Modifying Ligand-Induced and Constitutive Signaling of the Human 5-HT4 Receptor

G protein–coupled receptors (GPCRs) signal through a limited number of G-protein pathways and play crucial roles in many biological processes. Studies of their in vivo functions have been hampered by the molecular and functional diversity of GPCRs and the paucity of ligands with specific signaling effects. To better compare the effects of activating different G-protein signaling pathways through ligand-induced or constitutive signaling, we developed a new series of RASSLs (receptors activated solely by synthetic ligands) that activate different G-protein signaling pathways. These RASSLs are based on the human 5-HT4b receptor, a GPCR with high constitutive Gs signaling and strong ligand-induced G-protein activation of the Gs and Gs/q pathways. The first receptor in this series, 5-HT4-D100A or Rs1 (RASSL serotonin 1), is not activated by its endogenous agonist, serotonin, but is selectively activated by the small synthetic molecules GR113808, GR125487, and RO110-0235. All agonists potently induced Gs signaling, but only a few (e.g., zacopride) also induced signaling via the Gq pathway. Zacopride-induced Gq signaling was enhanced by replacing the C-terminus of Rs1 with the C-terminus of the human 5-HT2C receptor. Additional point mutations (D66A and D66N) blocked constitutive Gs signaling and lowered ligand-induced Gq signaling. Replacing the third intracellular loop of Rs1 with that of human 5-HT1A conferred ligand-mediated Gi signaling. This Gi-coupled RASSL, Rs1.3, exhibited no measurable signaling to the Gs or Gq pathway. These findings show that the signaling repertoire of Rs1 can be expanded and controlled by receptor engineering and drug selection.


INTRODUCTION
Heptahelical G protein-coupled receptors (GPCRs) are the largest family of human cell-surface receptors, encompassing more than 340 hormone receptors and 350-460 olfactory receptors [1,2]. They are activated by peptide hormones, odorants, photons, biogenic amines, phospholipids, and many other extracellular signals. Upon activation, GPCRs undergo conformational changes that allow active and reversible signaling through a limited number of G-protein pathways (G s , G i , G q , G 12/13 ). These signals mediate a wide variety of physiological responses, including heart rate, chemotaxis, cell proliferation, neurotransmission, and hormonal responses. Owing to their physiological importance, GPCRs are of great medical interest. Indeed, they are targets for at least 40% of modern pharmaceuticals [3].
Although many drugs target GPCRs, studies of GPCR signaling in vivo have been hampered by the lack of specific agonists and antagonists for many of the receptors. GPCRs display molecular and functional diversity, such as the type of G-protein signaling pathway associations, different levels of constitutive activity, and different signaling responses due to different ligand-selective conformations [4,5]. This diversity enables the receptors to transmit unique extracellular signals but hampers efforts to sort out the relative contributions of each signaling pathway or the roles of constitutive signaling for each receptor.
To better study the diversity of GPCRs, we developed receptors activated solely by synthetic ligands (RASSLs) by modifying their structures to render them unresponsive to endogenous hormones. Instead, RASSLs are activated by smallmolecule drugs [6], allowing them to be used to activate specific G-protein pathways rapidly and reversibly and to mimic the speed, localization, regulation, and amplification of endogenous GPCR signals [7].
Since it is impractical to convert all GPCRs into RASSLs, we and others have focused on representative G s -, G i -, and G qcoupled GPCRs, which stimulate adenylyl cyclase, inhibit adenylyl cyclase, and stimulate phospholipase-C, respectively. Ro1 (RASSL opioid 1), the prototype RASSL based on a G i -coupled k-opioid receptor [8], provided a proof of concept for this strategy. Its G i response to natural ligands is 0.001% of that of the wildtype receptor, but it is potently activated by the synthetic agonist spiradoline [8][9][10]. Ro1 decreases heart rate in mice [11] and affects taste sensation in the tongue [9]. In addition to serving as powerful tools to dissect the G-protein signaling in vivo, RASSLs can yield insights into fundamental aspects of receptor diversity [12]. For instance, constitutive signaling of Ro1 led to cardiomy-opathy [11], diminished bone formation [13], and induced hydrocephalus [10]. These constitutive signaling phenotypes would have been difficult or impossible to identify by studying endogenous receptors.
Multiple RASSLs have since been made, including a G scoupled RASSL based on the melanocortin-4 receptor [14], a G qcoupled RASSL based on the histamine 1 receptor [15], and a series of RASSLS based on muscarinic receptors [16]. These RASSLs are useful tools; however, it is still advantageous to derive a series of RASSLs with distinct G-protein signaling from the same parental GPCR. It can be difficult to compare the effects of RASSLs based on different parental GPCRs since these RASSLs could have different pharmacokinetics, constitutive activity, desensitization kinetics, and cellular localization.
To better study GPCRs, we built a series of RASSLs based on the human 5-HT 4 receptor (Figure 1), which has several advantages over other serotonin receptors. First, its pharmacological properties are well established [17]. Second, its agonists have milder effects (increased gastrokinesis [18], augmented memory acquisition and retention [19], increased chronotropic and inotropic cardiostimulation [20], and enhanced cortisol release [21]) than other serotonergic drugs. Third, it has a large number of synthetic ligands, which allows us to identify differences in their effects on that receptor. Fourth, a single mutation (D 100 A) in the mouse 5-HT 4 receptor dramatically reduced its affinity for serotonin, its endogenous ligand. This mutation also allows synthetic agonists and antagonists for the wildtype receptor to activate the mutant, turning 5-HT 4 -D 100 A into a RASSL [22].
Finally, we reasoned that novel receptors coupling to other signaling pathways could be created by making chimeras of the 5-HT 4 receptor with other family members. Altering the Gprotein selectivity of GPCRs is often difficult because it is based on receptor conformation determined by multiple regions of the receptor [23]. Changing multiple regions involves large internal mutations that often lead to receptor instability. A better strategy for altering G-protein signaling characteristics is to swap domains between structurally similar receptors within the same family. The 5-HT 4 receptor belongs to a family of at least 15 receptors, each with different subfamilies that engage different G-protein signaling pathways. The 5-HT 4 , 5-HT 6 , and 5-HT 7 subfamilies are G s coupled, the 5-HT 1 subfamily is G i coupled, and the 5-HT 2 subfamily is G q coupled [24]. These characteristics could expedite our efforts to make purely G s -, G i -, and G q -coupled RASSLs. Here, we describe a new series of RASSLs developed to modify the ligand-induced and constitutive signaling of the human 5-HT 4 receptor. These modified GPCRs will help us better study the effect of constitutive G s signaling and ligand-induced G s , G s /G q , and G i signaling in vivo.

