Oligomerization of wild type and nonfunctional mutant angiotensin II type I receptors inhibits galphaq protein signaling but not ERK activation.

The 7-transmembrane or G protein-coupled receptors relay signals from hormones and sensory stimuli to multiple signaling systems at the intracellular face of the plasma membrane including heterotrimeric G proteins, ERK1/2, and arrestins. It is an emerging concept that 7-transmembrane receptors form oligomers; however, it is not well understood which roles oligomerization plays in receptor activation of different signaling systems. To begin to address this question, we used the angiotensin II type 1 (AT(1)) receptor, a key regulator of blood pressure and fluid homeostasis that in specific context has been described to activate ERKs without activating G proteins. By using bioluminescence resonance energy transfer, we demonstrate that AT(1) receptors exist as oligomers in transfected COS-7 cells. AT(1) oligomerization was both constitutive and receptor-specific as neither agonist, antagonist, nor co-expression with three other receptors affected the bioluminescence resonance energy transfer 2 signal. Furthermore, the oligomerization occurs early in biosynthesis before surface expression, because we could control AT(1) receptor export from the endoplasmic reticulum or Golgi by using regulated secretion/aggregation technology (RPD trade mark ). Co-expression studies of wild type AT(1) and AT(1) receptor mutants, defective in either ligand binding or G protein and ERK activation, yielded an interesting result. The mutant receptors specifically exerted a dominant negative effect on Galpha(q) activation, whereas ERK activation was preserved. These data suggest that distinctly active conformations of AT(1) oligomers can couple to each of these signaling systems and imply that oligomerization plays an active role in supporting these distinctly active conformations of AT(1) receptors.

The 7-transmembrane or G protein-coupled receptors relay signals from hormones and sensory stimuli to multiple signaling systems at the intracellular face of the plasma membrane including heterotrimeric G proteins, ERK1/2, and arrestins. It is an emerging concept that 7-transmembrane receptors form oligomers; however, it is not well understood which roles oligomerization plays in receptor activation of different signaling systems. To begin to address this question, we used the angiotensin II type 1 (AT 1 ) receptor, a key regulator of blood pressure and fluid homeostasis that in specific context has been described to activate ERKs without activating G proteins. By using bioluminescence resonance energy transfer, we demonstrate that AT 1 receptors exist as oligomers in transfected COS-7 cells. AT 1 oligomerization was both constitutive and receptor-specific as neither agonist, antagonist, nor co-expression with three other receptors affected the bioluminescence resonance energy transfer 2 signal. Furthermore, the oligomerization occurs early in biosynthesis before surface expression, because we could control AT 1 receptor export from the endoplasmic reticulum or Golgi by using regulated secretion/aggregation technology (RPD TM ). Co-expression studies of wild type AT 1 and AT 1 receptor mutants, defective in either ligand binding or G protein and ERK activation, yielded an interesting result. The mutant receptors specifically exerted a dominant negative effect on G␣ q activation, whereas ERK activation was preserved. These data suggest that distinctly active conformations of AT 1 oligomers can couple to each of these signaling systems and imply that oligomerization plays an active role in supporting these distinctly active conformations of AT 1 receptors.
The 7-transmembrane (7TM) 1 or G protein-coupled receptors such as the angiotensin II type 1 (AT 1 ) receptor constitute the largest group of cell surface membrane receptors and mediate a vast array of biological effects in response to hormones, neurotransmitters, and sensory stimuli. The 7TM receptors activate or interact with numerous signaling proteins including heterotrimeric G proteins, arrestins, adaptor proteins, and extracellular signal-regulated kinases 1 and 2 (ERK1/2) at the intracellular face of the plasma membrane (1). Although 7TM receptors traditionally have been considered to work as monomeric entities, increasing evidence has shown that these receptors form both homo-and heterodimers or oligomers in vivo and in vitro. Whereas heterodimerization may provide a means to expand pharmacological diversity, the functional role of homodimerization is less well defined (2,3). 7TM receptor heterodimerization has been shown to modify biological function, receptor trafficking, ligand binding properties, and signal transduction of several 7TM receptors (2,3). The GABA B "receptor pair" provides an example of two different receptor subunits, the GABA B1 and GABA B2 , that both are necessary for receptor surface expression and function (4). The bradykinin B2 has been shown to potentiate AT 1 receptor signaling drastically (5), whereas the angiotensin II, type II, receptor inhibits AT 1 receptor signaling (6). The ␤ 1 /␤ 2 -adrenergic receptor heterodimerization regulates MAPK signaling efficacy, whereas G protein activation remains unaffected (7).
