Conformational biosensors reveal allosteric interactions between heterodimeric AT1 angiotensin and prostaglandin F2α receptors

G protein-coupled receptors (GPCRs) are conformationally dynamic proteins transmitting ligand-encoded signals in multiple ways. This transmission is highly complex and achieved through induction of distinct GPCR conformations, which preferentially drive specific receptor-mediated signaling events. This conformational capacity can be further enlarged via allosteric effects between dimers, warranting further study of these effects. Using GPCR conformation-sensitive biosensors, we investigated allosterically induced conformational changes in the recently reported F prostanoid (FP)/angiotensin II type 1 receptor (AT1R) heterodimer. Ligand occupancy of the AT1R induced distinct conformational changes in FP compared with those driven by PGF2α in bioluminescence resonance energy transfer (BRET)-based FP biosensors engineered with Renilla luciferase (RLuc) as an energy donor in the C-tail and fluorescein arsenical hairpin binder (FlAsH)-labeled acceptors at different positions in the intracellular loops. We also found that this allosteric communication is mediated through Gαq and may also involve proximal (phospholipase C) but not distal (protein kinase C) signaling partners. Interestingly, β-arrestin-biased AT1R agonists could also transmit a Gαq-dependent signal to FP without activation of downstream Gαq signaling. This transmission of information was specific to the AT1R/FP complex, as activation of Gαq by the oxytocin receptor did not recapitulate the same phenomenon. Finally, information flow was asymmetric in the sense that FP activation had negligible effects on AT1R-based conformational biosensors. The identification of partner-induced GPCR conformations may help identify novel allosteric effects when investigating multiprotein receptor signaling complexes.

G protein-coupled receptors (GPCRs) 3 are conformationally dynamic proteins that transmit information following interaction with different ligands that can promote multiple, yet specific downstream outputs depending on cellular context. It is widely accepted that this complexity is achieved through induction of distinct conformations in the receptor, whereby distinct conformational states preferentially drive specific receptormediated signaling events. The available conformational capacity, and therefore possible receptor functions, can be further expanded when considering allosteric effects. Receptor signaling partners, such as heterodimer partners, may provide novel conformational space generating new signaling modalities or protein life-cycle behaviors (1,2).
With the recent explosion in the number of GPCR structures, our understanding of the mechanics underlying GPCR function has evolved considerably. Structures of GPCRs bound to various ligands, nanobodies, and co-crystals with G proteins demonstrate the conformation complexity of individual receptors (3)(4)(5)(6)(7). Although X-ray crystallography provides high-resolution snapshots of the receptor, it is limited with respect to reporting on dynamic events and providing information on highly mobile protein domains or larger protein complexes. This is especially true when considering allosterically-mediated conformational changes in the context of receptor heterodimers. Although there have been some reports of receptor oligomerization observed in crystal structures, this has not been a consistent observation and it is difficult to determine whether such arrangements represent artifacts of crystal packing, or reflect physiologically relevant interactions that occur in live cells (8 -12). Some might argue that the paucity of GPCR oligomers in crystal structures refutes their existence, but many other biochemical and biophysical techniques reveal structural or functional profiles consistent with receptor oligomerization (13)(14)(15)(16)(17)(18). At any rate, GPCR oligomerization must be interro-gated from new perspectives that overcome some of these limitations.
Resonance energy transfer (RET) has been a key technology in characterizing GPCR oligomerization. High spatial and temporal resolution and applicability in live cell systems allows for highly robust assay development. RET experiments have been used extensively to support receptor dimerization (19 -24) but stringent controls are required to interpret the data (25)(26)(27)(28)(29)(30). Newer fluorescence-based RET (FRET) assays have been developed that can capture intra-molecular rearrangements in GPCRs in response to agonist. These sensors make use of a small fluorescent molecule, fluorescein biarsenical hairpin binder (FlAsH), as the acceptor and report on ligand bindingassociated conformational rearrangements in multiple GPCRs (31)(32)(33)(34)(35). Such sensors have also been used to examine conformational dynamics of GPCRs in oligomeric states, specifically a class C homodimer of mGluR1 and a class A heterodimer of ␣ 2A -adrenergic and -opioid receptors (36,37). Although valuable insight has been gained through the study of class C GPCRs, they are well accepted as obligate dimers. However, the latter article (37) sheds some light on conformational cross-talk in a putative class A receptor heterodimer. These authors demonstrated that morphine, targeting the -opioid receptor, affected the conformation of the ␣ 2A -adrenergic receptor in the presence of its ligand norepinephrine. This effect was shown to be G protein-independent and was still detected in isolated membranes, suggesting a simple mechanism of dimerization mediated through direct GPCR/GPCR contact.
We previously reported a functional interaction between FP and AT1R in vascular smooth muscle cells (38). Furthermore, a physical interaction between the receptors was reported in HEK 293 cells heterologously expressing both receptors as well as with endogenous receptors in vascular smooth muscle cells. It was also shown that signaling pathways modulated by the putative heterodimer could be either symmetrically or asymmetrically regulated (i.e. respond or not to stimulation of either partner in a similar or dissimilar way) depending on the signaling output being measured. This differential regulation of outputs lead to a functionally relevant bias attributed to AT1R/FP heterodimerization and may constitute a new druggable molecular target. To further understand the interplay between these two receptors, we generated a panel of FlAsH/BRET-based conformation-sensitive biosensors (34, 35, 39 -41) to explore the effect each of these receptors had on the conformational landscape of its dimer partner. We observed an asymmetrical transmission of conformational information from AT1R to FP whereby angiotensin II (Ang II) stimulation lead to a rearrangement in FP when tagged with biosensors engineered into the third intracellular loop with respect to its C-terminal RLuc moiety. This Ang II-induced conformational change was distinct from that driven by PGF2␣. The effect of Ang II on the conformation of FP was dependent on the presence of active G␣ q or G␣ 11 and appeared to be mostly independent of downstream signaling and conformational information flow between the two receptors seemed to be unidirectional or asymmetric. Finally, this transmission from AT1R to FP appeared to be specific as another G␣ q -coupled receptor was unable to induce similar conformational rearrangements in FP.

