Signal profiling of the β1AR reveals coupling to novel signalling pathways and distinct phenotypic responses mediated by β1AR and β2AR

A comprehensive understanding of signalling downstream of GPCRs requires a broad approach to capture novel signalling modalities in addition to established pathways. Here, using an array of sixteen validated BRET-based biosensors, we analyzed the ability of seven different β-adrenergic ligands to engage five distinct signalling pathways downstream of the β1-adrenergic receptor (β1AR). In addition to generating signalling signatures and capturing functional selectivity for the different ligands toward these pathways, we also revealed coupling to signalling pathways that have not previously been ascribed to the βAR. These include coupling to Gz and G12 pathways. The signalling cascade linking the β1AR to calcium mobilization was also characterized using a combination of BRET-based biosensors and CRISPR-engineered HEK 293 cells lacking the Gαs subunit or with pharmacological or genetically engineered pathway inhibitors. We show that both Gs and G12 are required for the full calcium response. Our work highlights the power of combining signal profiling with genome editing approaches to capture the full complement of GPCR signalling activities in a given cell type and to probe their underlying mechanisms.

G protein pathways engaged by the β 1 AR. In order to assess which G protein subtypes were activated by the β 1 AR, we used a panel of biosensors designed to capture the activation state of the heterotrimer by measuring the physical separation of Gα and Gβγ subunits via BRET 25 . Figure 1 shows BRET titration curves for 10 different Gα subunits, indicating that the βAR agonist isoproterenol leads to a robust activation of G s , G 12 and G z , as measured by a decrease in BRET. Smaller decreases in BRET in response to isoproterenol were also observed for other G proteins (G i2 , G oA and G oB ). No responses were observed for G q , G 13 , G i1 or G i3 . The lack of activation of different G protein isoforms did not result from insufficient biosensor sensitivity since robust responses were detected for the D2 dopamine receptor toward all Gα i subtypes ( Supplementary Fig. S2), for TPα receptors toward both G 12 and G 13 (Supplementary Fig. S3a,c) as well as for other GPCRs toward G q 17 , G 13 26 and G i 17,27 . To further characterize the activation of these G proteins and their downstream effector pathways by a panel of βAR ligands, full concentration-response curves and kinetic measurements were obtained using both the G protein sensors described above and biosensors detecting downstream events representative for each of these pathways. G s signalling. To probe the G s signalling pathway, an EPAC biosensor sensitive to cAMP levels 28,29 was used in conjunction with the G s activation biosensor (Fig. 2a,b). Similar activation kinetics were detected using both biosensors (Fig. 2c,d) and the effects of both full and partial agonists were revealed by the concentration-responses curves (Fig. 2e,f). A general pattern of concordance for both biosensors (Supplementary Tables S1 and S2) was observed with respect to efficacy and potency, although the differences between full and partial agonists were reduced when using the EPAC biosensor, consistent with the notion of signal amplification at the level of cAMP production. G i signalling. For each of the canonical PTX-sensitive Gi family members (G i1-3 , G oA and G oB , Fig. 3a), concentration-response curves were generated to further characterize the weak activation detected for some of these isoforms in BRET titration experiments. Reproducible kinetic responses and concentration/response curves could only be obtained for G i2 (Fig. 3b,c, Supplementary Tables S1, S2). Interestingly, only three agonists (isoproterenol, epinephrine and norepinephrine) could elicit responses from G i2 , suggesting more efficient coupling to G s . Of note, isoetharine and indacaterol, which were close to full agonist on G s activation and cAMP production could not evoke any G i activation. The fact that we only detected G i2 is not because this biosensor is more sensitive than those for other G i family members, as we could detect a more robust response for G i1 than G i2 for the D2-dopamine receptor which can activate all G i subfamily members ( Supplementary Fig. S2). Although not previously reported for the β 1 AR, previous studies have also demonstrated coupling of the β 2 AR to G i using conventional assays 30,31 and we could also detect β 2 AR coupling to G i2 using the BRET-based sensor (Fig. 3d,e).  