All cell culture media, reagents, and antibiotics were from Wisent Inc. (St-Bruno, Québec, Canada). All DNA primers for molecular cloning were custom-made by Integrated DNA Technologies Inc. (Coralville, IA, USA). All enzymes and other materials used for molecular cloning were from New England Biolabs Ltd. (Ipswich, MA, USA), except for the Pvu II and Taq I restriction enzymes that were both from Takara Bio Inc. (Noji Higashi, Kusatsu Shiga, Japan). Cell transfection reagent was from Invitrogen, Thermo Fisher Scientific Inc. (Waltham, MA, USA). All chemicals, including Ang II and the AT1R antagonists were from Sigma-Aldrich Inc. (St. Louis, MO, USA) unless otherwise specified. PGF2α and cloprostenol were from Cayman Chemical Company (Ann Arbor, MI, USA). Coelenterazine h was from NanoLight Technologies (Pinetop, AZ, USA).

As recently reported (14), the HEK 293 ΔGα_q/11/12/13 cell line wherein all the genes encoding for G_q, G₁₁, G₁₂, and G₁₃ proteins have been knocked out was generated by simultaneously targeting the GNA12 and the GNA13 genes of the previously established HEK 293 ΔGα_q/11 cells (15), using CRISPR/Cas9 as described previously (16).

The following plasmids pcDNA3/SP-FLAG-hAT1R-WT (12), pcDNA3.1/SP-FLAG-hAT1R-RlucII (17), pcDNA3/SP-HA-hFP-RlucII (18), β-arrestin 1-RlucII (19, 20), β-arrestin 2-RlucII (19, 20) were used. A K44 dynamin mutant in pcDNA3 was a generous gift from Dr. Denis Dupré (Dalhousie University). The different pcDNA3.1 plasmids encoding Gα_q, Gα₁₁, Gα₁₂, Gα₁₃ were from UMR cDNA resource center (www.cdna.org; University of Missouri-Rolla, Rolla, MO, USA). The pcDNA3.1/SP-FLAG-hAT1R-KPVAT-Venus plasmid was created from pcDNA3/SP-FLAG-hAT1R-WT that was amplified by PCR using the forward primer 5′-CCTAGCTAGCTCGAGGCCACCATGAACACGATCATC-3′ and reverse primer 5′-TACCGGTGGCGACCGGTTTCTCAACCTCAAAACATGGTGC-3′ (without a stop codon). The purified PCR fragment obtained was then subcloned in-frame into the NheI and AgeI restriction sites located in 5′ and 3′, respectively, of pcDNA3.1/Venus, which was a kind gift from Dr. Michel Bouvier (Université de Montreál, Montréal, Canada). In a same way, the pcDNA3.1/SP-FLAG-hAT1R[Δ325]-KPVAT-Venus plasmid was generated using the forward primer 5′-CCTAGCTAGCTCGAGGCCACCATGAACACGATCATC-3′ and reverse primer 5′-AAAGGGTGGCGACCGGTTTGGCTTTTGGGGGAATATATTTTAGAAGCTG-3′ (without a stop codon) to amplified pcDNA3/SP-FLAG-hAT1R-WT, and the resulting PCR fragment was then subcloned into the same restriction sites of pcDNA3.1/Venus. The HA-hOTR-Venus plasmid was made by the PCR amplification of a previously described hOTR-YFP construct (21) using the forward primer 5′-TTATGCCTGCGGATCCGAGGGCGCGCTCGCAGCCAACT-3′ and reverse primer 5′-CGCCACCTCCGGATCCCGCCGTGGATGGCTGGGA-3′. The purified PCR fragment obtained was then digested at the BamHI restriction site, and subcloned in-frame by recombination into pIRESpuro3/HA-Venus using the In-Fusion cloning system from Clontech Laboratories (Mountain View, CA, USA). The pcDNA3.1/SP-HA-hFP-WT plasmid was prepared by two rounds of PCR. First, a previously described pIRESpuro3/HA-hFP-WT construct (18) was amplified using the forward primer 5′-ATGAACACGATCATCGCCCTGAGCTACATCTTCTGCC-3′ and reverse primer 5′-ATCCGAATTCCTAGGTGCTTGCTGATTTCTCTGC-3′. The purified PCR fragment was submitted to a second round of PCR amplification using the forward primer 5′-CCTAGCTAGCTCGAGGCCACCATGAACACGATCATC-3′ and the same reverse primer. The resulting PCR fragment was then purified and digested at the XhoI and EcoRI restriction sites located at the 5′ and 3′, respectively, and subcloned into pcDNA3.1 from Invitrogen (Waltham, MA, USA). The pcDNA3.1/SP-HA-hFP-KPVAT-Venus plasmid was produced using the forward primer 5′-CCTAGCTAGCTCGAGGCCACCATGAACACGATCATC-3′ and reverse primer 5′-GGTGGCGACCGGTTTGGTGCTTGCTGATTTTGC-3′ to amplify pcDNA3.1/SP-HA-hFP-WT by PCR. The purified PCR fragment was then digested at the NheI and AgeI restriction sites located in 5′ and 3′, respectively, and subcloned into pcDNA3.1/Venus. All constructs were verified by bidirectional DNA sequencing. The constructs for enhanced bystander BRET to examine AT1R and FP trafficking, Lyn-rGFP, and rGFP-FYVE were used as previously described (17).

