Ligand-specific activation trajectories dictate GPCR signalling in cells

Activation of cell surface receptors by extracellular ligands is the hallmark of cell–cell communication and controls most physiological functions in humans. GPCRs are the largest class of such receptors^1,2. After ligand activation, GPCRs communicate their message into cells by recruiting and activating intracellular G proteins¹⁷. GPCR ligands can activate (agonist) or inactivate (antagonist/inverse agonist) receptors to various extents, and this diversity in ligand efficacy is exploited in drug therapy¹⁸. However, the molecular nature of ligand efficacy and the mechanisms of GPCR activation in living cells remain largely unclear.

Fluorescence spectroscopy studies with purified β₂-adrenergic receptors^19,20,21,22, and more recent nuclear magnetic resonance (NMR) and double electron–electron resonance (DEER) spectroscopy studies, have established that GPCRs do not operate as simple on/off switches but exist in a dynamic equilibrium of multiple inactive and active states^3,4,23,24,25. Partial agonists are believed to stabilize conformational states that are structurally different from those stabilized by full agonists^6,11,26,27. Likewise, single-molecule Förster resonance energy transfer (FRET) and advanced NMR spectroscopy studies have demonstrated that GPCRs can form distinct signalling complexes with G proteins^28,29, in which the G proteins may possess different nucleotide affinities^5,7. Partial agonists have been proposed to stabilize GPCR–G-protein complexes with reduced efficacy towards nucleotide exchange^5,7,30,31. However, all of these studies were performed in isolated systems using purified receptors reconstituted in detergent micelles or nanodiscs. Whether GPCRs in the physiological environment of a living cell also adopt different conformations and form distinct signalling complexes is essentially unclear.

Here we develop a new type of conformational GPCR biosensors to probe the existence and ligand modulation of such GPCR–G-protein-signalling complexes in intact cells. Using the M₂R as a model, we demonstrate that agonist activation leads to formation of an equilibrium of distinct GPCR signalling complexes along ligand-specific activation trajectories. These distinct activation trajectories and the relative abundances of these distinct receptor–G-protein complexes dictate the type and extent to which specific G proteins are activated.

GPCR activation results in large-scale receptor conformational changes^{1,32,33,34,35}. Most prominently, outward movement of the intracellular part of transmembrane domain 6 enables coupling to and activation of G proteins. In turn, G-protein binding promotes conformational changes in the receptor, including its extracellular domains and the ligand-binding pocket, which stabilize agonist binding^1,32,36. This communication between extra- and intracellular receptor domains represents the principle of allosteric coupling.

To track conformational changes at the extracellular surface of GPCRs at a high spatial resolution, we sought to develop a new type of biosensor. These biosensors should be genetically encoded, equipped with minimally sized labels (similar to those used in NMR, DEER and single-molecule FRET studies on isolated receptors) and retain an unmodified intracellular surface to preserve G-protein coupling. We chose the M₂R as a model because structural studies have shown that its extracellular conformational changes on activation, including closure of the binding pocket, are the most pronounced among all class A GPCRs of which the structures have been solved^36,37,38,39.

The least invasive way to attach probes to a GPCR at the single-residue resolution in living cells is by bioorthogonal chemistry on genetically encoded chemical anchors^12,13,14,15. In brief, a non-canonical amino acid (ncAA), also known as an unnatural amino acid, carrying an anchor for rapid catalyst-free labelling is incorporated into the receptor using genetic code expansion technology (GCE). The label is then attached post-translationally by ultrarapid strain-promoted inverse electron-demand Diels–Alder cycloaddition, which occurs within minutes without interfering with native functional groups. Using this strategy, we previously demonstrated quantitative labelling of GPCRs on the live cell surface⁴⁰.

We screened the entire extracellular surface of the M₂R to identify positions at which the click-ncAA trans-cyclooct-2-ene lysine (TCO*K)⁴¹ was efficiently incorporated and yielded robust labelling with a cell-impermeable, tetrazine-conjugated cyanine dye (Tet–Cy3) (Fig. 1a). Of 72 receptor mutants, 25 displayed good cell surface expression and labelling (Extended Data Fig. 1). The other 47 constructs were not expressed, not trafficked to the cell surface or showed no labelling at all (Fig. 1b and Supplementary Fig. 1). As a control, labelling in the absence of TCO*K produced no fluorescence (Supplementary Fig. 2).

a, Genetic incorporation of a ncAA (TCO*K) and bioorthogonal labelling of M₂R with Tet–Cy3 (Methods). Cells expressing biosensors are stimulated by continuous pressurized application of agonist (ACh) or buffer through a manifold tip. Created in BioRender. Thomas, R. (2025) https://BioRender.com/loxqlqf. b, Snake plot of M₂R indicating all positions that were robustly labelled (orange and grey) and showed activation-related changes in fluorescence intensity (orange). The numbers are the residue numbers. Positions that could not be labelled are indicated in white. ECL1, extracellular loop domain 1; ECL2, extracellular loop domain 2; ECL3, extracellular loop domain 3. c, Representative changes in fluorescence intensity (ΔF/F₀) were recorded over time from several individual HEK293T cells expressing M₂R biosensors labelled at the indicated amino acid positions. Cells were superfused with 1 mM ACh. The shaded areas indicate the duration of agonist addition, and the unshaded areas indicate agonist washout with buffer. d, Mean fluorescence intensity changes (ΔF/F₀) of all seven agonist-sensitive biosensors after activation with 1 mM ACh. Positive values indicate an increase in fluorescence; negative values indicate a decrease after ACh superfusion. Data are mean ± s.e.m., with each datapoint representing a single cell. M₂R⁸⁴ (35 cells examined over 17 independent experiments), M₂R¹⁷⁵ (44, 11), M₂R¹⁸¹ (76, 9), M₂R¹⁸⁸ (96, 15), M₂R⁴¹⁴ (33, 7), M₂R⁴¹⁵ (56, 9), M₂R⁴¹⁹ (34, 12). TM2, transmembrane domain 2; TM5, transmembrane domain 5; TM7, transmembrane domain 7. e, Top view of the X-ray crystal structure of the active M₂R (blue; Protein Data Bank (PDB): 4MQS). The positions of incorporated TCO*K yielding GPCR biosensors are colour coded according to the gradient, representing the mean ACh-induced changes in fluorescence intensity (ΔF/F₀). For comparison, the X-ray crystal structure of the inactive M₂R (PDB: 3UON) is shown in grey. The roman numerals indicate the number of the transmembrane helix. The constructs used were SP-M₂R^XXXTAG (Methods).