1 AR HEK 293 stable cell lines were transfected with a constant amount of Gα-Rluc (BRET donor) and untagged Gβ 1 , along with increasing amounts of Gγ 1 -GFP construct (BRET acceptor). Cells were stimulated (red curves) or not (black curves) with 1 μM isoproterenol and BRET values collected. NetBRET values were calculated by subtracting the background BRET signal detected in cells expressing the Rluc-fused constructs alone (donor-Rluc) from the BRET values obtained in cells expressing the energy donor and acceptor (donor-Rluc and acceptor-GFP). BRET titration curves were generated for 10 different Gα subunits: Gα s (b), Gα q (c), Gα 12 (d), Gα 13 (e), Gα i1 (f), Gα i2 (g), Gα i3 (h), Gα z (i), Gα oA (j) and Gα oB (k). Values represent mean ± SEM of 3 independent experiments performed in triplicate. Responses for Gα s , Gα i2 , Gα z and Gα 12 were further analyzed in subsequent sections. www.nature.com/scientificreports www.nature.com/scientificreports/ G z signalling. G z is an atypical member of the G i family as it is insensitive to pertussis toxin. We next investigated if this G protein could be activated by the β 1 AR. Our BRET-based biosensor (Fig. 4a) detected robust time-and concentration-dependent activation of G z by the β 1 AR (Fig. 4b,d). The concentration-response curves in Fig. 4d show that, in addition to the three full agonists for G s , indacaterol also acted as a weak partial agonist for G z . The lack of detection of the effects of other partial agonists was not caused by differential dynamic windows in each assay as isoproterenol responses were similar in both the G s and G z sensors (compare Fig. 1b with Fig. 1i). As G z coupling has never previously been documented for any βAR subtype, we also examined the ability of the β 2 AR to engage G z . The kinetics of activation in response to isoproterenol were similar for both receptor subtypes (Fig. 4b,c). Further, as expected, the rank order of potency of the ligands was distinct for each receptor subtype, generally reflecting their relative affinities for the two receptors (Fig. 4d,e). However, one notable exception was that despite its characterization as a β 2 AR-selective ligand, indacaterol 32,33 only produced a response in the β 1 AR, indicating that selectivity also depends on the signalling pathway examined; a further manifestation of functional selectivity. A similar effect was observed for isoetharine, activating G s through both β 1 -and β 2 AR (albeit with lower potency for the β 1 AR), but only resulting in G z coupling for the β 2 AR (Fig. 4e). G 12 signalling. Another novel feature of β 1 AR signalling captured by our biosensor panel was the activation of G 12 (Fig. 5a,c,e,g) with no detectable activation of G 13 (compare Fig. 1d,e). A previous report 34 showed a similar ability of GPR35 to discriminate between G 12 and G 13 . The lack of activation of G 13 by the β 1 AR did not result from lower sensitivity of the biosensor as we could readily detect its activation by a known activator, the TPα receptor (TPαR, Supplementary Fig. S3a,c). To confirm and further characterize G 12 activation, an additional . Gα s activation and cAMP production induced by the β 1 AR. (a) Schematic representation of the Gα s -Rluc/Gγ 1 -GFP biosensor used to study the Gα s induced β 1 AR signalling. (b) Schematic representation of the cAMP biosensor used to study the cAMP increase induced by the β 1 AR Gα s activation. HEK 293 cells were transfected with (c,e) Gα s -Rluc, Gγ 1 -GFP and untagged Gβ 1 or with the (d,f) EPAC biosensor, along with the β 1 AR. Kinetic curves represent time course of (c) Gα s activation (vehicle and isoproterenol, n=3) or (d) cAMP accumulation (vehicle and isoproterenol, n=3) expressed as absolute BRET ratio. Concentration-responses curves were generated for (e) Gα s activation and (f) cAMP accumulation following β 1 AR activation by the indicated ligands. Data were normalized to maximal isoproterenol response, which was taken as 100%, and are expressed as mean ± SEM values. Detail of the number of experiments, maximal responses, pEC 50 values and statistical comparisons of curve parameters for different ligands are provided in Supplementary Tables S1 and S2. assay monitoring downstream engagement of p115-RhoGEF was designed (Fig. 5b) and again validated using TPαR ( Supplementary Fig. S3b,d). This biosensor monitors recruitment of the rgRGS domain of the G 12/13 effector, p115-RhoGEF, directly to Gα 12 by monitoring BRET between Gα 12 -Rluc and p115-RhoGEF-GFP10. The expression of the biosensor components did not affect cell surface receptor expression ( Supplementary  Fig. S4a,b). Rapid G 12 activation kinetics and recruitment of p115-RhoGEF were noted in response to isoproterenol (Fig. 5d,f). The kinetics of the two biosensors were quite different. It is difficult to make direct kinetic comparisons between biosensors based on different designs. In the case of Fig. 5c, the G12 activation biosensor is based on dissociation of Gα from Gβγ whereas the biosensor for RhoGEF is based on recruitment of a subdomain of p115-RhoGEF to the Gα (Fig. 5d). The dynamics of such interactions are most likely different and possibly explain the difference in the kinetics observed. The other compounds tested displayed similar potencies and efficacies for either of the two G 12 pathway biosensors, with the exception of the weak partial agonist isoetharine, where activity could only be detected for the G 12 /Gβγ sensor (compare Fig. 5e,f). Compounds had similar rank order of potencies and efficacies as those observed for G z .