The role of 7TM receptor homodimerization is experimentally more difficult to assess. Studies of functionally compromised mutants co-expressed with wild type receptors suggest that homodimerization perhaps plays a role in receptor biogenesis, ligand binding, and receptor signaling (2,3). Mutant receptors of the ␤ 2 -adrenergic receptor, CCR5, and dopamine D 2 receptors have exerted a dominant negative effect on wild type receptor cell surface expression (8 -10). In contrast, surface expression of a mutant ␤ 2 -adrenergic receptor could be rescued by co-expression of WT receptor (11), and internalization of an ␣-factor receptor that could not internalize was rescued by co-expression of wild type receptor (12). Several studies suggest that dimerization may be important for G protein coupling. Wild type ␣-factor receptor signaling is inhibited by co-expression of a signaling-deficient receptor (12), and it is possible to inhibit both ␤ 2 -adrenergic receptor activation and dimerization with a peptide derived from transmembrane domain 6 of the receptor (13). A double cysteine M3 muscarinic receptor that has lost the ability to form covalent dimers and multimers displays an approximately 50-fold reduction in binding affinity compared with wild type and a greater than 10,000-fold reduc-tion in agonist potency (14), and recently mutations that abolish dimerization of the ␣-factor (15), CCR2, and CCR5 (16) receptors have been shown to also decrease or abolish agonistinduced signaling.
Although it is well established that 7TM receptors form homodimers, we are still far from understanding the functional consequences of this process. Two types of experiments are currently of great interest: experiments directed at defining the receptor domains involved in oligomerization, and studies examining the role of oligomerization for ligand binding and signaling. We speculated that the AT 1 receptor could be a means to advance our understanding of how receptor dimerization affects signaling. This receptor has been reported to engage in both homo-and heterodimerization (5,6,(17)(18)(19), and it interacts with a diverse array of signaling and scaffold proteins (20). Most important, it has been suggested that AT 1 receptor G protein and ERK activation pathways can be separated. Thus, different angiotensin II peptide analogues and AT 1 receptor mutants can activate ERK1/2 without stimulating G proteins (21)(22)(23)(24)(25). In the present study, we examined the effect of homodimerization on Ang II-induced G protein and ERK1/2 signaling. We created two receptor mutants defective in signaling or binding. Most interesting, when co-expressed with wild type receptors, the mutant receptors inhibit G protein signaling, whereas ERK activation and ␤-arrestin recruitment remained unaffected, suggesting that the receptor oligomerization has differential effects on different signaling systems. These results support and expand the notion that AT 1 receptor oligomerization plays an important role for activation of cellular signaling systems.

EXPERIMENTAL PROCEDURES
Materials-Coelenterazine h was purchased from Molecular Probes; DeepBlueC TM , pRLuc, and pGFP 2 vectors were from Biosignal Packard. LipofectAMINE TM 2000, cell culture media, and serum were from Invitrogen. Cell culture plates were from Techno Plastic Products AG (Trasadingen, Switzerland) except for 48-well plates that were obtained from Nunc. Angiotensin II, other ligands, and chemicals were all from Sigma.
cDNA Constructs-The EGFP-tagged bovine ␤-arrestin 2, as described previously (26), was kindly provided by Jeffery Benovic. The AT 1 -luc and GFP 2 and CaR-Luc are described in Ref. 27. ETa-luc was cloned by removing the stop codon in an N-terminally c-Myc-tagged version of the human ETa receptor, kindly provided by T. Nakagawa and as described previously (28), and inserting this product into the reading frame of the Rluc-NI at an EcoRV site. The PCR amplification was performed by using DNA oligonucleotides with 5Ј-phosphorylated ATGs (resembling an open EcoRV site). To generate the EGF TM -luc receptor, we PCR-amplified the first 2010 bases, i.e. the sequence coding for the N-terminal region and the transmembrane segment of the human EGF receptor (NM_005228; NCBI-Entrez Nucleotides Database), by using a 5Ј-DNA oligonucleotide with a HindIII site and three additional bases, AAG, to ensure digestion of the site, and a 3Ј-oligonucleotide with a KpnI site and three additional bases, AAG, to ensure digestion of the site. After digestion with KpnI and HindIII, this product was inserted in the reading frame of the pLuc N3 vector between the HindIII and KpnI sites. Both fusion receptors were verified by sequencing. Point mutations were introduced using the QuickChange mutagenesis protocol (Stratagene). Each mutant receptor was subcloned back into the original plasmid vector and sequenced to eliminate potential noncoding PCR-generated mutations.
