Conformational biosensors reveal adrenoceptor signalling from endosomes

Ligand binding to the extracellular surface of the β₂-adrenoceptor (β₂-AR) produces an activating conformational change in the receptor that promotes guanine nucleotide dissociation from the cytoplasmic GTP-binding protein Gα_s; this represents the critical biochemical event initiating classical GPCR signal transduction (Fig. 1a)¹³. Activated β₂-ARs are substrates for phosphorylation and binding of β-arrestins, events which inhibit interaction with G proteins and promote endocytosis of receptors via clathrin-coated pits (CCPs)^14,15. Acute β₂-AR Gα_s signalling is thus traditionally thought to be restricted to the plasma membrane^14,16,17. However, to our knowledge, this assumption has not been directly tested. To do so, we generated a biosensor of activated β₂-AR based on a conformation-specific single-domain camelid antibody (Nb80) used in recent structural studies^18,19. We reasoned that this nanobody, which selectively binds the agonist-occupied β₂-AR and is able to stabilize an activated receptor conformation when present in vitro at high concentration, might act as a sensor of receptor activation when expressed at relatively low concentration in intact cells (Fig. 1b). This proved to be the case; in cells maintained in the absence of agonist, Nb80 fused to enhanced GFP (Nb80–GFP) localized to the cytoplasm and not with β₂-ARs present in the plasma membrane(Fig. 1c, 0 min, top row; Pearson’s coefficient = 0.135). Line scan analysis verified the cytoplasmic distribution of Nb80–GFP before β₂-AR activation (Fig. 1d, top row) was as expected because the cytoplasmic concentration of Nb80–GFP achieved in our experiments (approximately equal to 20 nM) was considerably lower than the equilibrium dissociation constant estimated in vitro for Nb80 binding to purified β₂-ARs in the absence of agonist (0.76 ± 0.14 µM; Supplementary Fig. 1a–d). After application of the adrenergic agonist isoprenaline (10 µM), Nb80–GFP was rapidly recruited to the plasma membrane and co-localized there with β₂-ARs (Fig. 1c, middle row; Pearson’s coefficient = 0.625). Line scan analysis verified robust Nb80–GFP recruitment to the plasma membrane and concomitant depletion from the cytoplasm (Fig. 1d, middle row), consistent with the much higher affinity of Nb80 for isoprenaline-activated β₂-ARs (2.9 ± 0.5 nM; Supplementary Fig. 1d). Agonist-induced membrane recruitment of Nb80–GFP was specific because the D1 Dopamine receptor (DRD1), which is also G_s-coupled but does not bind Nb80 in vitro (data not shown), failed to recruit Nb80–GFP to the plasma membrane in response to dopamine (10 µM) application (Supplementary Fig. 2). Furthermore, β₂-AR–CFP and Nb80–YFP generated a pronounced fluorescence (Förster) resonance energy transfer (FRET) signal after isoprenaline application whereas DRD1–CFP did not (Supplementary Fig. 3a, b).

β₂-AR internalization began 1 to 2 min after Nb80–GFP recruitment to the plasma membrane, indicated by the emergence of surface-labelled β₂-AR in peripheral cytoplasmic vesicles. Nb80–GFP did not co-localize with β₂-AR-containing endocytic vesicles upon first appearance (Fig. 1c, middle row, arrow in merged image points to an example) but was recruited at later time points (Fig. 1c, bottom row, Pearson’s coefficient = 0.702; examples are indicated by arrowheads). Endosome recruitment of Nb80–GFP was evident by line scan analysis (Fig. 1d, bottom row; line scans are from the representative individual examples with further quantification in legend) and localized to EEA1-marked early endosomes (Pearson’s coefficient = 0.846; Supplementary Fig. 4) through which β₂-ARs iteratively cycle in the presence of agonist²⁰. β₂-AR-containing endosomes were initially devoid of Nb80–GFP and later acquired Nb80–GFP during their movement (Supplementary Videos 1 and 2). Interaction at endosomes was verified by β₂-AR–CFP and Nb80-YFP normalized FRET (nFRET) (Supplementary Fig. 3c). These results suggest that β₂-AR activation initiates a precisely choreographed series of events: Nb80–GFP is first recruited from the cytoplasm to the plasma membrane, then β₂-ARs internalize devoid of Nb80–GFP, followed by a second phase of Nb80–GFP recruitment to the internalized β₂-ARs.