The HEK 293 parental and ΔGα_q/11/12/13 cell lines were cultured in 75 cm² plastic flasks containing Dulbecco's Modified Eagle's Medium (DMEM) high glucose supplemented with 5% (v/v) fetal bovine serum (FBS) and 1% (w/v) penicillin–streptomycin (P-S) antibiotics. Cells were cultured in a controlled, humidified environment maintained at 37°C and 5% CO₂ atmosphere until a confluency of 80–100% was reached. For transfection, cells were trypsinized, and 125,000 cells/well in 1 mL of the same fresh media were seeded into 12-well plates. Cells were incubated overnight in at 37°C. After 24 h, media from the plate was replaced by 1 mL/well of DMEM supplemented with 5% (v/v) FBS. Cells were transfected with the appropriate plasmids encoding for AT1R (0.3 μg/well), FP (0.1 μg/well), and the sensors (β-arrestin 1/2-RlucII (0.0025–0.025 μg/well), along with pcDNA3.1(-) for a total of 1.0 μg per well using Lipofectamine® 2000 [2.5(Lipo2000):1(DNA) ratio] and following the manufacturer's instructions. Twenty-four hours post-transfection, the cells were trypsinized, and 40,000–50,000 cells/well were transferred into white 96-well plates (Costar® #3917, Corning, NY, USA) previously coated with poly-L-ornithine hydrobromide (Sigma-Aldrich Inc., St. Louis, MO, USA). Cells were left in a cell culture incubator for another 24 h before performing the experiments. Mycoplasma testing was carried out periodically on all cell lines using the MycoAlertTM kit from Lonza (Rockland, MD, USA).