AT1R ligand binding induces a conformational change in FP
To investigate the interplay between the protomers in the FP/AT1R heterodimer, we began by co-expressing each of our previously published FP-ICL3-RLucII conformational biosensors (FlAsH "walked" through the third intracellular loop (ICL3) at 5 different positions (39)) with wild-type AT1R (AT1R-WT) in HEK 293 cells. In this configuration, the assay strictly reports on conformational rearrangements between different vantage points in ICL3 and the C terminus of FP induced in response to ligand stimulation, i.e. only changes in FP conformation are reported. We noted a similar pattern across the different conformational biosensors in response to PGF2␣ as previously observed when expressed alone (Fig. 1A). When comparing the position of the FlAsH tags, the ICL3 P4 biosensor was again the FlAsH tag position that showed the largest ⌬BRET in response to PGF2␣ in the panel (39). This response was dose-dependent (supplemental Fig. S1). Interestingly, when the cells were stimulated with Ang II, we not only observed a change in the BRET across all the FP conformational biosensors in the panel but opposite in direction, suggesting that activation of the heterodimer partner caused a distinct conformational change in FP as compared with its cognate agonist (Fig. 1B). As with direct stimulation of the conformational biosensor with PGF2␣, the FP ICL3 P4 sensor also reported the largest ⌬BRET in response to Ang II and the response was dosedependent (supplemental Fig. 1B). We therefore selected this biosensor as a focus for further study.
To assess the specificity of these effects, we next pretreated cells with antagonists for either FP or AT1R (Fig. 1C). HEK 293 cells transfected with the FP-ICL3 P4-RLucII conformation sensor and AT1R-WT were pretreated with 10 M AS604872, a FP antagonist, or 10 M losartan, an AT1R antagonist, followed by stimulation with 1 M PGF2␣ or 1 M Ang II. AS604872 was able to block the response to PGF2␣ but interestingly had no effect on the response driven by Ang II, suggesting the Ang II-induced conformational rearrangements were uncoupled from the orthosteric binding pocket of FP. Pretreatment with losartan had no effect on the PGF2␣-induced conformational rearrangement in FP but was able to block the effect of Ang II. To ensure the dependence of AT1R to drive conformational change in FP across the heterodimer, HEK 293 cells were transfected with only the FP-ICL3 P4-RLucII conformation-sensor and in this case, were insensitive to either Ang II or losartan (Fig. 1D).
Although we represent the data as the ligand-induced change in BRET by subtracting the average reading before stimulation from the average reading post-ligand addition for simplicity of analysis, all our data sets were also temporally resolved. We noted a rapid response of the FP-ICL3 P4-RLucII sensor in response to 1 M PGF2␣ when expressed alone (Fig. 1E). This response reached a plateau and was sustained for the length of the recording, suggesting that such conformational changes are stable as long as agonist is present. When co-expressed with AT1R-WT, the response to 1 M PGF2␣ was similar to FP alone, whereas the response to 1 M Ang II had a slower time constant before reaching a plateau (Fig. 1F). Additionally, co-expression of AT1R lead to an increase in the ligand-naive BRET, suggesting expression of AT1R, independent of ligand binding, modulates the conformation of FP (compare Fig. 1, E versus F). However, how this conformational information was transmitted was not clear from these experiments.