Nb80–GFP recruitment to endosomes required β₂-ARs because a phosphorylation-deficient mutant version of the β₂-AR (β₂AR-3S) that couples to Gα_s but is impaired in agonist-induced endocytosis²¹ recruited Nb80–GFP to the plasma membrane but produced much less recruitment to endosomes (Fig. 1e, top row). Nb80–GFP co-localized with β₂-AR-3S after agonist-induced activation (Pearson’s coefficient = 0.674) but this was largely restricted to the plasma membrane (line scan analysis in Fig. 1f is representative; further quantification in the figure legend). Nb80–GFP membrane recruitment was also reversible because the biosensor returned to a cytoplasmic distribution after addition of the competitive antagonist CGP-12177 (Fig. 1e, f, middle and bottom rows; Pearson’s coefficient = 0.106), verified by recovery of Nb80–GFP fluorescence intensity in the cytoplasm (Fig. 1g; see also Supplementary Video 3). Thus, Nb80–GFP detected activated β₂-ARs both in the plasma membrane and endosomes after acute agonist application.

Discrete Nb80–GFP recruitment phases were clearly resolved by total internal reflection fluorescence (TIRF) microscopy that selectively detects events occurring in the plasma membrane and extending approximately equal to 100 nm into the peripheral cytoplasm²². First, within 2 min after agonist application, Nb80–GFP was progressively recruited to the plasma membrane in a diffuse distribution (Fig. 2a, compare first and second columns from left; quantification in Fig. 2b). Nb80–GFP did not co-localize with β₂-ARs when they clustered in relatively static punctae characteristic of receptor-containing CCPs²³ (Fig. 2a, third column). Second, over a period of several additional minutes, Nb80–GFP was recruited to a discrete population of highly mobile punctae (Fig. 2a, fourth column) representing peripheral β₂-AR-containing endocytic vesicles²³.

Sequential TIRF imaging emphasized the distinction between relatively static β₂-AR puncta not co-localizing with Nb80–GFP (Fig. 2c, arrowhead indicates an example) and mobile endosomes that did (Fig. 2c, arrow indicates a representative example; many are visible in Supplementary Video 4). Later recruitment of Nb80–GFP occurred rather suddenly, typically within approximately 5 s (Supplementary Video 5), as also evident in kymographs of individual endosome trajectories (Fig. 2d). Ruling out potential artefacts of wavelength-dependent differences in TIRF microscopy illumination depth, later recruitment of the biosensor to endosomes was similarly observed when the excitation wavelengths used to detect receptor and biosensor were reversed (Fig. 2e).

Nb80–GFP did not detectably concentrate in CCPs labelled with either of two independent markers, the adaptor protein β-arrestin (Fig. 3a and Supplementary Video 6) or the coat protein component clathrin light chain (Fig. 3b and Supplementary Video 7). Representative examples are shown and this was verified across multiple cells and experiments (Pearson’s coefficient = 0.365 and 0.319, respectively, with numbers of replicates specified in the legend). In contrast, and as expected based on previous work, extensive co-localization of β₂-AR with β-arrestin was observed under the same conditions (Pearson's coefficient = 0.677). Separation of Nb80–GFP localization from that of either β-arrestin or clathrin was clear in line scans (bottom panels of Fig. 3a, b were derived from the images shown and are representative; further quantification is in the legend).

A simple interpretation is that Nb80–GFP associates with activated β₂-ARs in the plasma membrane but then dissociates before receptors cluster in CCPs and internalize. This was surprising because β₂-AR clustering in CCPs occurs much more rapidly²³ than the overall reversal rate of Nb80–GFP recruitment observed after agonist washout in intact cells (Fig. 1g), or the kinetics of Nb80 dissociation from purified β₂-ARs in vitro (Supplementary Fig. 1b, e). One possibility is that Nb80-GFP dissociation is accelerated during the clustering process, by mechanisms such as by receptor phosphorylation or steric exclusion mediated by β-arrestins or other CCP-associated components. An alternative possibility is that the fraction of β₂-ARs that have bound Nb80–GFP in the plasma membrane by the time of the clustering reaction are unable to enter CCPs, and only those β₂-ARs not initially bound to Nb80–GFP in the plasma membrane are able to cluster in CCPs and subsequently internalize. In either case, the data clearly indicate that Nb80–GFP associates with β₂-ARs after endocytosis, and after uncoating of the endocytic vesicle has occurred (Fig. 3c). Accordingly, Nb80–GFP recruitment to β₂-AR-containing endosomes cannot represent an artefact of persistent nanobody binding from the plasma membrane; instead, this observation reveals that β₂-ARs present in early endosomes are in an activated conformation.