BRET 2 Assay-Except for minor differences, the BRET 2 assay was performed as described previously (27). (BRET 2 is a specific patented assay from Packard Biosciences that uses bioluminescence from luciferase to GFP 2 , a second generation modification of GFP.) Briefly, 0.5 million COS-7 cells per well were seeded into a 6-well plate and grown in 10% fetal calf serum/Dulbecco's modified Eagle's medium (DMEM) overnight. After 24 h the cells were transfected using Lipofect-AMINE TM 2000 (Invitrogen) according to the manufacturer's protocol (varying amounts of DNA were used to ensure equal expression levels between the different receptors). After 2 days, the cells were washed twice with PBS to remove the indicator dye before detachment in PBS. The cells were then spilt into two portions. The first portion was used to examine the GFP levels by fluorescent measurements as described previously (27), and next the Rluc expression was determined by measuring the Coelenterazine h induced luminescence as described previously (27). The second portion of the cells was submitted to DeepBlueC excitation, and the luminescence at the dual bands (515/30 nm and 410/80) was measured on a fusion reader (Packard Instrument Co.). The BRET 2 ratio was determined according to the principle described previously (29). BRET 2 ratio equals ((emission (515/30) Ϫ (emission (410/80) ϫ Cf))/(emission (410/80), where Cf denotes the cross-Rluc luminescence cross-talk ratio into the 515/30 filter defined as emission (515/30)/(emission (410/80) when Rluc expressed alone is excited.
Regulated Secretion/Aggregation Technology-This assay was developed by Rivera et al. (30) and is available from Ariad Pharmaceuticals at www.ariad.com/regulationkits. To perform this assay, we seeded 2.5 ϫ 10 6 COS-7 cells in a p10 dish, and we transfected the cells with 10 g of F m4-AT 1 -EEA(A) cDNA and in the co-expression experiments also with 2 g of wild type receptor DNA using LipofectAMINE TM 2000 (Invitrogen) according to the manufacturer's protocol. After 24 h, the cells were split into 48-well plates coated with 0.25% poly-L-lysine (100,000 cells/well) and stimulated (with or without AP21998 at 2 M) for 2 h prior to performing the inositol phosphate (IP) accumulation or binding assays as described below.
Whole Cell Competitive Radioligand Binding Assay-Except for minor changes, this assay was performed as described previously (31). Cells were prepared as described above for the regulated secretion/aggregation technology. Next, cells were washed once in cold Hanks' balanced salt solution (HBSS), supplemented with CaCl 2 (0.9 mM) and MgCl 2 (1.05 mM), and then cooled down to 4°C for 30 min. The HBSS was aspirated, and cells were incubated at 4°C for 3 h in 0.25 ml of HBSS containing the radioligand ([tyrosyl-3,5-3 H]angiotensin II (5-L-isoleucine)) at various concentrations either alone or in the presence of 10 Ϫ5 M unlabeled Telmisartan. After incubation, cells were washed twice with ice-cold HBSS before addition of 0.5 ml of lysis buffer (1.0% Triton X-100, 50 mM Tris/HCl, pH 7.5, 100 mM NaCl, 5 mM EDTA) and incubated for 30 min at room temperature. Finally, the lysis buffer was transferred to a vial containing 4 ml of Ultima-Gold (Packard Instrument Co.); vials were capped and shaken, and 1 h later total radioactivity was measured on a Tri-Carb 2900 TP liquid scintillation analyzer (Packard Instrument Co.).
IP Assay-Briefly, cells were transfected as described for the regulated secretion/aggregation technology. After 1 day, the cells were seeded into 48-well plates coated with 0.25% poly-L-lysine (100,000 cells/well) and incubated in inositol-free DMEM supplemented with myo-[2-3 H]inositol (1 Ci/ml) (Amersham Biosciences). Twenty hours after application of the radioligand, the cells were assayed as described previously (27).
ERK Phosphorylation Assay-Except for minor modifications, this assay was performed as described previously (32,33). Briefly, 2.5 million COS-7 cells were seeded into a p10 dish and grown in 10% fetal calf serum/DMEM overnight. After 24 h, the cells were transfected using LipofectAMINE TM 2000 according to the manufacturer's protocol. On the day after transfection they were seeded onto 6-well plates. Next, 45 h after transfection, the cells were serum-starved for 3 h, incubated with agonist for 12 min at 37°C, and then lysed. SDS-PAGE and immunoblotting were performed as described previously (33), and the bands were visualized using the enhanced chemiluminescence system (Amersham Biosciences). To quantify the densitometry of the bands, the gels were scanned, and the density of each band was measured using NIH-Image 1.62.
␤-Arrestin Recruitment-This assay was performed as described previously (29) except for minor changes. Briefly, 2.5 million COS-7 cells were seeded into a p10 dish and grown in 10% fetal calf serum/DMEM overnight. After 24 h, the cells were transfected using Lipofect-AMINE TM 2000 according to manufacturer's protocol. We used 3 g of EGFP-bovine-␤-arrestin 2; either 3 g of vector or AT 1 receptor together with varying amounts of Rluc-tagged receptor DNA to reach equal expression levels between the different receptors were studied. Next, 48 after transfection cells were washed twice in PBS and detached. Cells were divided into two parts; the first part was used to examine GFP and luciferase expression as described above, and the other part was submitted to either basal or agonist treatment for 20 min at room temperature before measuring the BRET 2 ratios as described above.