AT1R-induced conformational change in FP is dependent on G␣ q expression and activity
We next explored how the information was transmitted from AT1R to FP. Our initial focus was one step removed from the receptor, in the G proteins presumably shared by the heterodimeric receptor. Both FP and AT1R have been demonstrated to couple to G␣ q , G␣ 12 , and G␣ i (42)(43)(44)(45)(46). To investigate involve-ment of different G␣ subunits, we used small molecule inhibitors, microbial toxins, and cell lines with CRISPR-mediated knock-out of G proteins and then re-assessed the ability of PGF2␣ or Ang II to induce a conformational change in FP.
First, using the G␣ q inhibitor FR900359 (47) on HEK 293 cells co-expressing the FP-ICL3 P4-RLucII conformational biosensor and AT1R-WT, we noted a slight effect on the PGF2␣induced response where the magnitude of the response was slightly larger when G␣ q activity was inhibited ( Fig. 2A). In contrast, the Ang II response was completely abrogated upon G␣ q inhibition. We next used a HEK 293 cell line made devoid of functional G␣ q , G␣ 11 , G␣ 12 , and G␣ 13 using CRISPR/Cas9 (⌬G␣ q/11/12/13 line, Fig. 2B) (35). This cell line was especially valuable, as we could perform rescue experiments reintroducing individual G␣ subunits. Similar to our observations made using FR900359, there was minimal effect of the loss of these G proteins on the FP-ICL3 P4-RLucII response to PGF2␣. As with the small molecule inhibition of G q/11 , however, the ⌬G␣ q/11/12/13 line also showed the dependence of Ang II-mediated conformational alterations in FP on re-expression of G␣ q or G␣ 11 , highlighting the necessity of both expression and activity of G␣ q/11 for Ang II to affect FP conformation. How-ever, co-expression of either G␣ 12 or G␣ 13 did not re-establish cross-talk between FP and the AT1R in the ⌬G␣ q/11/12/13 line. We also examined the possible involvement of G␣ i in the transmission of conformational information between the AT1R and FP using pertussis toxin (PTX, Fig. 2C). Pre-treatment of the ⌬G␣ q/11/12/13 line expressing FP-ICL3 P4-RLucII, AT1R-WT, and various co-expressed G proteins with PTX had no effect on where ** ϭ p Ͻ 0.01. C, a two-way analysis of variance was used to analyze the 2 graphs. For both graphs, the G protein factor and interaction were nonsignificant, whereas the PTX factor was significant (PGF2␣ ϭ p Ͻ 0.01, Ang II ϭ p Ͻ 0.001). Bonferroni corrected tests were used to make post hoc comparisons where: * ϭ p Ͻ 0.05; ** ϭ p Ͻ 0.01. D and E, kinetic traces of conformational biosensors responding to ligand in the presence and absence of G proteins. ⌬G␣ q/11/12/13 cells were transfected with FP-ICL3 P4-RLucII and AT1R-WT (D) or FP-ICL3 P4-RLucII, AT1R-WT, and G␣ q (E). The specified ligand was injected onto the cells at the 50th repeat measure as denoted by an arrow. Inset to D: offset basal traces for PGF2␣ and Ang II to better demonstrate differences in biosensor responses. Traces represent the mean of 3 independent experiments.
pressed suggesting a more complicated cross-talk we did not explore further here. When examining the kinetic data, we noted that PGF2␣ induced a sustained response regardless of the expression of G␣ q , whereas Ang II only induced a sustained response when G␣ q was present (Fig. 2, D and E). When compared with native HEK 293 cells, an increase in the time for the PGF2␣ response to reach a plateau was observed (compare Fig. 1F versus 2D). The effect was more clearly seen in the inset to Fig. 2D, where we have offset the baselines to match. This was also noted when G␣ q was re-introduced for both PGF2␣ and Ang II and therefore may be a property of the cell line as opposed to simple G protein expression per se (compare Fig. 1F versus 2E). We also observed an increase in the ligand-naive BRET when G␣ q was re-introduced compared with native HEK 293 cells (compare Fig. 1F versus 2E). This may again be due differences in gene expression profiles between the cells or the differences in G protein biosynthesis from endogenous and exogenous genetic templates. Taken together, our data showed that transmission of information from AT1R leading to a conformational change in FP was dependent on both the expression and activity of G␣ q/11 and independent of the other G␣ proteins tested.