As for G z coupling discussed above, engagement of G 12 had never been previously described for any βAR subtype. Therefore, the ability of the β 2 AR to activate this pathway was also examined. The β 2 AR was also capable www.nature.com/scientificreports www.nature.com/scientificreports/ of activating G 12 with potencies that were similar between the two receptors (Fig. 5g). To further characterize the downstream consequences of the engagement of Gα 12 by the β 1 AR, we next assessed activation of the Rho pathway, a known downstream effector of G 12 . Recruitment of PKN to the plasma membrane was measured using an ebBRET 17 biosensor based on PKN translocation to the plasma membrane upon Rho activation. This sensor monitors RlucII-tagged PKN (Rluc-PKN) density at the plasma membrane, which is labelled with a BRET acceptor, Renilla reniformis GFP (rGFP-CAAX, Fig. 6a). TPαR, a known activator of the G 12 /Rho/PKN pathway [35][36][37] , was used to validate this biosensor. As shown in Fig. 6c, agonist-mediated activation of TPαR led to an increased BRET signal, reflecting membrane recruitment of PKN. This response was blocked by the TPαR antagonist, SQ 29548. Consistent with the ability of this biosensor to detect Rho activity, a constitutively active mutant form of RhoA, Q63L-RhoA 38 , promoted robust BRET from the PKN-based sensor (Fig. 6d). To further validate the PKN recruitment assay as a faithful readout of G 12/13 activity, we assessed the effect of co-expressing either WT or constitutively active forms (CAM) of Gα 12 or Gα 13 . As seen in Fig. 6e, expression of WT Gα 12 and to a greater extent of CAM Gα 12 and CAM Gα 13 significantly increased recruitment of PKN to the plasma membrane. Finally, we designed a plasma membrane-targeted inhibitor of G 12/13 activity by fusing the rgRGS domain of P115-RhoGEF (p115-RGS) to the membrane anchoring CAAX domain (p115-RGS-CAAX) (Fig. 6b). As shown in Fig. 6f, expression of p115-RGS-CAAX completely blocked the ability of U46619 to promote PKN recruitment to the plasma membrane as assessed by ebBRET 39 . The G q inhibitor YM254890 did not affect the response, confirming that PKN recruitment resulted mainly from G 12/13 activation. The selectivity of p115-RGS-CAAX Figure 4. Gα z -induced activation by the β 1 AR and β 2 AR. (a) Schematic representation of the Gα z -Rluc/Gγ 1 -GFP biosensor used to study the Gα z induced βAR signalling. HEK 293 cells were transfected with Gα z -Rluc, Gγ 1 -GFP and untagged Gβ 1 , along with (b,d) β 1 AR or (c,e) β 2 AR. Kinetic curves represent time course of Gα z activation by (b) β 1 AR (vehicle n = 1; isoproterenol n = 2) or (c) β 2 AR (vehicle and isoproterenol, n = 2) expressed as absolute BRET ratios. Concentration-responses curves for Gα z activation following (d) β 1 AR or (e) β 2 AR activation by indicated ligands. Data were normalized to maximal isoproterenol response, which was take as 100%, and are expressed as mean ± SEM values. Details of the number of experiments, maximal responses, pEC 50 values and statistical comparisons of curve parameters for (d) β 1 AR activation by different ligands are provided in Supplementary Tables S1 and S2. For (e) β 2 AR activation, n = 3 for all ligands.