Source data

Next, we applied the endogenous M₂R agonist acetylcholine (ACh) to single cells expressing one labelled construct each using a superfusion device (Fig. 1a). Seven of the 25 Cy3-TCO*K-M₂R constructs exhibited robust and reproducible changes in their fluorescence emission intensities (Fig. 1c,d), therefore serving as reporters for M₂R activation. We quantified the labelling efficiency at these M₂R biosensors using fluorescence correlation spectroscopy^40,42. Four M₂R biosensors showed quantitative Cy3 labelling (positions Thr84 (M₂R⁸⁴), Glu175, Ala414 and Pro415) and three reached labelling efficiencies of 60–80% (positions Phe181, Phe188 and Asn419; Extended Data Fig. 1). Importantly, all seven M₂R biosensors robustly activated G proteins and internalized after ACh exposure, indicating full functionality (Extended Data Fig. 2). Notably, labelling position Ala414 with either the rhodamine dye TAMRA or the cyanine dye Cy5 also resulted in M₂R activation biosensors (Supplementary Fig. 3).

Cyanine fluorophores such as Cy3 are environmentally sensitive^23,42,43. We propose that the observed fluorescence changes arise from local microenvironmental changes (such as transitions to either more hydrophobic or to more polar microenvironments) caused by ligand-promoted receptor conformational changes. Mapping the position of ACh-sensitive labels onto high-resolution structures of the inactive and active M₂R (Fig. 1e) shows that all positions featuring ACh-promoted fluorescence changes lie in extracellular receptor domains that move during receptor activation^37,38,39. All seven biosensors retain the ability to respond to the positive allosteric modulator LY2119620 (ref. ³⁸), which binds at the extracellular allosteric binding site (Extended Data Fig. 3). Thus, we infer that the M₂R biosensors report on activation-related conformational changes of the receptor. Supporting this, binding of the antagonist N-methylscopolamine did not induce fluorescence changes at any of these positions (Extended Data Fig. 2).

ACh-stimulated fluorescence changes varied in both direction (fluorescence increase or decrease) and amplitude (ΔF/F₀, −16% to +10%) (Fig. 1c,d and Supplementary Table 1). All changes were strictly dependent on the presence of ACh and returned to the baseline after ACh washout (Fig. 1c). Moreover, ACh-mediated changes in fluorescence were also concentration dependent, with potencies matching reported ACh affinity values (around 1–6 µM; Extended Data Fig. 4), which further supports the specificity of the observed effects.

The M₂R biosensor panel enables monitoring receptor conformational changes across the entire extracellular surface (Fig. 1e) while leaving the intracellular domains, which couple to G proteins, untagged. This makes it an ideal tool for investigating the existence of distinct, ligand-specific GPCR active states and their coupling to G proteins in intact cells. To study agonist effects, we selected M₂R agonists differing in their reported efficacies in classical pharmacological assays: the endogenous full agonist ACh, the superagonist iperoxo^39,44,45,46, and the partial agonists arecoline and pilocarpine. All of the agonists stimulated G-protein activation at all seven biosensors in a concentration-dependent manner, similar to wild-type (WT) M₂ receptors (Extended Data Fig. 5). We noted that higher agonist potency tended to correlate with a smaller dynamic range of the functional response (Extended Data Fig. 4i). This would be compatible with the notion that some biosensors might display slightly different spontaneous activity compared with WT M₂ receptors. Only at the M₂R¹⁸¹ biosensor were all of the agonists less potent and less efficacious. Nonetheless, the rank order of agonist potencies and efficacies at all biosensors matched WT receptors (Extended Data Fig. 5 and Supplementary Table 2). This demonstrates that conformational data obtained with the panel of biosensors can be projected to WT receptors.

Stimulation of the M₂R biosensor family with a saturating concentration of iperoxo changed fluorescence emission at six out of the seven ACh-responsive M₂R biosensors (Fig. 2 and Extended Data Fig. 6). Compared with ACh, iperoxo produced larger responses at five biosensors (M₂R⁸⁴, M₂R¹⁷⁵, M₂R⁴¹⁴, M₂R⁴¹⁵ and M₂R⁴¹⁹), but smaller fluorescence changes at the M₂R¹⁸⁸ biosensor. Notably, the M₂R¹⁸¹ biosensor did not respond at all (Fig. 2a,b and Extended Data Fig. 6) at any of the sequences of agonist addition (Supplementary Fig. 4). Absolute ΔF/F₀ amplitudes after iperoxo exposure (Fig. 2b and Supplementary Table 1) were normalized to ACh at each M₂R biosensor and plotted in a radar plot (Fig.