Ang II-induced conformational change in FP required an intact cell membrane
Knowing that specific G proteins were required to mediate transmission of conformational information between the two protomers, we next focused on the involvement of other signaling proteins downstream of the receptor. To dissociate the receptor from more distal cytosolic components associated with molecular cross-talk, we first prepared membranes from the ⌬G␣ q/11/12/13 line expressing the FP-ICL3 P4-RLucII conformation sensor with and without AT1R-WT and G␣ q . We initially examined the effect of the membrane preparation on basal BRET as reported from the FP conformational biosensor in the absence of ligand (Fig. 3A). We observed an increase in BRET in membranes that were reduced upon co-expression of G␣ q . Western blot analysis on the membrane preparation samples confirmed co-expression of G␣ q (Fig. 3B). We then compared intact cell and membrane preparation samples following stimulation with PGF2␣ ( Fig. 3C) or Ang II (Fig. 3D). In response to PGF2␣, we observed larger ⌬BRET values in the membrane preparation samples compared with intact cells, but the effects were similar in direction. However, following stimulation with Ang II we noted a distinct ⌬BRET (of opposite direction) when comparing intact cells with membranes regardless of whether or not AT1R or G␣ q was co-expressed.
As membrane preparation disrupted the ability of Ang II to induce ⌬BRET and because G␣ q/11 was required to mediate transmission of information from AT1R to FP in intact cells, we assessed whether this was correlated with disruption of a physical interaction between the receptors per se. We demonstrated that neither membrane preparation itself, nor expression or absence of G␣ q , affected immunoprecipitation of FP with AT1R (Fig. 3E). These observations suggest that cell integrity (and probably proper stiochiometric association of multiple signaling partners) is necessary to manifest allosteric interac-tions in the dimer even though a physical interaction between the two protomers was still maintained.

Proximal but not distal signaling partners affect Ang IIinduced conformational rearrangement in FP
Having identified a requirement for both G␣ q expression and activity as well as an intact cell membrane to allosterically connect FP and AT1R, we wanted to more explicitly examine whether conformational cross-talk from AT1R to FP was mediated through canonical molecular cross-talk. Because the membrane preparation procedure altered communication of conformational information from AT1R to FP, we investigated involvement of downstream signaling partners. We began by examining the distal effector protein kinase C (PKC) in the G␣ q signaling pathway shared by both receptors. Using either an inhibitor (Gö6983, Fig. 4A) or an activator (phorbol 12-myristate 13-acetate, Fig. 4B), we detected no effect on the ability of either PGF2␣ or Ang II to induce conformational changes in FP-ICL3 P4-RLucII when co-expressed with AT1R-WT in HEK 293 cells. We next examined a more proximal signaling protein in the G␣ q signaling cascade, phospholipase C (PLC). When HEK 293 cells expressing the FP-ICL3 P4-RLucII conformation sensor and AT1R-WT were pre-treated with the small molecule PLC inhibitor U73122, we noted no effect on the PGF2␣-mediated ⌬BRET and a small but significant reduction in the response to Ang II suggesting that the receptor/ G␣ q/11 /PLC complex may be shared by the heterodimer (Fig. 4C).