www.nature.com/scientificreports www.nature.com/scientificreports/ inhibitory action on G 12/13 was confirmed by its lack of effect on the activation of G q , G i2 or G s -mediated cAMP production stimulated by TPαR, D4R and β 1 AR, respectively (Supplementary Fig. S5b-d). Expression of the biosensor component did not influence the cell surface receptor expression ( Supplementary Fig. S4c). As shown in Fig. 6g,h, stimulation of the β 1 AR promoted PKN recruitment to the plasma membrane that was blocked by the expression of the G 12/13 activity dominant negative p115-RGS-CAAX, confirming engagement and activation of G 12 by the β1AR. (a) Schematic representation of the Gα 12 -Rluc/Gγ 1 -GFP biosensor used to study the Gα 12 induced βAR signalling. (b) Schematic representation of the Gα 12 -Rluc/ p115-RhoGef-GFP (p115-GFP) biosensor used to study the Gα 12 induced βAR signalling. HEK 293 cells were transfected with (c,e,g) Gα 12 -Rluc, Gγ 1 -GFP and untagged Gβ 1 or with (d,f) Gα 12 -Rluc, p115-GFP and untagged Gγ 1 and Gβ 1 , along with β 1 AR or (g) β 2 AR. Kinetic curves represent time course of (c) Gα 12 activation (vehicle and isoproterenol n = 3) or (d) Gα 12 -p115 biosensor activation (vehicle n = 2, isoproterenol n = 3), expressed as absolute BRET ratio. Concentration-responses curves for (e) Gα 12 activation or (f) Gα 12 -p115 biosensor activation following β 1 AR activation by indicated ligands. (g) Concentration-responses curves for Gα 12 activation following isoproterenol-induced β 1 AR or β 2 AR stimulation (n = 6). Data were normalized to maximal isoproterenol response (100%), and are expressed as mean ± SEM values. Detail (e,f) of the number of experiments, maximal responses, pEC 50 values and statistical comparisons of curve parameters for (e,f) β 1 AR activation by different ligands are provided in Supplementary Tables S1 and S2. www.nature.com/scientificreports www.nature.com/scientificreports/ Calcium signalling. Using a bioluminescent obelin calcium biosensor 40 , we observed that similar to β 2 AR activation 20 , β 1 AR stimulation leads to a calcium response 40 . As shown in Fig. 7, only three of the seven ligands tested (isoproterenol, epinephrine and norepinephrine) were able to stimulate a calcium response (Fig. 7c,d). Interestingly, these three compounds were also identified as full agonists using the G 12 biosensor. Ligands such as isoetharine and indacaterol that were almost full agonists for the G s pathway but only weakly activated G 12 were unable to promote significant calcium mobilization, suggesting that G 12 could play a role in the calcium response. To test this notion, we assessed βAR-mediated calcium mobilization in the presence or absence of the G 12/13 dominant negative p115-RGS-CAAX construct. Isoproterenol-stimulated calcium mobilization through either βAR isoform (Fig. 7e) was blunted in the presence of p115-RGS-CAAX, confirming the functional relevance of G 12 coupling. Again, both receptors were expressed at similar levels which was not altered by the presence of p115-CAAX ( Supplementary Fig. S4d). Given that the receptor calcium responses were only partially affected by the inhibition of G 12/13 signalling, we next examined potential contributions of other G proteins activated by these receptors. As shown in Fig. 7f, the isoproterenol-mediated response for either β 1 AR or β 2 AR was significantly reduced in CRISPR-Cas9-generated cells lacking Gα s 20 (ΔGs), again with no change in levels of receptor expression ( Supplementary Fig. S4e). Restoration of G s expression in the ΔGs cells rescued β 1 AR and β 2 AR-mediated calcium influx, confirming a role for G s in this response. Taken together, the data indicate that both G 12/13 and G s contribute to βAR-mediated calcium mobilization.
Global signalling profiles. Given the granularity of the signalling profiles obtained, we analyzed the potential for functional selectivity amongst the pathways engaged by the β 1 AR. The operational model 42 Table S3). The propensity of each compound to activate the pathways considered was illustrated using a radial graph format in which each of the vertices represented activity towards one of the biosensors tested. Both maximal response and the efficiency expressed as Log(τ/K A ) are shown. This analysis revealed that maximal response and efficiency profiles of isoproterenol, epinephrine or norepinephrine were practically identical for all the pathways analyzed. However, the patterns were quite different when considering the effects of isoetharine, salmeterol, indacaterol and xamoterol. These 4 compounds were partial agonists to different extents toward the G s /cAMP pathway and they displayed different abilities to promote detectable activation of G i2 , G z , G 12 and β-arrestin2. The most efficient of the four partial agonists toward G s /cAMP, indacaterol, was also able to recruit β-arrestin2, and activate Gα 12 , and Gα z . Isoetharine could activate G 12 and G s , and salmeterol and xamoterol could only activate G s . These different signalling profiles are illustrated in Supplementary Fig. S7 in which the relative efficiency is color-coded for each of the signalling pathways studied. Whether this reflects true functional bias or pure partial agonism towards pathways coupled to the activated receptor is difficult to unambiguously determine. When comparing the efficiency of each ligand towards the different pathways using the operational model (log(τ/K A )), the rank order of coupling efficiency of each compound (except for xamoterol) was respected for all pathways measured ( Supplementary Fig. S7). However, although indacaterol and xamoterol had similar efficiencies as epinephrine and norepinephrine to activate the G s /cAMP pathways, they were much weaker than these two compounds at activating G i2 , G z , G 12 or promoting the recruitment of β-arrestin2, in many cases not activating them at all. To a lesser extent, a similar comment can be made for isoetharine and salmeterol, although their efficiency to activate G s and cAMP production is slightly less than norepinephrine and epinephrine. In any case, the data clearly reveals distinct signalling profiles for the different ligands that translate into distinct cellular outputs.

