclases. As a promising target for treating substance use and other neuropsychiatric disorders^11,57,58, D3R offers an additional layer of selectivity by modulating specific signaling pathways within defined cellular or subcellular contexts. Such enhanced selectivity holds the promise for the development of precisely targeted therapeutic strategies with reduced side effects. In this context, receptor-selective agonists are valuable both as research tools and potential treatments, offering improved outcomes compared to non-selective compounds due to differences in both receptor distribution and signaling⁵⁹. Sumanirole, a D2R-selective agonist with higher affinity for D2R over D3R^60,61, exemplifies this approach by selectively initiating D2R-mediated effects in animal models, including prolactin release, regulation of striatal acetylcholine levels, and mediating autoreceptor-driven inhibition of nigrostriatal firing⁶⁰. Similarly, pramipexole, a drug initially developed as a D2R agonist and later found to have high potency and selectivity for D3R⁶², has a demonstrated ability to delay L-DOPA-induced dyskinesia in the treatment of Parkinson’s disease^62–64, suggesting that its clinical benefits may be partly due to D3R unique signaling properties. Although evidence for D3R-specific physiological effects is still emerging, the recent development of ML417 as a highly selective D3R agonist represents a major step forward in understanding D3R function and therapeutic potential³⁴. ML417 has already been shown to possess neuroprotective properties, along with promising safety and pharmacokinetic profiles that would make it a strong lead compound for further drug development. Overall, D3R represents a unique case among GPCRs as no other GPCR with a primary coupling to Gα_i/o/z proteins has been reported to be incapable of activating a subset of the family members. These insights not only clarify previous inconsistencies in the literature but also open new avenues for selective D3R pharmacology.

Methods

Ethics statement

Procedures involving mice strictly followed NIH guidelines and were approved by the Institutional Animal Care and Use Committee at Augusta University. We have complied with all relevant ethical regulations for animal use. Mice were housed at a consistent temperature with unlimited access to food and water under a constant 12-h light/dark cycle. The following previously described strains were utilized: (i) cAMP Encoded Reporter (CAMPER) (C57BL/6-Gt(ROSA)26Sortm1(CAG-ECFP*/Rapgef3/Venus*)Kama/J) (RRID:IMSR_JAX:032205)⁴⁴, (ii) CaMKIIα-Cre (B6.Cg-Tg(Camk2a-cre)T29-1Stl/J) (RRID:IMSR_JAX:005359)⁴⁵. The two strains were crossed to obtain experimental CaMKIIα-Cre:CAMPER^+/– mice. Both sexes were used and animal identification was verified by automated genotyping PCR (Transnetyx).

DNA plasmids and chemicals

The plasmids encoding human codon-optimized DRD2, DRD3, and DRD4 were subcloned from the PRESTO-Tango library (Addgene; #66269, #66270, and #66271), a generous gift from Dr. Bryan Roth (UNC, Chapel Hill, NC). The subcloning involved the removal of the V2-tail and tTA sequences from receptor C-termini, but leaving the N-terminal HA signal peptide, followed by a FLAG tag before the receptor coding sequences. Gβ1-venus^156-239 and Gγ2-venus^1-155 were generous gifts from Dr. Nevin A. Lambert (Augusta University, Augusta, GA). Gα proteins and masGRK3CT-Nluc constructs were generous gifts from Dr. Kirill A. Martemyanov (UF Scripps Biomedical Research, Jupiter, FL). pGloSensor™-22F was purchased from Promega. Dopamine hydrochloride (#3548), quinpirole hydrochloride (#1061), forskolin (#1099), PG-01037(#3887), NGB-2904(#2635), and L-741,626 (#1003/10) were purchased from Tocris, while rotigotine (#HY-75502), sumanirole maleate (HY-70081A), and ML417 (HY-136390) were purchased from MedChemExpress. All chemicals were resuspended according to manufacturers’ instructions, aliquoted, and stored at –20 °C until use. The constructs for βarr2-LgBiT and receptor-SmBiT were generous gifts from Dr. Andy Chevigné (Luxembourg Institute of Health, Luxembourg). DRD2, DRD3, and DRD4 were sub-cloned into the SmBiT vector with an N-terminal FLAG tag using In-Fusion technology (Takara Bio).