␤-Arrestin-biased ligands of AT1R are also capable of inducing a conformational change in FP
There is considerable interest in developing ␤-arrestin or G protein-biased ligands as therapeutics. However, given that receptors are constitutively associated with G proteins, there is still a lot we do not understand about how G proteins might be involved in how biased ligands signal. Thus, we next used AT1R-biased ligands that selectively activate ␤-arrestin over G␣ q (48). Using all 5 of the FP-ICL3-RLucII conformational biosensors, we screened this group of ligands to see if they could induce conformational changes in FP (Fig. 5A). Our initial screen detected the ability of both SBpA and SI, ␤-arrestinbiased AT1R agonists, to induce a small positive ⌬BRET in the FP conformational biosensors at saturating concentrations. As with Ang II, the FP-ICL3 P4-RLucII conformational biosensor showed the largest magnitude ⌬BRET across the panel of ligands and was therefore the focus of the subsequent experiment. To ensure that the heterodimerization between FP and AT1R did not alter the defined functional profile of these ligands as biased, we assessed their ability to activate G␣ q using a BRET-based G␣ q activation biosensor (Fig. 5A, inset). Neither SBpA nor SI were able to elicit a response in this biosensor, whereas a robust response was detected in response to Ang II. We tested SBpA and SI in the ⌬G␣ q/11/12/13 line to assess whether there was any requirement for G␣ q in mediating conformational rearrangement in FP. Interestingly, both SBpAand SI-induced conformational changes were ablated in the ⌬G␣ q/11/12/13 line, but could again be rescued when G␣ q was reintroduced through transient co-expression (Fig. 5, B and C, comparing G␣ q versus pcDNA3.1 vector control, respectively). These observations suggest that even biased ligands require G␣ q to transmit information to FP even though they do not necessarily result in downstream G␣ q signaling.

The G␣ q -coupled OTR does not induce a similar conformational change in FP
Although one distal effector pathway was not involved in the conformational cross-talk between FP and AT1R, we did not systematically interrogate other pathways. However, to affirm receptor dimerization as a mechanism underlying the response of FP conformational biosensors to Ang II, we explored whether another G␣ q -coupled GPCR could lead to a similar response in FP (49,50) where similar patterns of molecular cross-talk might also be expected. We compared HEK 293 cells transfected with the FP-ICL3 P4-RLucII conformational biosensor and either AT1R-WT or OTR-WT (Fig. 6, A and B). We noted FP conformational responses to Ang II and not OT, although the cells were transfected with AT1R-WT or OTR-WT. To confirm that the absence of response to OT was not due to relative expression levels of the receptors, we titrated the expression of either AT1R-WT (Fig. 6C) or OTR-WT (Fig. 6D) and noted a saturating response to AT1R-WT/Ang II conditions but no response across any of the OTR-WT/OT conditions. Finally, we confirmed that both AT1R-WT (Fig. 6E) and OT-WT (Fig. 6F) constructs were able to activate a BRET-α -Gα q α -Na + /K + ATPase based G␣ q biosensor. This demonstrated that G␣ q activation per se was not driving the conformational change in FP-ICL3 P4-RLucII, suggesting a shared G protein in the context of the AT1R/FP heterodimer.

Asymmetrical transmission of conformational information between protomers of the FP/AT1R heterodimer
Finally, we considered whether ligand binding to the FP protomer could induce a similar conformational change in its AT1R partner. Using a corresponding panel of conformational biosensors build into AT1R (35), we screened whether ligand binding to WT-FP induced changes in the relative orientation between the ICL3 and C terminus of AT1R. Examining multiple vantage points in the ICL3 of AT1R demonstrated robust, position-dependent responses to orthosteric ligand (Ang II) in our AT1R conformational biosensors (Fig. 7A). Intriguingly, stimulation with PGF2␣ was not able to induce a response in any of the AT1R conformational biosensors suggesting that arrangement of the putative heterodimer was asymmetric (Fig.  7B). Examination of the kinetic data demonstrated a sustained response of the conformational biosensor to Ang II, similar to what we had observed with the FP sensor (Fig. 7, C and D). The effect of co-expressing FP-WT was not as large on ligand-naive BRET compared with AT1R-WT with the FP biosensor (compare Fig. 7, C versus D, and 1, E versus F). As before, we noted no change in BRET across the entire recording period when in response to PGF2␣ (Fig. 7D). When comparing traces from the ⌬G␣ q/11/12/13 with and without re-introducing G␣ q , we noted an increase in the sustained response to Ang II when G␣ q was present (Fig. 7, E versus F). This is similar to observations we made when examining the sensor expressed alone without the co-expression of FP-WT (35).