Human 5-HT 4 D 100 A is a G s -coupled RASSL
The D 100 A mutation in the mouse 5-HT 4 receptor converts it into a RASSL [22], but its effects on the human 5-HT 4 receptor have not been tested. We now extend these findings to the human 5-HT 4 receptor ( Figure 1). To determine if antagonists for the wildtype 5-HT 4 receptor also activate the human 5-HT 4 -D 100 A mutant, we tested a variety of compounds. The mutant receptor was not activated by serotonin (Figures 2, 3A), but it was activated by agonists (cisapride, zacopride), partial agonists (RS23597, RS39604, RS67333), antagonists (GR113808, RO110-0235), and an inverse agonist (GR125487) for the wildtype 5-HT 4 receptor ( Figure 2). Interestingly, GR125487 showed a specific response, as demonstrated by the steep concentration-response curve ( Figure 3D). In addition, GR113808, GR125487, and RO110-0235 potently activated G s signaling of 5-HT 4 -D 100 A without stimulating the wildtype receptor ( Figures 3B-E). The mutant receptor was selectively activated by multiple synthetic ligands (GR113808, GR125487, and RO110-0235) but not serotonin. We named it Rs1 (RASSL serotonin 1) (Figure 1).

Rs1 has a high level of constitutive signaling
We next examined the constitutive signaling of Rs1 in more detail. Rs1 showed greater constitutive signaling than the wildtype receptor at all levels of transfection ( Figure 4A). Constitutive activity was observable when only 25 ng of receptor cDNA per 5610 6 HEK293 cells was transfected ( Figure 2). The highest level of constitutive activity, achieved with 5.4 mg of receptor cDNA per 5610 6 HEK293 cells, was 1.5 times greater than that of the wildtype 5-HT 4 receptor (49.661.25 nM vs. 32.564.04 nM, p,0.005) and .10-fold higher than that of the control receptors (the b 2 -adrenergic and parathyroid hormone receptors), which have low levels of constitutive signaling ( Figure 4A). Despite the high level of constitutive signaling, both the 5-HT 4 receptor and Rs1 could still be further activated by zacopride ( Figure 4B).