Cell cultures and transfections

HEK293T/17 cells were purchased from ATCC, while Gα_i/o/z-deficient HEK293A cells (ΔGα_i/o/z), a generous gift from Dr. Asuka Inoue (Tohoku University, Japan), were previously generated using three rounds of CRISPR–Cas9 mutagenesis and validated⁶⁵. Cells were cultured in DMEM (Gibco, 10567-014) supplemented with 10% FBS (Biowest, S1520), non-essential amino acids (Gibco, 11140-050), 100 units/ml penicillin and 100 µg/ml streptomycin (Gibco, 15140-122), and 250 µg/ml amphotericin B (ThermoFisher, 15290-018) at 37 °C and 5% CO₂. For transfection, two million cells were seeded in each well of 6-well plates in 1.5 ml medium per well without antibiotics and containing 10% dialyzed FBS (Biowest, S181D) for 4 h and then transfected with a mixture containing a 1:3 ratio of DNA plasmid (2.5 µg) and polyethylenimine (PEI; 7.5 µl) (VWR, AAA43896) in a final volume of 500 µL per well. The following amount of each plasmid was transfected, unless otherwise indicated: 0.21 µg GPCR (D2R, D3R, or D4R), 0.83 µg Gα protein (Gα_oA, Gα_oB, Gα_i1, Gα_i2, Gα_i3, Gα_z, Gα_s, Gα_q, Gα₁₂), 0.21 µg Gβ1-venus^156–239, 0.21 µg Gγ2-venus^1–155, and 0.013 µg masGRK3CT-Nluc. An empty vector, pcDNA3.1, was used to normalize the ratio of transfected plasmids. Transiently transfected cells were incubated for 18–22 h before being tested.

CHO cells were cultured in DMEM (Gibco, 10567-014) supplemented with 10% FBS (Biowest, S1520), non-essential amino acids (Gibco, 11140-050), 100 units/ml penicillin and 100 µg/ml streptomycin (Gibco, 15140-122), and 250 µg/ml amphotericin B (ThermoFisher, 15290-018) at 37 °C and 5% CO2. For transfection, two million cells were seeded in each well of 6-well plates in 1.5 ml medium per well without antibiotics and containing 10% dialyzed FBS (Biowest, S181D) for 4 h and then transfected with a mixture containing 2.5 µg of DNA, 3 µL PLUS reagent, and 6 µL LTX (Invitrogen, 15338100) in a final volume of 500 µL per well.

G protein nanoBRET assay

The day after transfection, HEK293T/17 cells were briefly washed with PBS, resuspended in BRET buffer (PBS supplemented with 0.5 mM MgCl₂ and 0.1% glucose), collected in 1.5 ml tubes, and centrifuged for 5 min at 500 x g. Pelleted cells were resuspended in 300 µl of BRET buffer, and 25 µl of cells were plated in 96-well white microplates (Greiner Bio-One). The Nluc substrate Hikarazine-103 was purchased from Synthelis and used according to the manufacturer’s instructions. BRET measurements were obtained using a POLARstar Omega microplate reader (BMG Labtech). All measurements were performed at room temperature, and the BRET signal was determined by calculating the ratio of the light emitted by Gβ1γ2-venus (collected using the emission filter 535/30) to the light emitted by masGRK3CT-Nluc (475/30). In kinetics assays, the baseline value (basal BRET ratio) was averaged from recordings of the five seconds before agonist injection. In concentration-response experiments, 25 µl of cells per well were plated and mixed with the Nluc substrate. Initial readings were performed to establish basal BRET ratio, and then agonists were added. The BRET signal was recorded for 60 s. ΔBRET ratios were obtained by subtracting the basal BRET ratio from the maximal amplitude measured. G protein nanoBRET assays performed in CHO cells follow the same protocol of HEK293T/17 cells, with the exception that the day after transfection, cells were collected in BRET buffer supplemented with 12.5 µM EDTA at pH 8.0.

cAMP inhibition assay

1 million ΔGα_i cells were seeded in each well of 6-well plates. 4 h later, cells were transfected with plasmids for mammalian expression of individual dopamine receptors (D2R, D3R, or D4R), pGloSensor™-22F cAMP plasmid (Promega), and individual Gα proteins Gα_oA, Gα_i2, or Gα_z. 18–22 h after transfection, cells were collected in PBS, centrifuged at 500 x g for 5 min at room temperature, and the supernatant was discarded. The pelleted cells were resuspended in 300 μl of BRET buffer (PBS supplemented with 0.5 mM MgCl₂ and 0.1% glucose), and 40 μl was plated into each well of a white 96-well plate, followed by the addition of 10 μl of D-Luciferin potassium salt substrate (GoldBio; #LUCK-100; final concentration 600 µg/ml). After incubation at 37 °C for 1h, the plate was placed in a BMG Omega microplate reader and maintained at 28 °C until a stable baseline value was recorded. Cells were then treated with 1 μM quinpirole for 30 s before adding 0.5 μM forskolin and recording luminescence for 25 min at 72-s intervals. A fold change (FC) was calculated for each sample by dividing the maximum relative light units (RLU) obtained after forskolin application by the average baseline RLU. For the quinpirole concentration-response curve, the FC for samples treated with the vehicle was averaged and considered 100% of forskolin-induced cAMP levels. The FC of all other samples was then normalized to the vehicle control FC average and expressed as a percentage of cAMP inhibition.