Discussion
Here, we demonstrate asymmetric transmission of conformational information between protomers of the putative FP/AT1R heterodimer (summarized in Fig. 8). The AT1R-induced conformational rearrangement in FP was dependent on both expression and activation of G␣ q and possible involvement of the proximal G␣ q -effector PLC. This is consistent with reports showing that PLC␤ is stably associated with G␣ q (51). Furthermore, we demonstrate the AT1R-driven conformational change in FP was predominantly independent of a key distal downstream receptor signaling pathway. We propose that the transmission of information occurs at the level of the membrane and is most likely propagated via a shared G protein as part of a signaling complex. As we noted that even in the absence of G␣ q , the AT1R/FP heterodimer remains intact, suggesting that G␣ q subunits are not critical to the assembly of the receptor heterodimer, although other G protein heterotrimers might also serve this role in their absence. G␤␥ subunits are also important in the formation of GPCR dimers and their associated signaling complexes, as suggested by our previous work (52,53). Our data here suggests that G␣ q acts as a conduit, allosterically connecting the two receptors once assembled into a signaling complex. Surprisingly, ␤-arrestin-biased AT1R ligands (48) also demonstrated a dependence on G␣ q although they elicited no G␣ q activation per se. This would further support the notion that G␣ q plays a key structural role enabling conformational cross-talk between receptors, regardless of the nature of the bound ligand. Therefore, we demonstrate a novel mechanism in which allosteric interactions can transmit information between protomers of a GPCR heterodimer. Our observations are in contrast with a previous report demonstrating independence of the G proteins for conformational cross-talk between the receptors (37) suggesting that heterodimer-spe-

Allosteric interactions in GPCR heterodimers
cific arrangements are possible. We have preliminary data using BRET that FP and OTR do not dimerize (data not shown).
We previously showed that within the FP/AT1R complex, each receptor was capable of modulating the functional output of the other through asymmetric allosteric interactions (38). Asymmetric structural arrangements have been noted in luteinizing hormone oligomers (54), rhodopsin (55), mGluR2/3 heterodimers (56), and leukotriene B4 receptor dimers (18). These studies support the notion that individual protomers in a receptor dimer may interact with a shared G protein through distinct interfaces (see also Ref. 2), suggesting that structural asymmetries may translate into functional or conformational asymmetries. Our results here further strengthen the case for functional FP/AT1R heterodimeric complexes and provide insight into the mechanism by which the two receptors communicate. Although the precise functional consequences of AT1R-induced change in FP conformation are yet to be determined, it also appears to be asymmetric in nature. We observed asymmetry in allosteric communication between receptors, with AT1R modulating FP but not the converse. Furthermore, the AT1R to FP conformational cross-talk in the heterodimer may be biased toward G␣ q/11 , as no effect was observed when we altered G␣ 12/13 or G␣ i function or levels. This could represent a coupling preference of the heterodimer or it may be possible that our biosensors are sensitive to conformations driven by particular G proteins coupled to the heterodimer. Capitalizing on such signal bias and asymmetric conformational crosstalk may provide novel venues for targeting heterodimers, ignored in most current drug discovery programs (57)(58)(59). As we have demonstrated previously, ligand binding to AT1R can modulate the functional output of FP (38). Because both AT1R and FP couple to G␣ q , it is difficult to explore the functional effect of the AT1R-induced conformational effects on FP, as they share a number of common signaling outputs. It is also important to acknowledge that there is also the possibility that the induced conformation may be silent with respect to signaling (60). A larger understanding of unique and shared receptor signaling outputs may help settle this question. FP and AT1R are important targets at the core of many biological functions. AT1R is a primary target in the treatment of hypertension with AT1R antagonists of the sartan family being widely prescribed (61). A role for FP has also been demonstrated in regulating blood pressure where its blockade has been suggested to reduce blood pressure (62). FP is involved in parturition with enhanced PGF2␣ signaling initiating labor by causing smooth muscle contraction of the myometrium (63,64). The AT1R is also expressed in the myometrium with increased levels coinciding with pregnancy (65)(66)(67). Examining the receptor complex may yield novel drug targets in these tissues. An understanding as to how these two receptors commu- nicate at a structural level could facilitate rational drug design as our data indicates that at least as regards receptor conformation, Ang II is a biased ligand for FP, whereas the converse is not true for PGF2␣ and the AT1R. Regardless of the mechanisms, our conformational biosensors could be used to identify new conformational and allosteric connections between known and orphan GPCRs, without the requirement for knowledge about how downstream signaling or receptor trafficking is altered. This may help identify new targets for drug discovery as ligands for one receptor may act as allosteric modulators of heterodimer partners, provide a new vantage point from which to understand receptor dynamics, and foster the development of receptor screens that are portable from cell type to cell type regardless of a priori knowledge about downstream signaling.