BRET experiments were performed as described elsewhere (13, 14, 18, 22, 23). For intermolecular BRET experiments, energy-donor proteins were fused to Renilla luciferase (Rluc) II and energy-acceptor proteins were fused to the yellow florescent protein (YFP). In all experiments, coelenterazine h was used as the Rluc/RlucII substrate to generate light with a maximal emission peak at 480 nm, allowing YFP excitation. In a typical experiment, transfected cells from a white 96-well plate were carefully washed once with 150 μL of Kreb's buffer [146 mM NaCl, 4.2 mM KCl, 0.5 mM MgCl₂, 1 mM CaCl₂, 5.9 mM glucose, and 10 mM HEPES pH 7.4, 0.1% (v/v) glucose], and then the cells were incubated in the absence (i.e., with only Kreb's buffer alone) or presence (i.e., with AT₁R antagonists or endocytosis inhibitors) of pretreatments in 95 μL/well of the same fresh Kreb's buffer. The plate was protected from light, and the cells were incubated for 1 h at 37°C for BRET experiments which were conducted at 37°C. A TriStar² LB 942 multimode microplate reader from Berthold Technologies Inc. (Bad Wildbad, 75323, Germany) equipped with the predetermined BRET1 filter pair F485/F530, and a VICTOR X-light multilabel plate reader from Perkin Elmer Inc. (Waltham, MA, USA) equipped with the predetermined BRET1 filter pair F460/F535 nm (200 msec./filter, alternatively) were used to measure BRET ratios. For enhanced bystander BRET we used 410 nm (donor) and 515/40 nm (acceptor) filters. Under these conditions, both plate readers were able to complete one cycle of BRET ratio measurement in 2 min for a 96-well plate. Experiments with 96-well plates were also carried out at once following four consecutive steps in a timely fashion so that a same time interval apply between each well throughout the process: (1) adding the coelenterazine h substrate for light generation (using a repeater pipette); (2) measuring the BRET ratios of basal, unstimulated cells; (3) stimulating the cells with either vehicle or ligand (using a multichannel pipette); (4) measuring the BRET ratios of stimulated, ligand-induced cells. Following a 1 h incubation in Krebs buffer, 25 μL/well of a 30 μM coelenterazine h solution (for a final concentration of 5 μM) was sequentially added to the cells so it takes 2 min to fill up the entire 96-well plate. The basal, unstimulated BRET ratios were then immediately measured over a period of 10 min (5 measuring cycles). Upon completion, the 96-well plate was rapidly removed from the plate reader, and 30 μL/well of either vehicle or 5 μM ligand (e.g., Ang II or PGF2α, for a final concentration of 1 μM in 150 μL) was sequentially added to the cells so, again, it takes 2 min to fill up the entire 96-well plate. The ligand-induced BRET ratios were then immediately measured over a period of 90 min (45 measuring cycles).

BRET ratios were calculated as BRET = λ_530or535/λ_485or460, and ligand-induced BRET changes as ΔBRET = BRET_Ligand − BRET_Vehicle. Conditions associated with ligand and vehicle were performed in triplicate (i.e., 3-wells each) for each biological replicate, and averaged data were used in all calculations. As shown in the figures, BRET data were reported as kinetic traces or bar graphs integrating the area under the curve. Statistical analysis and curve fitting were all carried out using the GraphPad PRISM software v6.0 (La Jolla, CA, USA). For Figure 1, half-time values were calculated using a one-phase association equation and a least squares method to fit the data on a curve by non-linear regression. For Figure 2, ΔBRET data were fit on a curve by non-linear regression using a one-phase exponential dissociation decay and a least squares method. Statistical analysis on normalized ΔBRET data was performed using a Mann-Whitney U-test (p < 0.05), followed by a post-hoc Dunn's test to correct for multiple comparisons (all conditions vs. 0 μg). For Figure 4 statistical analysis on ΔBRET data was performed using an unpaired Student's t-test (p < 0.01), followed by a post-hoc Bonferroni test to correct for multiple comparisons (all conditions vs. PGF2α or cloprostenol). EC₅₀ values from Table 1 were calculated with the following three parameter equation: Response = Bottom + [(Top - Bottom) / (1 + 10(logEC₅₀ − log[A])], where Top and Bottom represented maximal and minimal asymptotes of the curve, [A] is the agonist concentration expressed in (M) and EC₅₀ is the agonist concentration (M) that generated a response half way between the top and bottom.