Discussion
The β 1 AR was chosen here as prototypical GPCR where functional selectivity and biased signalling has been examined, albeit to a limited extent 7,8 . In previous studies, different ligands were reported to act in a biased fashion toward either ERK1/2 MAPK or adenylyl cyclase signalling and, in some cases antagonists or inverse agonists for the adenylyl cyclase pathway acted as agonists for the MAPK pathway. These studies and similar ones for other GPCRs 15,17,43 open the possibility that fine-tuning receptor-specific signalling outputs with functionally selective ligands could generate superior therapeutic options for a number of diseases by disfavouring signalling cascades associated with adverse effects without compromising beneficial pathways. To do this would first require a broader approach to the signalosome downstream of the receptors to capture the ensemble of pathways that can be engaged by any given receptor. However, in most cases, only a limited number of pathways have been n = 6, (e) n = 3). Statistical comparisons were done using two-way ANOVA followed by post-hoc comparison with Tukey's test. #### p < 0.0001 compared to (d) Q63L or (e) TPαR Vehicle Mock. www.nature.com/scientificreports www.nature.com/scientificreports/ examined. In addition, comprehensive signal profiling of proximal and distal outputs can reveal novel signalling pathways downstream of a given receptor and how these pathways might interact. The approach proposed in the present study was to combine BRET-based signal profiling with genome editing and the use of pharmacological and genetically engineered inhibitors to allow a more detailed dissection of the relevant signalling architecture in a model cell type, establishing proof of principle.
G protein profiling was first used to identify novel signalling partners for the β 1 AR. Our data shows for the first time that both the β 1 AR and β 2 AR are coupled to the G 12 signalling pathway. This was revealed not only by the activity of G 12 itself, but also using novel biosensors detecting the engagement of Rho-GEF and the recruitment of PKN downstream of G 12 activation, which was also found to contribute to βAR-stimulated calcium mobilization. Interestingly, our data suggest that neither receptor was coupled to G 13 , despite the fact that robust activation of G 13 could be detected in response to TPαR activation using our biosensors. Such selectivity between the two members of the G 12/13 family was also noted for GPR35 34 , which was found to be better coupled to G 13  Supplementary Tables S1 and S2. (e) HEK 293 cells were transiently transfected with β 1 AR or β 2 AR along with the obelin biosensor, with or without the p115-CAAX inhibitor. Data were normalized to maximal A23187 response (100%), determined from the area under the curve (AUC), and are expressed as the mean ± SEM values (n = 3). Statistical comparisons were done using two-way ANOVA followed by post-hoc comparison with Tukey's test. (f) Parental HEK 293 or ΔGαs cells were transfected with β 1 AR or β 2 AR, along with obelin, with or without Gα s . Data normalization and statistical analysis were done as described in (e) (n = 3). (2020) 10:8779 | https://doi.org/10.1038/s41598-020-65636-3 www.nature.com/scientificreports www.nature.com/scientificreports/ over G 12 . The implications for G 12 signalling downstream of βAR activation may include cell shape changes as well as cell migration, mediated through activation of the Rho pathway.
Interestingly, we noted that both G s and G 12 seem to be involved in βAR-mediated calcium mobilization, suggesting a role in calcium signalling, beyond the well-characterized activation of voltage-gated calcium channels by βAR in excitable tissues [44][45][46] . β 2 AR-mediated calcium mobilization in non-excitable HEK 293 cells was previously reported to require transactivation of the purinergic P2Y receptor through a G s -dependent but cAMP-independent mechanism 20 . Whether G 12 activation is also involved in this transactivation remains to be determined. For compounds such as isoproterenol that activated both G s and G 12 , we found that the two pathways contributed to rise in cytoplasmic calcium, as cells deleted for either G s or in which G 12/13 was inhibited by the G 12/13 dominant negative construct, p115-RGS-CAAX showed calcium responses significantly lower than those observed in the parental cells. Interestingly, compounds such as indacaterol, isoetharine and salmeterol that were strong partial agonists for G s but weak partial agonist (indacaterol, isoetharine) or unable to activate G 12 (salmeterol) could not generate significant calcium mobilization. This is consistent with an important role of G 12 in calcium mobilization. Dominant-negative RhoA or the ROCK inhibitor Y-27632 were previously shown to block GPR55-mediated Ca 2+ transients 47,48 , consistent with a role for the G 12/13 pathway in calcium responses. In a recent paper, the β 1 AR was proposed to couple to G 14 using DREADDs in combination with CRISPR gene deletions 39 . In theory, part of the calcium response could originate from this pathway as well. Although this cannot be excluded, this is unlikely since RNA-Seq analysis did not show detectable expression of G 14 in the cells we used (Supplementary Table S4).