Reagents
Unless noted otherwise, all chemicals were reagent-grade and purchased from Sigma. FR900395 was purchased from the Institute of Pharmaceutical Biology (University of Bonn, Germany). PGF2␣ was from Cayman Chemical, whereas AS604872 was a kind gift from Dr. Stéphane Laporte. SBpA, SI, SII, and DVG were synthesized by Lifetein. Western Lightning Plus-ECl was purchased from PerkinElmer Life Sciences.

Membrane preparation
FlAsH-labeled cells were resuspended in 1 ml of homogenizing buffer (10 mM Tris-HCl, pH 7.4, 5 mM EDTA). Cells were homogenized with a glass Teflon homogenizer with 20 updown strokes on ice. The homogenate was transferred to 1.5-ml microtubes and the membrane was pelleted by centrifugation at 13,000 rpm at 4°C for 30 min in a Microlite RF from Thermo IEC. The pelleted membrane was resuspended in TME buffer (50 mM Tris-HCl, pH 7.4, 4.8 mM MgCl 2 , 2 mM EDTA, 100 mM NaCl, 1 M GDP, 30 nM GTP, 250ϫ protease inhibitor mixture (Sigma)). Total protein was quantified using a Bradford reagent (Bio-Rad) and 10 g of protein was loaded/well in a white 96-well plate (Corning).

Conformational biosensor assays
Data collection with the FlAsH/BRET-based conformation biosensors was previously described (39). Briefly, labeled cells in white 96-well plates were treated with 2 M coelenterazine H (Nanolight) for 5 min. Plates were loaded into a Victor X Light Luminescence plate reader (PerkinElmer Life Sciences) and luminescence was captured through F460 and F535 filters for BRET 1. Filtered signals were collected every 1 s with an integration time of 0.2 s for 50 reads. The drug was then injected into the well using the attached injector module, and signals through each filter were collected for another 100 reads. To calculate BRET, the signal through the F535 filter was divided by that through the F460 filter. The basal BRET refers to the average BRET across the first 50 reads prior to injection and the change in BRET due to drug binding (⌬BRET) is calculated by subtracting the basal BRET from the average BRET across the last 50 reads post-injection. Kinetic traces are the raw BRET calculated from the F535/F460 at the sampling frequency.