Data Analysis-All pharmacological data were analyzed using GraphPad Prism. For whole cells radioligand binding data were analyzed by one-site competitive binding analysis, whereas phosphatidylinositol hydrolysis data were analyzed using nonlinear regression curve fitting. Regarding statistical analysis, all values obtained for variant receptors were directly compared with those observed for WT receptor. A Bartlett test confirmed that there were no significant differences in the variance of the compared groups of data sets. Subsequently, onetailed, paired Student's t tests were performed. All statistically significant results at the p Ͻ 0.05 level are reported.

AT 1 Receptors Form Constitutive Homodimeric Complexes-
We used two different approaches to examine AT 1 -homodimerization: the BRET 2 technology from Packard Instrument Co., and the regulated secretion/aggregation technology (RPD TM ) from ARIAD Pharmaceuticals (30). To analyze homodimerization using BRET 2 experiments, AT 1 and control receptors were fused to either Renilla luciferase (Rluc) or green fluorescent protein 2 (GFP 2 ) as described previously (27), and these constructs were expressed in COS-7 cells either alone or together (Fig. 1A). When AT 1 -luc and AT 1 -GFP 2 were co-expressed and Rluc was excited with the substrate DeepBlueC, we detected a robust BRET 2 signal with a value of 0.15 Ϯ 0.01 (Fig. 1A). The BRET 2 signal was not affected by incubation with the receptor agonist Ang II or the inverse agonist Telmisartan (Fig. 1A).
To verify that the BRET 2 signal results from specific proteinprotein interactions, we performed two types of experiments. First, we co-expressed three other luciferase-tagged receptors with the AT 1 -GFP 2 receptor at the same expression levels (expression data not shown). For this experiment, we chose two 7TM receptors (the endothelin 1a (Eta) receptor and the calcium-sensing receptor (CaR)) and a truncated version of the EGF receptor dubbed EGF TM . Although we observed BRET 2 signals in these experiments, these were significantly smaller (0.01-0.08; p Ͻ 0,05) than that of the AT 1 receptors (Fig. 1A). Next, we co-expressed the wild type receptors (AT 1 , Eta, CaR, and EGF TM ) together with AT 1 -luc and AT 1 -GFP 2 receptors at similar expression levels as above to see if they could reduce the BRET 2 signal. Confirming that most of the BRET 2 signal is caused by specific homodimeric protein-protein interactions, the AT 1 -WT lowered the AT 1 -luc/AT 1 -GFP 2 BRET 2 signal significantly to a value of 0.06 (Ϯ0.02; p Ͻ 0,05), whereas none of the three other wild type receptors did (Fig. 1B).
Studies by others have indicated that some 7TM receptors form homodimeric/homo-oligomeric complexes as early as the ER (34 -36). To examine whether AT 1 receptors dimerize during biosynthesis, we used the regulated secretion/aggregation technology (RPD TM ) developed by Rivera et al. (30). In principle, the protein of interest is N-terminally fused to a protein (dubbed F m ) that accumulates as aggregates in the ER and Golgi. However, incubation with a synthetic small molecule drug (AP21998) alleviates aggregation and allows the fusion proteins to escape the ER and Golgi in an AP21998-gated fashion. Furthermore, a furin cleavage site has been implemented so that the F m repeats can be enzymatically removed in the Golgi to ensure irreversible disaggregation of the proteins of interest (30). Our idea of using this assay to study homooligomerization was to generate a mutant receptor that showed AP21998-gated surface expression but did not activate Ang II-mediated IP accumulation. If this receptor dimerizes with the AT 1 -WT, the prediction would be that co-expression of these two receptors would result in AP21998-gated Ang IImediated signaling. We therefore N-terminally fused a signaling-deficient AT 1 R mutated in the DRY motif (D125E/R126E/ Y127A/M134A) with four consecutive F m fusion domains, and we expressed this fusion receptor (dubbed F m4 -AT 1 -EEA(A)) in COS-7 cells. When exposed to AP21998, the F m4 -AT 1  (Ϯ0.02) to 17 (Ϯ0.04) fmol/10 5 cells, p Ͻ 0,05), but as expected the receptor was still not able to induce IP accumulation in response to Ang II (Fig. 1, C and D). We then co-expressed AT 1 -WT and F m4 -AT 1 -EEA(A) receptors, and we analyzed the effect of AP21998 on the Ang II-induced IP accumulation. Preincubation with AP21998 approximately doubled the IP response, showing that the AT 1 -WT expression is indeed AP21998-gated p Ͻ 0.05 (Fig. 1E). This finding strongly suggests that AT 1 receptor dimerization occurs early during biosynthesis and that the dimer interaction occurs with relatively high affinity. Furthermore, to analyze if F m4 -AT 1 -EEA(A) retention of the AT 1 -WT was because of a specific protein-protein interaction rather than ER or Golgi clotting that unspecifically may retain protein, we subsequently co-expressed the ETaR and CaR with the F m4 -AT 1 -EEA(A). These experiments confirmed the specificity of the response because the endothelin II and calcium-induced IP accumulation responses were independent of AP21998 (Fig. 1F).