Another novel observation of our study is the subtype-selective coupling of β 1 AR to G i family members. Although coupling to G i had already been shown for the β 2 AR 30,31 , our study is the first report showing that the β 1 AR can also couple to G i family members. When considering the pertussis toxin-sensitive G i isoforms, β 1 AR-mediated activation could only be detected for G i2 , indicating a possible selectivity among the G i protein family members. This apparent selectivity did not result from a different sensitivity of the BRET biosensors themselves, since dopamine D2R-mediated activation of G i1 was more robust than that of G i2 while those of G i3 , www.nature.com/scientificreports www.nature.com/scientificreports/ G oA and G oB were equivalent to G i2 . Regardless of whether this also reflects functional selectivity for the D2R G i/o subtypes engagement, our biosensors have the dynamic range to capture distinct levels of G protein activation by different GPCRs. Thus, as observed for the G 12/13 family, β 1 AR can display selectivity among members of a same G protein subfamily. This observation may provide interesting avenues to further explore functional differences among different members of the same Gα subfamily and start to understand the evolutionary pressures that maintained these closely related subtypes.
Our data also reveal for the first-time robust coupling of both the β 1 -and β 2 AR to G z , a member of the G i subfamily lacking an ADP ribosylation site, making them pertussis toxin-insensitive. This subtype has been shown to inhibit adenylyl cyclase types I, V and VI 49 and has been found to have a very low intrinsic GTPase activity compared to other G protein subtypes, including other G i isoforms. It is widely expressed in many tissues and is found in high levels in the central and peripheral nervous systems as well as in the platelets. In these circulating cells, G z knockout mice were found to display abnormal platelet aggregation at physiological concentrations of epinephrine 50 . The physiological implications of this G z coupling to the βARs remain to be investigated but the possibility of a counter regulatory action on the G s -stimulated production of cAMP in tissues expressing G z and the adenylyl cyclase sensitive to this G i subunit will be worth exploring. It should be noted that the potency of ligands to activate G z is an order of magnitude lower than that for G s activation, which would be consistent with a biphasic regulation of adenylyl cyclase, allowing for tight regulation and possibly oscillatory behaviours.
The present paper clearly illustrates that broader profiling of signalling pathways that can be engaged by a receptor can reveal signalling pathways that had not been anticipated before, even for receptors as well studied as the β 1 -and β 2 AR. We acknowledge that apparent differences in coupling efficacy between the β 1 -and β 2 AR could result from different expression levels, as they were not directly compared. The data also reveals that different ligands produce distinct signalling profiles leading to the activation to specific subsets of a given receptor's repertoire. Whether these different profiles result from bona-fide biased signalling or are consequences of different level of partial agonism and different coupling strengths of the different effectors engaged is difficult to unambiguously determine at this point, especially when no bias factor can be calculated because of the absence of response for some of the pathways. Nevertheless, the selective activation of weakly coupled pathways seems to correlate with the overall relative efficacy of the compounds tested, indicating that partial agonism plays an important role in the apparent functional selectivity. However, whether the different signalling profiles result from biased signalling or partial agonism, the end result is that some ligands activate certain pathways but not others and that this is bound to have functional consequences. Many of our biosensors are based on overexpression of G protein subunits which may alter native receptor/G protein stoichiometries. An additional caveat must be considered when using gene deleted lines where rewiring of signalling pathways might occur. Resolving such issues will require a closer focus using distinct approaches.
What impact might the novel signalling pathways uncovered in the present study and the ligand-selective signalling profiles observed have on clinical indications where β-adrenergic agonists or antagonists are used will be an interesting area for further study. For this purpose, it will be critical to move biosensors into more physiologically relevant cell systems as the HEK 293 cell may not reflect important differences in signalling pathways in disease-relevant cells. Our results showed a limited functional bias for the various agonists tested among the different pathways tested downstream of the β 1 AR, but was able to robustly discriminate between partial and full agonists. Thus, as a homogenous platform, our approach allows a more global appreciation of the signalling profile of a given cell.

Material and Methods
Cell culture and transfection. HEK 293 and ΔGα cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 100 U/ml penicillin/streptomycin at 37 °C in 5% CO 2 . The HA-β 1 AR HEK 293 stable cell line was maintained in complete medium supplemented with 1 μg/ml of puromycin. Except for G s titration and concentration-response curves with β 1 AR, all transfections were carried out as follows: 48 h before the experiments, cells were transfected in suspension using polyethylenimine at 3:1 PEI/DNA ratio and seeded (~3×10 4 cells/well) in 96-well/plates pre-treated with poly-D-lysine. Each of the expression vectors and biosensor constructs were diluted in PBS, and the total quantity of DNA was completed to 1 µg/row with salmon sperm DNA. For G s titration and concentration-response curves, stable HA-β 1 AR cells were plated at 9000 cells/well in a poly-ornithine treated white 96-well plate the day before transfection and were transfected using PEI (2.5ug PEI:1ug DNA ratio) at 150 ng total DNA (i.e. biosensor constructs, HA-β 1 AR vector and empty pcDNA3.1 vector) per well. 24 h post-transfection, the medium was changed. The HA-β 1 AR HEK 293 stable cell line was used for all β 1 AR experiments, except for G q and G 13 saturation curves and when experiments were done in parental and ΔGα (ΔGα s ) cell lines. β 1 AR-encoding plasmid was also added in the transfection mix for all experiments, except for EPAC and obelin biosensors, where native responses were measured.