We used BRET-based biosensors to examine real-time recruitment of β-arrestin 1/2-RlucII to AT1R-Venus alone, FP-Venus alone, and the AT₁R/FP dimer wherein Venus was fused to either AT1R or FP. Ang II stimulation induced a rapid, robust, and sustained β-arrestin 1/2-Rluc recruitment to AT₁R-Venus alone (Figure 1A) typical of a class-B GPCR interaction with β-arrestin (24). While showing an identical kinetic profile, co-expression of FP-WT to form the putative AT1R-Venus/FP-WT dimer dampened the extent of β-arrestin recruitment (Figure 1A). However, we cannot say with certainty that this effect is not due to reduced AT1R expression in the presence of co-expressed FP or simply a different physical spacing of BRET donor and acceptor caused by formation of the heterodimer. β-arrestin recruitment could blocked by the AT1R antagonist losartan and was not detected using a mutated AT1R[Δ325]-Venus with a markedly reduced ability to recruit β-arrestin (Figure 1B). On the other hand, in the AT1R-WT/FP-Venus dimer configuration, however, Ang II failed to recruit β-arrestin 1/2-Rluc (Figure 1A). Altogether, these findings suggest that Ang II occupancy of the AT1R/FP dimer induced a recruitment of β-arrestin 1/2 that is symmetric, or cis, to the AT1R protomer, and that FP may act to reduce this response. By way of comparison, PGF2α stimulation did not induce recruitment of β-arrestin 1/2-Rluc to FP alone, as previously shown (25, 26), or to the AT1R-WT/FP-Venus dimer (Figure 1C). Interestingly, however, it induced a comparatively slower, less robust, yet prolonged β-arrestin 1/2-Rluc recruitment to the AT1R-Venus/FP-WT than Ang II did, suggesting this time that the PGF2α occupancy of FP recruits β-arrestin 1/2 to its AT1R partner protomer in an asymmetrical, or trans, fashion (Figure 1C). Recruitment of β-arrestin was not seen when FP-WT was co-expressed in the context of a dimer with β-arrestin 1/2 recruitment-deficient AT1R[Δ325]-Venus mutant or another peptidergic class-A GPCR, the oxytocin (OT) receptor (OTR, Figure 1D). We previously showed that OTR does not dimerize with FP (13) so results presented here are consistent with this earlier study. In all cases, β-arrestin 2 was shown to be a better responder than β-arrestin 1 for both Ang II and PGF2α stimulation (Figures 1A,C). All responses were also specific since no recruitment was detected when AT₁R alone or FP alone were stimulated with their non-cognate ligands (Figures 1B,D), or when β-arrestin 1/2-RlucII were transfected alone. Furthermore, the PGF2α-induced recruitment of β-arrestin 2-RlucII to the AT1R-Venus/FP-WT dimer was titrated by coexpression of increasing amounts of AT1R-WT to disrupt the reporting dimer by competition, demonstrating the specificity of AT1R-Venus/FP-WT dimer formation critical for asymmetric recruitment of β-arrestin 2 (Figure 2).

In order to provide an independent measure of how the dimer resulted in a new ability for FP to induce trafficking of the AT1R/FP dimer, we used enhanced bystander BRET (17) to follow association of AT1R-RlucII or FP-RlucII with a membrane BRET acceptor (Lyn-rGFP) or early endosome acceptor (rGFP-FYVE). In the case where AT1R-RlucII was co-expressed with WT FP and either of the BRET acceptors, stimulation with either Ang II (Figure 3A) or PGF2α (Figure 3B) led to a time-dependent loss of the membrane signal and an increase in the endosomal population. When FP-RlucII was co-expressed with WT-AT1R and the two BRET acceptors, again PGF2α induced a loss of membrane-localized receptor and a transient increase in an early endosomal population (Figure 3C). This suggests that the two ligands might drive receptors into different trafficking itineraries, discussed below. We also attempted to measure what would happen when we stimulated this combination of receptors (FP-RlucII and WT-AT1R) with Ang II. There was little change in bystander BRET for either location-specific biosensor likely because in our hands the AT1R expresses so much better than the FP and thus most of the AT1R in this configuration was not coupled to FP (data not shown).