Co-expression of Mutant Receptors and Wild Type Inhibit G␣ q Protein but Not ERK Activation-Co-expression studies of wild type and mutant receptor have so far had three different outcomes. 1) The mutant receptor does not affect wild type signaling; 2) the mutant receptor inhibits wild type signaling either by inhibiting surface expression or the activation of signaling (8 -10); or 3) the wild type receptor can functionally rescue the mutant receptor (11,12,37). In addition, co-expression of two nonfunctional receptors can create functional receptors (17,38).
Several studies employing the AT 1 receptor and other 7TM receptors imply the existence of distinct "active" conformations of 7TM receptors. Thus, it has been suggested that some receptor conformations selectively can activate ␤-arrestin and ERK activation, whereas a different set of receptor conformations cause G␣ q signaling (21)(22)(23)(24)(25). We therefore asked the following question: Does receptor dimerization play a role for these states of activation? If so, would a selective binding or signalingdeficient receptor mutant affect all confirmations equally? To answer these questions, we generated two mutants as follows: the AT 1 -EEA(A) that does not activate G␣ q protein but binds Ang II, and AT 1 -K199A, mutated in the Ang II-binding site. We co-expressed these with AT 1 -WT and monitored the effects on signaling, including G␣ q -induced IP accumulation, ERK activation, and ␤-arrestin 2 recruitment.

AT 1 Mutants Exert a Dominant Negative Effect on Wild Typemediated G Protein
Signaling-To be able to detect a dominant negative effect of a mutant or to detect a functional rescue of such a receptor, it is imperative that the assay can detect both decreases and increases in signaling. We therefore transfected increasing amounts of AT 1 receptor cDNA into COS-7 cells, and we measured the Ang II-induced IP accumulation ( Fig. 2A). We Concentration-response curves for Ang II-induced IP accumulations are shown by representative curves from at least three experiments. Data are presented in cpm/well versus the Ang II concentration. A, to identify the "dynamic range" of receptor levels, increasing amounts of AT 1 -WT receptor DNA were transfected into COS-7 cells. B, to test the ability of the mutant receptors to induce IP accumulation, 2 g of AT 1 -WT or 6 g of mutant receptor DNA was transfected into COS-7 cells. C and D, to test the effect of the mutant receptors on wild type receptor IP hydrolysis, wild type (2 g) and mutant receptor (6 g) DNA were co-expressed into COS-7 cells either with (D) or without (C) G␣ q cDNA (2 g). E, to test whether the mutant AT 1 receptors inhibit AT 1 -WT IP accumulation specifically, Eta receptor (EtaR) and CaR were co-transfected with the AT 1 receptor mutants into COS-7 cells, and endothelin II (1 M) or Ca 2ϩ (50 mM)-induced IP accumulation was assessed. Data were normalized to the wild type receptor agonistinduced response when it was expressed alone. chose to use 2 g of DNA per 10-cm culture dish, because the system is very dynamic at this level of receptor expression ( Fig.  2A). We next confirmed that the mutants did not activate IP accumulation (Fig. 2B) by testing varying amounts of transfected receptor DNA (0.1-20 g). No signaling was observed at any receptor amount, although surface expression was confirmed by ligand binding experiments for AT 1 -EEA(A) and fluorescent antibody staining against an N-terminally incorporated Myc tag for AT 1 -K199A (data not shown).
To test the effect of the mutants on G␣ q activation by the wild type AT 1 receptor, we co-expressed the receptors. Most interesting, the mutant-WT receptor combination exhibited impaired IP accumulation in response to Ang II as compared with WT receptors alone p Ͻ 0.05 (Fig. 2C), suggesting the mutant receptors exert a dominant negative effect on WT receptor G␣ q signaling. Co-transfection with G␣ q did not change this result, suggesting the dominant negative effect was not because of trapping of G protein with p Ͻ 0.05 (Fig. 2D). Furthermore, the dominant negative effect was AT 1 -WT-specific because the IP accumulation response of the ETa and CaR was not affected by the presence of the AT 1 -receptor mutants (Fig. 2E). Finally, to analyze for a functional rescue, we co-expressed the two mutant receptors defective in either ligand binding or G protein activation; however, none of this combination enhanced IP hydrolysis in response to Ang II (data not shown).