Ligand-specific G q signaling in Rs1
Before constructing Rs1-5HT 2C chimeras to make a G q -signaling RASSL, we assayed inositol phosphate 1 (IP1) accumulation by Rs1 via constitutive or ligand-induced signaling. Rs1 showed no measurable difference in constitutive G q signaling ( Figure 5) as compared to mock-transfected cells in the IP1 and calcium mobilization assays. Upon activation by cisapride, zacopride, RS23597, RS39604, or RS67333, Rs1 showed 2-3.5-fold higher G q signaling than the wildtype 5-HT 4 receptor (p,0.005) ( Figure 5). Surprisingly, GR113808, GR125487, and RO110-0235 activated predominately G s (Figure 2) and little G q signaling ( Figure 5 Figure 2. Rs1 is a G s -coupled RASSL. Rs1 was efficiently activated by small compounds known to be full agonists (cisapride, zacopride), partial agonists (RS39604, RS67333, and RS23597), antagonists (GR113808, RO110-0235), or inverse agonists (GR125487) for the wildtype 5-HT 4 receptor. It was not activated by its endogenous agonist (serotonin   activate the G s signaling of Rs1 without activating the G s signaling of the wildtype 5-HT 4 receptor ( Figure 2). Therefore, we could use drugs with distinct chemical structures ( Figure S1) to activate G s or G s /G q signaling of Rs1. The once controversial use of conformation-specific ligands to alter G-protein coupling and other receptor functions has now been demonstrated in several other GPCRs [25][26][27][28][29]. This is the first time that agonist-dependent functional selectivity has been shown in a RASSL.
Replacing the second (i2) or third intracellular loop (i3) of Rs1 eliminated both G s and G q signaling ( Figure S3). Only the carboxyl chimera (Rs1-C-5-HT 2C ) showed enhanced G q signaling in response to cisapride, zacopride, RS23597, RS39604, and RS67333 ( Figure 5), showing that the ligand-induced specificity of signaling was preserved. Constitutive and ligand-induced G s signaling were also largely preserved ( Figures S3A, S3B).
These data indicated that i2 and i3 are both necessary for G scoupling of the 5-HT 4 receptor. They also suggested that the Cterminus of 5-HT 2C contains a G-protein coupling domain for G q signaling or that the C-terminus of 5-HT 4 receptor promotes G q signaling. However, we were unable to completely alter the Gprotein preference of this receptor for G q signaling. We amplified the G q signaling of Rs1 by domain swapping the C-terminus with 5-HT 2C . None of the other 12 chimeras showed enhanced G q signaling, even when multiple internal segments were combined. We did not proceed further with these experiments because substitutions of multiple internal domains also decreased cellsurface expression of the receptors (data not shown). Since G q signaling of Rs1-C-5-HT 2C was activated by cisapride, zacopride, and RS23597 but not by serotonin, we named it Rs1.1 (Table 1).

A purely G s signaling RASSL with low levels of constitutive signaling
Since the high constitutive activity of the 5-HT 4 -D 100 A mutant causes significant phenotypes in transgenic mice [30] and could not be controlled by inverse agonists [31], we attempted to lower the Rs1 constitutive activity by making additional point mutations. We focused on the D 66 N and W 272 A mutations, which reduce constitutive signaling of the mouse 5-HT 4 receptor [31,32]. Rs1-D 66 A, Rs1-D 66 N, and Rs1-W 272 A significantly reduced constitutive signaling ( Figure 6A). The cell-surface expression of Rs1-D 66 A and Rs1-D 66 N was similar to that of Rs1 ( Figure 6B), so the reduction in constitutive signaling was probably not linked to lower cell-surface expression. Surprisingly, the D 66 A and D 66 N mutations also abolished zacopride-induced G q signaling ( Figure 6C). Thus, we created two RASSLs exhibiting pure G s signaling and low constitutive signaling. Unfortunately, the efficacy of the ligandinduced G s response was also significantly compromised, diminishing the utility of these receptors.

Engineering Rs1 for G i Signaling
To engineer a G i -signaling RASSL based on Rs1, we replaced its intracellular loops with those of 5-HT 1A , a G i signaling receptor [33]. Of four Rs1-5HT 1A chimeras (Figures 7, S4, S18, S19, S20, S21), only the two containing i2 and i3 from Rs1 were expressed at a level similar to that of Rs1 ( Figure S5D). Replacing those loops abolished constitutive and ligand-induced G s signaling at both low and high levels of receptor cDNA (25 ng and 4.8 mg per 5610 6 cells) ( Figures  S5A, S5B). Interestingly, this RASSL showed no evidence of constitutive signaling via the G i or G s pathway. These findings  strongly imply that both i2 and i3 are required for G s signaling of Rs1. In addition, activation of the Rs1-i3-5-HT 1A chimera with zacopride significantly inhibited cAMP accumulation induced by 10 mM apomorphine (agonist for dopamine 1 receptors) in HEK293 cells co-transfected with 1.5 mg of Rs1 receptor and 0.5 mg of dopamine 1 receptor (per 5610 6 cells; Figure S5C and Figure S8A). This inhibition was smaller than that of m-opioid receptor stimulated by [D-Ala2, D-Leu5]-enkephalin (DADLE). Both responses were abolished by 50 nM pertussis toxin, indicating the involvement of G i signaling ( Figure 7B). Unfortunately, the potency (amount of drug needed to reach an effect) of ligand-induced G i signaling was significantly reduced ( Figure 7C). While these results are encouraging, future experiments are needed to determine if low potency may be due to nonspecific effects, or will reproduce in other cell types. Since Rs1-i3-5HT 1A exhibited G i but not G s signaling, we named it Rs1.3 ( Figure 1, Table 1).