G␣ q -activation assay
As mentioned above, we used an established polycistronic G␣ q activation sensor that is based on physical separation of G␣ and G␥ subunits following receptor activation (69). Briefly, cells in white 96-well plates (Corning), transfected with the biosensor, and GPCR of interest were washed once and then left with In the absence (A) of G␣ q/11/12/13 , PGF2␣ binding to the sensor receptor can elicit a conformational (black arrow), but this response was blunted for the partner AT1R (red arrow). However, when G␣ q/11 is present (B), full responses to in the FP biosensor are observed in response to either receptor. When the AT1R is tagged with the conformational biosensor, responses are detected in response to Ang II but not PGF2␣ (C and D, respectively), indicating that such conformational responses might be asymmetric depending on the relative positions of the sensor tags and the organization of the dimer/G protein complex.
80 l of Kreb's buffer in a white 96-well plate. Cells were left to rest for 2 h at room temperature before starting the assay. Data were collected on a Victor X Light luminescence plate reader (PerkinElmer Life Sciences) using 410/80 and 515/30 filters for BRET 2 in well mode. Prior to data collection, 5 M coelenterazine 400a (Goldbio) was added to wells and incubated at room temperature for 5 min. The signal was collected through each filter with an integration time of 0.2 s every 0.6 s for 30 reads. The injector module then injected the drug of interest and the data were collected for another 60 reads. BRET was calculated by dividing the signal through the 515/30 filter by that captured through the 410/80 filter. The ⌬BRET was then calculated by subtracting the average BRET pre-injection from the last 30 reads post-injection.

Co-immunoprecipitation
Transfected cells were lysed in solubilization buffer (25 mM HEPES, pH 7.4, 5 mM EDTA, 50 mM NaCl, 10% glycerol, 1% Triton X-100, 1/250 protease inhibitor mixture (Sigma)) for 1 h at 4°C. Lysates were cleared by centrifugation at 13,500 rpm in a Microlite RF (Thermo IEC) microcentrifuge for 10 min. The supernatant was collected, and protein was quantified using a Bradford assay (Bio-Rad). FLAG M2 beads (Sigma) were loaded with 500 g of protein and incubated at 4°C by rocking overnight. The next day, beads were washed 3 times with solubilization buffer pelleting the beads at 1,600 rpm for 2 min at 4°C between each wash. Protein was eluted from the beads in elution buffer (50 mM Tris, pH 7.4, 150 mM NaCl, and FLAG peptide (3 l/100 l of TBS, Sigma) for 20 min at 4°C. Beads were spun down at 1,600 rpm for 2 min and the supernatant was collected. 20 l of sample buffer (62.5 mM Tris pH 6.8, 16.3% glycerol, 2% SDS, 5% ␤-mercaptoethanol, 2 mg of bromphenol blue) was added and samples were heated to 65°C for 15 min.

SDS-PAGE and Western blotting
Samples were loaded on a 10% polyacrylamide gel and run for 1 h at 120 V. Resolved protein was transferred to PVDF membrane under 100 V for 1 h. Membranes were blocked for 1 h in 5% milk in TBST. Anti-RLuc (Millipore) or anti-FLAG (Sigma) primary antibodies were prepared in TBST ϩ 5% milk at a 1:2,000 dilution. The membrane was incubated with primary antibodies at 4°C with gentle rocking overnight. The next day, the membrane was washed 3 times with TBST. Secondary antibodies conjugated to horseradish peroxidase (Sigma) were diluted at 1:20,000 in TBST ϩ 5% milk and incubated with the membrane for 1 h at room temperature. The membrane was washed 3 times with TBST and then treated with enhanced chemiluminescence reagent (PerkinElmer Life Sciences).

In-cell Western blotting
Cells plated in black 96-well plates (Corning), transfected with the proteins of interest, were washed once with Kreb's buffer. Cells were then fixed with 2% paraformaldehyde that had been warmed to 37°C for 10 min at room temperature. Cells were then washed 3 times with Kreb's ϩ 1% BSA. Primary antibodies for HA and myc (Covance) and FLAG (Sigma) were diluted in Kreb's ϩ 1% BSA at 1:200. Cells were incubated with primary antibodies for 1 h at room temperature. Cells were washed 3 times with Kreb's buffers and then incubated with anti-mouse (HA or myc) or anti-rabbit (FLAG) secondary antibodies conjugated to Alexa 488. Secondary antibodies were diluted at 1:500 in Kreb's ϩ 1% BSA and incubated with the cells for 1 h at room temperature. Hoechst DNA stain was also added during the incubation with secondary antibodies at a concentration of 1 g/ml. Fluorescent signals were measured using a Synergy 2 multimode plate reader (BioTek) using 360/40 excitation and 460/40 emission filters for Hoechst and 485/20 excitation and 528/20 emission filters for Alexa 488. The Alexa 488 signal was normalized to the Hoechst signal to quantify the average surface expression of the GPCR of interest/cell.