p115RhoGEF biosensor to monitor Gα 12 activity: A BRET-based biosensor composed of RGS-homology (RH) domain (amino acids 1-246) of p115RhoGEF fused to GFP10 and G 12 -84RlucII was used to measure Gα 12 activity. Cells were transfected with 40 ng of G 12 -84RlucII, 500 ng of p115RhoGEF-GFP10 and 300 ng of receptor per row of a 96-well plate. For concentration response experiments, BRET was monitored 2 minutes after agonist addition.
PKN biosensors to monitor Rho activation: Upon receptor stimulation, activated G proteins promote Rho activation through RhoGEF recruitment to the plasma membrane. The activated Rho then recruits the RlucII-tagged effector (PKN) to the PM. Rho activity is monitored using the BRET between RlucII and a membrane-bound rGFP. The Rho binding domain of PKN1 was tagged with RlucII (BRET donor) to monitor its recruitment to the plasma membrane using enhanced bystander BRET with rGFP-CAAX (kRAS) 23  p115-CAAX-based inhibitor of G protein-mediated Rho activation. p115-CAAX is composed of the G protein binding domains (rgRGS) of p115RhoGEF, targeted to the plasma membrane using the prenylated polybasic sequence from kRAS. Upon receptor stimulation, p115-CAAX is recruited to activated G 12 or G 13 , preventing recruitment of WT RhoGEF and thus activation of Rho. The rgRGS domain is known to promote GTPase activity, further inhibiting G protein-mediated signalling. To validate p115-CAAX as an inhibitor of G 12 /G 13 pathway, HEK 293 cells were transfected with HA-TPαR (10 ng), PKN-RlucII (1 ng) and rGFP-CAAX (300 ng) along with p115-CAAX or ssDNA (mock). Cells were pre-treated with YM254890 or vehicle for 30 min at 37 °C and stimulated with TPαR agonist U46619 (100 nM) or vehicle for 6 min. For calcium mobilization experiments, HEK 293 cells were transfected with 500 ng of Obelin biosensor, 10 ng of HA-β 1 AR or HA-β 2 AR and 20 ng of p115-CAAX. Kinetic data for isoproterenol (1 µM) and A23187 (5 µM) were collected for 30 sec at 37 °C, with an integration time of 0.5 sec. To assess the effect of p115-CAAX on G q /G i signalling, cells were transfected with 10 ng of receptor, 5 ng of Gα q /Gα i2 -RlucII, 100 ng of Gβ 1 WT, 200 ng of GFP10-Gγ 1 and 20 ng of p115-CAAX. BRET values were collected after 6 min of agonist stimulation (TPαR: U46619, D4R: dopamine) at 37 °C and addition of Prolume purple. The influence of p115-CAAX on cAMP response from endogenous β 2 AR or overexpressed β 1 AR (10 ng) was assessed using EPAC biosensor (25 ng). BRET values were collected after 10 min of Isoproterenol stimulation and addition of coelenterazine 400a, as described in calcium transfection (see below).