AT 1 Mutants Do Not Affect ERK Activation-We also tested the mutant receptor effects on Ang II-induced ERK phosphorylation. Most interesting, both of the mutant-WT AT 1 R combinations induced ERK phosphorylation to the same degree as the wild type AT 1 receptor expressed alone (Fig. 3C). In contrast, neither AT 1 -EEA(A) nor K199A expressed alone activated ERK (Fig. 3B). As mentioned above, the receptors were expressed at levels where we could detect both decreases and increases of ERK-activated signaling (Fig. 3A). We also coexpressed the mutant receptors to look for functional rescue of ERK signaling, but we failed to find any (data not shown). The finding that the mutant-WT combination fully activated ERK signaling also supported the notion that the observed dominant negative effect on G␣ q signaling was caused by a specific protein-protein interaction at the plasma membrane rather than reduced surface expression of wild type AT 1 receptor.
Compromised AT 1 Mutants Do Not Affect Wild Type-induced ␤-Arrestin 2 Recruitment-Because the mutant-WT AT 1 R combinations exhibited dampened G protein activation, but were able to support ERK phosphorylation, we tested the effect of the mutants on AT 1 -WT-luc Ang II-stimulated ␤-arrestin 2 recruitment. In general, 7TM receptors are thought to recruit ␤-arrestin in a G protein-dependent manner, because GRK2 phosphorylation is a preceding event (39). In contrast, some receptors including the AT 1 receptor have been shown to support G protein-independent ␤-arrestin recruitment in specific contexts (21)(22)(23)(24)(25). Furthermore, ␤-arrestin recruitment has been reported to precede and enhance ERK activation (25,40).
Thus, in line with the ERK activation data, we found the AT 1 receptor mutants did not compromise the ability of the AT 1 -WT-luc to robustly recruit ␤-arrestin 2 in response to Ang II. To perform this analysis, AT 1 -luc was co-expressed with EGFPtagged ␤-arrestin 2 with and without the receptor mutants (Fig. 4A).
AT 1 R-WT Activation Rescues ␤-Arrestin 2 Recruitment by AT 1 -K199A-luc-We fused Rluc to the C-terminal end of the AT 1 -K199A receptor to analyze its ability to recruit ␤-arrestin 2 in the presence of 100 nM Ang II. This receptor did not produce any significant ␤-arrestin 2 recruitment in response to Ang II (Fig. 4B). However, when we also co-expressed the wild type AT 1 receptor, Ang II stimulation produced a robust in-crease in the BRET signal of 1.7-fold (p Ͻ 0.05). In contrast, co-expressing the AT 1 -WT with other luciferase-tagged receptors (CaR-luc, Eta-luc, and EGF TM -luc) did not produce any significant increase in the BRET signal in the presence of Ang II, suggesting that the wild type AT 1 receptor specifically rescues ␤-arrestin 2 recruitment of AT 1 -K199A (Fig. 4B).

FIG. 3. Mutant AT 1 receptors do not inhibit wild type-induced ERK activation.
Western blot analysis of AT 1 -mediated ERK phosphorylation is shown. A, to verify that the level of ERK activation would be sensitive to changes in receptor function when 2 g of receptor was expressed, we tested the dynamic range of receptor levels by expressing increasing amounts of AT 1 -WT receptor DNA in COS-7 cells and stimulated as described under "Experimental Procedures." Phosphorylated and total ERK (T-ERK) are shown for un-stimulated cells and cells stimulated with either 1 or 100 nm of Ang II. Gel images are representatives from four different experiments. ERK phosphorylation (P-ERK) was assessed by densitometric analysis of the phosphorylated ERK band intensity. B, 2 g of AT 1 -WT or 6 g of mutant AT 1 receptor DNA was transfected into COS-7 cells and stimulated as described under "Experimental Procedures." Phosphorylated and total ERK are shown for un-stimulated cells or cells stimulated with either 1 or 100 nm of Ang II. Gel images are representatives from four different experiments. C, to test the effect of the mutants on wild type receptor ERK activation, wild type (2 g) and mutant receptors (6 g) receptor DNA were co-expressed into COS-7 cells, and the cells were treated as described under "Experimental Procedures." Phosphorylated and total ERK proteins are shown for un-stimulated cells or cells stimulated with either 1 or 100 nm of Ang II, by representative gel images from nine experiments. To assess the relative efficacies of activation, densitometric gel quantification of the phosphorylated ERK band intensities was performed, and the density of the bands was normalized with reference to the wild type receptor maximum response values.