Because endosomes contain activated β₂-ARs, we next asked if receptors engage their cognate G protein from this compartment. Heterotrimeric G proteins and adenylyl cyclase can be observed in endosomes as well as at the plasma membrane, supporting the concept of endosome-based G protein signalling^4,6,10,24. To directly investigate the subcellular location of G protein activation, we developed a distinct biosensor based on another nanobody, Nb37, which specifically recognizes the guanine-nucleotide-free form of Gα_s representing the catalytic intermediate of G protein activation (Fig. 4a)²⁵. We hoped that because Nb37 binds a surface of the alpha-helical domain that is accessible only in the nucleotide-free form, we would be able to detect production of this critical but fleeting activation intermediate in living cells. This was indeed the case because Nb37–GFP localized in the cytoplasm of untreated cells and was rapidly recruited from the cytoplasm to the plasma membrane in response to isoprenaline application (Fig. 4b, yellow arrowhead, Pearson coefficient = 0.627). Membrane recruitment of Nb37–GFP was agonist-dependent even in cells overexpressing Gα_s and, importantly, the cytoplasmic concentration of Nb37–GFP achieved in our experiments (also approximately 20 nM) was substantially lower than that producing detectable inhibition of G protein activation in vitro (Supplementary Fig. 5). Together, these observations suggest that Nb37–GFP indeed functions as a specific biosensor of G_s activation under the experimental conditions used.

Nb37–GFP was recruited not only to the plasma membrane (Fig. 4b, yellow arrowhead) but also to β₂-AR-containing endosomes. Notably, endosome recruitment of Nb37–GFP occurred after the appearance of β₂-ARs in the endosome. Such later recruitment was evident in dual label confocal image series (Fig. 4b shows a representative example: white arrowhead indicates recently internalized β₂-AR, white arrow indicates Nb37–GFP recruitment, Pearson coefficient = 0.710; see also Supplementary Video 8). Nb37–GFP recruitment was specifically dependent on receptor activation because a mutant β₂-AR that is defective in G-protein coupling (β₂-AR-Cys 341 Gly)²⁶ did not produce detectable recruitment (Supplementary Fig. 6), whereas the distinct G_s-coupled DRD1 recruited Nb37–GFP to both the plasma membrane and endosomes (data not shown). Nb37–GFP localized uniformly on β₂-AR-containing endosomes in cells overexpressing Gα_s (Fig. 4b and Supplementary Video 8) but was recruited non-uniformly in cells expressing endogenous levels (Fig. 4c, arrowhead indicates an example). Moreover, Nb37–GFP recruitment to endosomes had a scintillating appearance, fluctuating rapidly when viewed in live image series (Supplementary Video 9), suggesting that G protein activation detected by Nb37–GFP occurs dynamically from limited regions of the endosome membrane.

We then asked if active β₂-AR G_s coupling from endosomes contributes to the cellular response, focusing on cAMP as a classical second messenger carrying the downstream signal, using a previously described real-time assay of cAMP accumulation in living cells and normalizing to the signal produced by receptor-independent activation of adenylyl cyclase with forskolin (5 µM)²⁷. Isoprenaline produced a rapid increase in cytoplasmic cAMP accumulation, reaching a maximum within approximately 10 min (Fig. 4d, blue line). Dyngo-4a, a dynamin inhibitor that blocks β₂-AR endocytosis by inhibiting CCP function^28,29, did not detectably affect the earliest portion of the forskolin-normalized cAMP accumulation curve but significantly reduced its later rise (Fig. 4d, green line, grey dots indicate P values), consistent with the two phases of endosome-localized activation detected by the nanobody-derived biosensors. A similar inhibition was observed in cells expressing a mutant β₂-AR with a C-terminal alanine residue added that prevents efficient recycling (β₂-AR-Ala)³⁰ (Supplementary Fig. 7), suggesting that the Dyngo-sensitive component of the cAMP response did not represent a secondary consequence of resensitization by receptor recycling¹⁷. Dyngo-4a did not produce the same effect on cAMP accumulation elicited by the internalization-defective β₂-AR-3S mutant receptor²¹ (Fig. 4e, f), further supporting the conclusion that the second signalling phase indeed requires receptor localization in endosomes.

The present findings revise a long-held tenet of molecular pharmacology, that acute signal transduction mediated by the canonical β₂-AR G_s signal transduction mechanism is plasma membrane delimited. The results provide direct evidence supporting the hypothesis that β₂-AR endocytosis contributes to a second phase of the acute cellular cAMP response, which represents a significant component of the overall biochemical signal developed within several minutes after the initial agonist application. Thus, while it remains clear that β₂-ARs can elicit G_s-mediated signal transduction from the plasma membrane, the present data reveal a discrete component of the acute signalling response that is initiated from endosomes (Fig. 4g). It remains unknown if β₂-ARs are continuously bound by agonist in endosomes, as depicted in the figure for simplicity, but conformational activation of G_s in endosomes is both receptor and agonist-dependent. Unambiguous detection of endosome-based activation of acute G-protein-linked signalling is presently limited to the β₂-AR G_s system for which the critical nanobodies are available. However, we believe that endosome-based contribution to the acute signalling response is likely widespread in the GPCR superfamily because the β₂-AR belongs to the largest group (family A) of GPCR and is often considered a prototype. We also suggest, more generally, that nanobody-based biosensors represent a versatile strategy for probing other types of dynamic conformational change with high spatiotemporal resolution in living cells.