BRET experiments. BRET was in general performed as previously decribed 29,51,52 . 48 h after transfection, cells were washed once with PBS and incubated for 1 h at 37 °C in Tyrode's-HEPES buffer (137 mM NaCl, 0.9 mM KCl, 1 mM MgCl 2 , 11.9 mM NaHCO 3 , 5.5 mM glucose, 3.6 mM NaH 2 PO 4 , 25 mM HEPES, and 1 mM CaCl 2 , pH 7.4). The expression levels of the energy acceptor GFP10-tagged proteins were measured as total fluorescence using a FlexStation II microplate reader (Molecular Devices, Sunnyvale, CA, USA) with excitation at 400 nm and emission at 510 nm. Before β 1 AR activation, cells were exposed to ICI 118,551 (10 nM) for 30 min prior to the experiment in order to inhibit β 2 AR activity. For concentration-response experiments, cells were treated, with or without ligands, for the indicated times, and BRET was measured using a TriStar2 LB 942 Multidetection Microplate Reader (Berthold Technologies), equipped with a BRET400-GFP2/10 filter set (acceptor, 515 ± 20 nm; and donor, 400 ± 70 nm filters), 5 min after the addition of 2.5 µM of coelenterazine 400a. Absolute BRET signals (BRET) were derived from emissions detected with the energy acceptor filter divided by emission detected using the energy donor filter, while netBRET signals were obtained by subtracting BRET signals obtained in cells expressing the Rluc-fused donor constructs alone. ΔBRET was calculated by subtracting ligand-induced BRET from vehicle BRET. For kinetic experiments, cells were preincubated with coelenterazine 400a for 5 min, followed by readings for the indicated times, following addition of drugs. For G s and G 12 titration and concentration-response curves, 48 h post-transfection, wells were washed once with Kreb's/HEPES buffer (146 mM NaCl, 4.2 mM KCl, 0.5 mM MgCl 2 , 1 mM CaCl 2 , 5.9 mM Glucose and 10 mM HEPES pH 7.4) and 80 μl of Kreb's/HEPES buffer was added and plates left for 2-3 h at 37 °C. Then, 10 nM ICI 118,551 was added and plates left another 30 min at 37 °C to block endogenous β 2 AR. Finally, agonists were added and BRET assessed similarly as described above, using a FLUOstar Optima (BMG) equipped with BRET2 filter set (410 nm/515 nm). Saturation assays were performed initially to determine optimal donor to acceptor ratios for kinetic experiments (2020) 10:8779 | https://doi.org/10.1038/s41598-020-65636-3 www.nature.com/scientificreports www.nature.com/scientificreports/ and concentration-response curves and read 1 to 3 minutes after agonist addition (concentration 1 uM isoproterenol). Kinetic measurements were performed 48 h post-transfection, 5 minutes after addition of coelenterazine 400a (2.5 µM) or Prolume purple (1 µM, 6 min).
Calcium mobilization. An obelin biosensor was used as a calcium reporter as described previously 6,54,55 .
Stable HA-β 1 AR HEK 293 cells were transfected in suspension with 500 ng of WT obelin-pLVXi2H/per row. For experiments performed in parental and cells lacking Gα s proteins (ΔG s ), cells were transiently transfected with 100 ng of receptor and WT obelin. 48 h after transfection, cells were washed once with Tyrode's-HEPES buffer and incubated with the obelin substrate, coelenterazine cp (1 µM; Biotium) for ~2 h in the dark. For concentration-response experiments, increasing concentrations of agonists, diluted in Tyrode's buffer, were injected into the wells and luminescence was measured using a SpectraMax L (Molecular Devices). Kinetics of activation were determined for each ligand concentration for 60 s and concentration-response curves were determined from the area under the curve (AUC). For Gα S complementation experiments, HEK 293 T parental or cells lacking Gα S were transfected with 10 ng of β 1 or β 2 AR, 500 ng of obelin, in the presence or absence of 10 ng of Gα S . Luminescence was measured using a Mithras LB940 microplate reader.
Cell surface ELISA. Parental HEK 293 cells or cells lacking Gα s were transfected with HA-β 1 AR, HA-β 2 AR or HA-TPαR. 48 h after transfection cells were washed with PBS and fixed with 3% paraformaldehyde for 10 min at RT. Cells were labelled with mouse anti-HA-HRP antibody (3F10; 1:2000) for 1 h at RT in Wash B buffer (PBS supplemented with 0.5% BSA), followed by extensive washing (3×10 min). Vybrant DyeCycle Orange stain was added for 30 min at a final concentration of 10 μM at RT. Vybrant fluorescence was measured with excitation at 519 nm and emission at 563 nm using a FlexStationII microplate reader (Molecular Devices, Sunnyvale, CA, USA) to control for the number of cells/well. Total luminescence was measured 2 min after the addition of the HRP substrate Western lightning plus ECL using a Mithras LB940 microplate reader. Receptor cell surface expression was calculated as ratio of total luminescence to vibrant fluorescence.

Data analysis. Concentration response curves describing ligand responses by different ligands were analyzed
with Graphpad Prismv6 (GraphPad Software, La Jolla, CA), using built-in 3 or 4 parameter logistic equations to obtain independent pEC50 and maximal response values for different receptor-biosensor pairs y = a + (b-a)/ (1 + 10 (logEC50-x) *c) where: y is the measured response; a is the minimal asymptote, b is the maximal asymptote; b-a is maximal response and c is the slope.
Data were also fit with the operational model of Black and Leff 42,56 using a set of equations kindly provided by Dr. Arthur Christopoulos. These equations were introduced into Graphpad Prism6: A one-phase exponential decay was used to calculate the half-life in kinetic experiments. Statistical significance of ligand-induced changes at the different biosensors was established using one-way ANOVA to reveal concentration effects. pEC 50 and maximal responses of different ligands were compared to corresponding isoproterenol values by means of one-way ANOVA followed by post-hoc Dunnett test. Specific details on statistical comparisons are provided in legends of Figures and Tables.