AT 1 Receptors Dimerize Constitutively and Specifically-By
using BRET 2 analysis and a novel technique that enables ligand-gated release of receptors to the plasma membrane, we have shown that AT 1 receptors specifically form homo-oligomeric complexes in living cells. This oligomerization is constitutive, i.e. it is unaffected by both receptor agonist or antagonist, and it occurs early during biosynthesis before surface expression, probably as early as the ER. In addition, we have shown that a receptor combination consisting of AT 1 wild type and mutant receptors, deficient in either ligand binding or signaling, fully supported ERK phosphorylation and ␤-arrestin 2 recruitment but not G␣ q signaling. These observations support and extend previous studies of 7TM receptor oligomerization. Previously, AT 1 receptor dimerization has been suggested by using various approaches. 1) Using Angiotensin affinity labeling, or chemical cross-linking experiments, AT 1 receptor complexes with a molecular weight corresponding to a dimeric complex have been identified in transfected cell lines and in several organs such as liver adrenal pituitary and brain (41)(42)(43)(44)(45). 2) High affinity ligand binding has been reconstituted by co-expression of two receptors (AT 1 -K102A and -K199A) that did not bind ligand when expressed individually (17). 3) The AT 1 has been suggested to heterodimerize with three different receptors as follows: the bradykinin B2; angiotensin II, type II, and the ␤ 2 -adrenergic receptor (5,6,19). In contrast to our findings, AbdAlla et al. (18) observed increased levels of dimerization in the presence of Ang II using cross-linking and immunoprecipitation techniques. Inconsistent conclusions as to whether 7TM receptor oligomerization is ligand-dependent have also been noted in studies on other receptors and can probably be attributed to different specific contexts such as cell cultures, techniques, and materials employed (16).
AT 1 Receptor Dimerization Occurs during Biosynthesis-We have found that AT 1 receptors dimerize with relatively high affinity during biosynthesis, before surface expression, most likely in the ER or Golgi. This notion is in line with studies of other 7TM receptors such as the vasopressin V1/V2, C5a, and GABA B receptors, suggesting that dimerization during biosynthesis is a general phenomenon for 7TM receptors (34 -36). To make our conclusion, we used the ligand-gated cell surface expression/secretion system, denoted regulated secretion/aggregation technology (RPD TM ) from ARIAD Pharmaceuticals and developed by Rivera et al. (30). With this system, we show that the wild type receptor when co-expressed with an AP-21998-gated F m fusion receptor (F m4-AT 1 -EEA(A)) appears AP-21998-gated, whereas both ETa and CaR were unaffected. As described under "Results," this system was originally designed to enable regulation of protein secretion in an AP21998-gated fashion. This system should be useful as a tool for others to regulate 7TM receptor surface expression and to study proteinprotein interactions of these receptors.
Different States of Activated Receptor Conformations Exist-By employing co-expression of wild type receptors and mutated receptors defective in either binding or signaling, we found that G protein activation was abolished, whereas ERK activation and ␤-arrestin 2 recruitment remained intact. Furthermore, we performed careful controls to confirm that both the G protein and the ERK activation assays could detect activity changes if any were present. These observations suggest that the structural demands for ERK activation and ␤-arrestin recruitment are fulfilled in a complex consisting of wild type and mutant receptors, whereas the conformation required for G protein activation cannot be achieved. Therefore, as supported by previous findings (21)(22)(23)(24)(25), distinct active conformations of AT 1 receptors that activate different signaling partners exist. Most important, although some of these receptor states may be able to activate all signaling proteins, some states are clearly only capable of activating distinct signaling proteins such as ERKs and ␤-arrestin.
Two different scenarios can explain the observation that the mutant receptors inhibit wild type receptor-mediated G protein signaling. Either oligomer formation with intact receptors is required for G protein coupling or, alternatively, receptor dimerization may not be necessary for coupling; however, the mutant receptor somehow compromises the ability of the wild type receptors to stabilize the active state required for G protein activation. Presently, we cannot distinguish between these possibilities because we lack a reagent that specifically prevents AT 1 receptor dimerization.
Six observations by others suggest that two 7TM receptors are necessary or are advantageous for G protein coupling. 1) Applying chemical cross-linking of purified leukotriene B 4 receptor BLT1 and G␣ i2 , ␤ 1 , and ␥ 2 proteins in a reconstituted system, Baneres and Parello (46) used a combination of mass spectrometry and neutron scattering to established that only one G protein trimer binds to the dimeric complex of this receptor.