βarrestin2 recruitment assay

HEK293T/17 cells were transfected with 1.25 µg receptor-SmBiT, 0.05 µg βarrestin2, and 1.3 µg empty vector (pcDNA3.1) 4 h after seeding in 6-well plates. The following day, cells were collected in PBS, centrifuged, and resuspended in 300 μl of PBS supplemented with 0.5 mM MgCl₂, and 40 μl was plated into each well of a white 96-well plate. Basal luminescence was measured after adding 40 μl of the substrate (Hikarazine-103 diluted 1:100) using a LUMIstar Omega microplate reader (BMG Labtech). Thereafter, either dopamine or quinpirole was added at the indicated final concentration, and luminescence was measured. All readings were measured at room temperature. Fold change was calculated for each receptor and plotted as a concentration-response curve, normalized to the response observed upon treatment with the lowest concentration of ligand (treated as 1).

2-photon FRET cAMP imaging

Acute brain slices were prepared from adult mice (2–4 months of age) weighing 24–33 g, as similarly described^66,67. Anesthesia was rapidly achieved using the drop method with ~3% isoflurane for vapor inhalation⁶⁸ followed by decapitation and rapid extraction of the brain, which was subsequently agarose-mounted on a vibratome (Precisionary VF-310-0Z) in ice-cold oxygenated buffer consisting of (in mM): KCl (2.5), NMDG (93), glucose (25), HEPES (20), sodium ascorbate (5), sodium pyruvate (3), thiourea (2), NaH₂PO₄ (1.2), CaCl₂ (0.5), MgCl₂ (10), NaHCO₃ (30). 300-µm thick coronal slices containing the hippocampus were sectioned and incubated for 1 h at 34 °C in an oxygenated recovery buffer consisting of (in mM): NaCl (126), KCl (2.5), CaCl₂ (2), MgCl₂ (2), NaHCO₃ (18), NaH₂PO₄ (1.2), glucose (10). Slices were then maintained at ambient temperature oxygenated recording buffer consisting of (in mM): NaCl (125), KCl (2.5), CaCl₂ (2), MgCl₂ (2), NaH₂PO₄ (1.25), NaHCO₃ (25), glucose (25). Individual brain sections were then constantly perfused in recording buffer at approximately 2 ml per minute in a recording chamber (Warner Instruments) for FRET imaging on a Zeiss 780 multiphoton confocal microscope (20X W Plan-Apochromat objective). Excitation of the cAMP FRET donor (mTurquoise2) was achieved by tuning a Ti: Sapphire laser (Coherent Chameleon Vision S) to 850 nm. Individual photomultiplier tubes were utilized to simultaneously capture emission from the FRET donor (455-509 nm) and FRET acceptor (Venus; 526-571 nm) at 2.5 s intervals. ML417 (MedChemExpress HY-136390), Sumanirole maleate (MedChemExpress HY-70081A), and isoproterenol hydrochloride (Iso; TCI America I0260) were bath applied as indicated in the text. FRET values were calculated using standard ImageJ tools from the raw fluorophore intensity at the neuronal cell body. Non-responsive cells were excluded based on a previously established cutoff criterion (two multiplied by the standard deviation of the baseline prior to drug treatment)^69,70.

Statistics and reproducibility

Statistical analysis was performed using GraphPad Prism version 10 software. Concentration-response curves were fitted to a sigmoidal four parameter logistic function (variable slope analysis) to quantify agonist potencies (pEC₅₀), maximal responses (E_max). Concentration-response curves were fitted and Hill slope values for each agonist were close to 1. At least three independent biological replicates were used for each experiment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

C.O. designed the research. L.Z., L.F., S.S., Y.L., and B.S.M. performed research and analyses of the data. L.Z. wrote the original manuscript draft. L.Z., L.F., and C.O. wrote the manuscript with input from all authors. C.O. and B.S.M. acquired the funding. All authors read and approved the manuscript.

Peer review

Peer review information

:Communications Biology thanks the anonymous reviewers for their contribution to the peer review of this work. Primary Handling Editors: Dr. Nilanjan Banerjee. A peer review file is available.

Funding

This work was supported by NIH awards R01 DC022104 to C.O and R01 NS129554 to B.S.M.

Data availability

All processed data that support the findings of this study are available within the paper and its supplemental data and deposited on Figshare https://doi.org/10.6084/m9.figshare.29825135. The following plasmids have been deposited in Addgene: G_i2G_oA(6) (ID: 258615); G_i2G_oA(10) (ID: 258616); D2R^D3-ICL3 (ID: 258617); D2R^D3-ICL2 (ID: 258618); D2R^D3-ICL2/ICL3 (ID: 258619); D3R^D2-ICL2 (ID: 258620); D3R^D2-ICL3 (ID: 258621); and D3R^D2-ICL2/ICL3 (ID: 258622). Any additional information required to reanalyze the data reported in this paper is available from the corresponding author upon request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10587-0.