Next, we assessed to what extent the AT1R-Venus/FP-WT dimer could asymmetrically recruit the β-arrestin 2-Rluc by examining a panel of five chemically distinct FP ligands, some of which have clinical indications (27). Except for AL-8810, all the ligands we tested, cloprostenol, fluprostenol, latanoprost, and tafluprost, were found to recruit β-arrestin 2-Rluc in an asymmetric fashion with EC₅₀ values in the low nanomolar to the picomolar range (Table 1). While all the other tested ligands exhibit full agonist properties, AL-8810 is the only one shown to display partial agonist properties (28), and had an EC₅₀ value for β-arrestin 2-RlucII recruitment in the micromolar range (Table 1). Further, as AL-8810 promotes activation of ERK1/2 by transactivation of the epidermal growth factor receptor (25), this may be possible explanation for this discrepancy. It is clear, however, that a number of FP ligands can recruit β-arrestin to the dimer.

Having previously shown allosteric crosstalk between FP and AT1R in that antagonists for either receptor modulated signaling via the partner receptor, we next tested how different AT1R antagonists modulated PGF2α-mediated recruitment of β-arrestin to the dimer. PGF2α responses were regulated by both temilsartan and azilsartan (Figure 4A), while cloprostenol responses were regulated by temilsartan only (Figure 4B). On the other hand, other AT1R antagonists we tested, irbesartan, candersartan, and losartan, did not allosterically modulate the responses induced by either PGF2α or cloprostenol.

Using a HEK 293 line with deletion of G_q/11 and G_12/13 (13), we next assessed how different G proteins might be involved in recruitment of β-arrestin to the FP/AT1R dimer. Ang II-induced recruitment of β-arrestin 1/2-RlucII to the AT₁R-Venus/FP-WT dimer was measured in the parental and the ΔGα_q/11/12/13 cell line. The absence of G proteins reduced Ang II-induced response by approximately 50% for β-arrestin 1 and β-arrestin 2 (Figures 5A,B). Individual G proteins were then restored to the ΔGα_q/11/12/13 cell line to assess which might be able to rescue this loss of β-arrestin recruitment. Replacement of Gα₁₁, Gα₁₂, or Gα₁₃ partially rescued β-arrestin 1 recruitment but surprisingly, not Gα_q (Figure 5A). A similar finding was noted for β-arrestin 2 (Figure 5B). Although the data were considerably noisier for the responses measured after treatment with PGF2α, similar patterns were detected for β-arrestin 1 recruitment (Figure 5C, also see inset). Interestingly, in this case, re-expression of any of the four Gα subunits was able to partially rescue β-arrestin 2 recruitment mediated by PGF2α (Figure 5D). Although we did not control for expression of the different G protein subunits, these data suggest a complicated interplay between G protein and β-arrestin coupling, inconsistent with a simple interpretation of these events.

We previously showed that stimulation of the receptor pair with either ligand resulted in receptor internalization of the dimer when reconstituted using a split Venus protein complementation (12). The data in that paper, using confocal imaging, hinted that distinct endocytic routes might be taken in response to PGF2α compared to Ang II. Given the role of β-arrestin recruitment in the post-stimulation trafficking of receptors, we examined how different inhibitors of endocytosis affected interactions between the FP/AT1R heterodimer and β-arrestin1/2. As before, we first examined the effects of Ang II on β-arrestin1/2 recruitment. In both cases, neither concanavalin A or a dominant negative dynamin K44 mutant had significant effects compared to control (Figures 6A,B). However, the extent of the recruitment of β-arrestin was decreased significantly by phenylarsine oxide and potentiated by sucrose. The responses to PGF2α were qualitatively different. In this case, again there was no effect of dynamin K44, but both sucrose and phenylarsine oxide reduced β-arrestin1/2 recruitment while concanavalin A potentiated the interaction (Figures 6C,D). These results again suggest that the trafficking itineraries of the heterodimer are distinct following stimulation with either Ang II or PGF2α. Taken together our data suggest that the dimer adopts different fates when stimulated by distinct ligands.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.