2) The size of the cytoplasmic surface of a single 7TM receptor is most likely too small to anchor both the ␣and the ␤␥-subunits. Recently, Palczewski and co-workers (47-50) used their atomic force microscopy data (obtained from native retina membranes in concert with the crystal structures of rhodopsin and the G protein) to propose a model where the monomers in a rhodopsin dimer complex cooperate to dock a single G␣ t and, furthermore, where a neighboring rhodopsin dimer might bridge the G␤␥ interaction. 3) In the heterodimeric GABA B complex, GABA B2 catalyzes G protein activation, whereas the FIG. 4. Mutant AT 1 receptors do not inhibit wild type-induced bovine ␤-arrestin 2 recruitment. A, to test whether the mutant receptors interfere with the wild type receptor ␤-arrestin 2 recruitment, we co-expressed the AT 1 -luc-tagged receptor with EGFP-tagged ␤-arrestin 2 either alone or together with untagged mutant receptors. The cells were treated with either basal vehicle or with an Ang II saturating dose (100 nM) as described under "Experimental Procedures." Data are depicted as fold responses and represent the average (ϮS.D.) from three experiments performed in duplicate. B, to see whether the AT 1 -WT receptor could induce a "functional rescue" of the ␤-arrestin 2 recruitment by the AT 1 -K199A-Rluc, we co-expressed AT 1 -K199A-luc and EGFP-tagged ␤-arrestin 2 either alone or together with untagged AT 1 -WT receptor, and we analyzed ␤-arrestin 2 recruitment as aforementioned. Furthermore, to analyze whether the AT 1 -WT receptor would affect the ability of other receptors to interact ␤-arrestin 2, luciferase-tagged versions of the Eta receptor, CaR, and EGF TM receptor were also co-expressed with EGFP-tagged ␤-arrestin 2 and the untagged AT 1 -WT receptor. These expression combinations were treated with basal vehicle or with a saturating Ang II dose as described under "Experimental Procedures." Data are depicted as fold response and represent the average (ϮS.D.) from four to six experiments performed in duplicate and compared by using a Student's t test (*, p Ͻ 0.05). GABA B1 subunit binds GABA and improves GABA B2 G protein coupling but cannot activate the G protein itself, suggesting optimal activation requires both receptors (4,51). 4) A double cysteine M3 muscarinic receptor that has lost the ability to form covalent dimers and multimers is almost unable to activate the G protein, although the multimer still binds ligand but with reduced affinity (14). 5) Co-transfection of the wild type platelet-activation factor receptor and a mutant receptor (D63N) that does not couple to G protein when expressed alone result in a receptor complex with high constitutive activity. When wild type and mutant receptors were expressed at a ratio of 1/1, the agonist-induced response was greater than that of the wild type expressed alone. When the ratio was increased to 1/3, the agonist could no longer induce a response, possibly because the receptors were already desensitized because of the high constitutive activity (37). 6) It has been shown recently that point mutations that abolish dimerization of the ␣-factor (15), CCR2, and CCR5 (16) receptors also decrease or abolish agonist-induced signaling, although the receptors maintained their high affinity ligand binding.
There are three ways to explain the ERK data. First, dimerization with wild type receptors enables the mutant receptors to activate ERK and recruit ␤-arrestin. Second, because the mutants did not prevent ERKs and ␤-arrestin activations, these processes may only need one functional receptor in a complex to activate these molecules, or in fact a monomeric receptor may be capable of this. Third, it seems likely that the mutant receptors affect specific activation states differently, in which case receptor complex might still activate ERK and ␤-arrestin but not G␣ q . It is possible that AT 1 receptors activate ERK more robustly than G proteins. Thus, in line with our observations, various AT 1 receptor ligands that activate ERKs but not G protein have been identified, whereas no ligand solely activating G protein has been described (24,25). In addition, mutant AT 1 receptors that induce ␤-arrestin recruitment and ERK phosphorylation or internalization, but not G protein activation, have been described (21,23).
Along with our observation that AT 1 -WT receptor can rescue ␤-arrestin 2 recruitment by AT 1 -K199A (Fig. 4B), several other studies suggest that a dimeric receptor complex might be needed for interaction with ␤-arrestin and ERKs. Accordingly, ␤ 2 -adrenergic receptors heterodimerize with AT 1 receptors, and receptor antagonists against either of these receptors abolish both G protein and ERK activation from both receptors (19). Moreover, endocytosis of receptors that do not internalize alone has been functionally rescued by either WT or different receptors in homo-or heterodimeric complexes (12,52) as well as a mutant M3 receptor that has been reported to exert dominant negative effects specifically on WT M3 receptor ERK activation (53). Finally, the crystal structure of ␤-arrestin may contain two binding sites for 7TM receptors (54,55).
Conclusions-We have shown that AT 1 receptors specifically and ligand independently dimerize during biosynthesis before surface expression. Furthermore, AT 1 receptor mutants exerted a dominant negative effect on G␣ q activation, whereas ERK activation was preserved, suggesting that separate activation conformations of AT 1 receptors can be coupled to each of these signaling systems. These data imply oligomerization plays an active role in supporting distinct active conformations of AT 1 receptors.