DISCUSSION A new series of RASSLs
We report here a new series of RASSLs to study multiple Gprotein signaling pathways. Many GPCRs activate multiple Gprotein signaling pathways and exhibit a wide range of constitutive signaling activities. Our new RASSLs will help us better study the effect of stimulating canonical signaling pathways (G s /G q , G s , G i ), with different constitutive signaling using a single receptor backbone and different synthetic agonists. These new RASSLs will also allow a systematic examination of the different functional domains of the 5-HT 4B receptor.
These new RASSLs could help us better dissect the physiological significance of constitutive signaling in vivo. Constitutive *** , and the Rs1-C-5-HT 2C chimera exhibit different G q signaling properties. G q signaling was analyzed by measuring the accumulation of inositol 1 phosphate (IP1). The 5-HT 4 receptor showed increased G q signaling when activated by serotonin, cisapride, and zacopride but not by RS39604 or RS67333, relative to mock treatment with HBSS. Rs1 showed significantly higher G q signaling than the wildtype 5-HT 4 receptor when activated by cisapride, zacopride, RS39604, and RS67333 but not by serotonin. G q signaling of Rs1-C-5HT 2C chimera was activated by cisapride, zacopride, RS23597, RS39604, and RS67333 and was minimally activated by RO110-0235. GR113808 and GR125487 did not activate G q signaling by any of the three receptors. ***p,0. signaling is crucial for many physiological processes and many diseases. Up to 40% of all GPCRs [34], including the 5-HT 4 receptor [35], show significant constitutive activity, and Rs1 could be a good model for these receptors. Like the mouse 5-HT 4 -D 100 A receptor [22], Rs1 had a higher level of constitutive signaling than the wildtype 5-HT 4 receptor. Using the tetracycline transactivator (Tet) system, we have already made an Rs1 transgenic mouse in which Rs1 expression is driven by the osteoblast-specific Col1a-1 2.3-kb promoter fragment. These mice exhibited dramatically increased bone formation [30]. These and other findings strongly suggest that constitutive signaling can drive potent phenotypic changes in vivo. We found only constitutive G s signaling in Rs1 but did not observe any constitutive G q signaling. It would be interesting to generate new or modify existing RASSLs with increased G i or G q constitutive signaling in the future to examine the effect of both agonist-mediated and constitutive signaling in various physiological processes. Since signaling would only be dependent on expression rather than on circulating hormones, these RASSLs have an intrinsic advantage for studying constitutive signaling when combined with a Tet system. In addition, the 5-HT 4 RASSLs have a large set of agonists with different activities that could be useful in future studies. Many of the Rs1 agonists used in this study activated Rs1 with an EC 50 in the nanomolar range, allowing to us activate Rs1 effectively. Having an abundant selection of ligands for Rs1 helped us find strong evidence of ligand-induced functional selectivity on Rs1. Having a greater variety of drugs to choose from could be valuable for in vivo studies. For instance, we could use GR125487 to activate Rs1 while suppressing the basal activity of wildtype 5-HT 4 . Alternatively, we could potentially study Rs1 constitutive signaling by using drugs such as RO116-0086 or RO116-1148 [31] to decrease constitutive signaling of the wildtype 5-HT 4 receptor but not of Rs1. Finally, this new series of RASSLs may make it possible to perform in vivo studies in which Rs1 is activated with minimal side effects. Since knockout of the 5-HT 4 receptor does not cause overt side effects [36], treatment with GR113808, GR125487, and R0110-0235 (antagonists and inverse agonists for the wildtype receptor) may have minimal side effects as well. This eliminates the need to knock out the endogenous 5-HT 4 receptors for most in vivo experiments [10]. Since the same agonists can be used to activate all of the RASSLs within this series and therefore+ engage different G-protein signaling pathways (Table 1), we can more easily compare the effect of activating the G s, G s/q , or G i pathway.
While these new RASSLs may be useful for studying G protein signaling in vivo, they have several limitations. Despite significant efforts to screen all available antagonists, we could not identify inverse agonists that would lower constitutive signaling by Rs1. Fortunately, we can use conditional expression systems such as the tetracycline transactivator system to control Rs1 expression and constitutive activation. Although we were successful in finding mutations that reduced Rs1 constitutive signaling these same mutations adversely affected the agonist mediated signaling (Rs1.2), suggesting that the Other mutations could be explored that may provide a more optimal reduction in G s basal activity without affecting ligand-induced receptor activation. In addition, the relatively small G i signaling of Rs1.3 can only be activated by a relatively high concentration of zacopride, thus limiting its usefulness in vivo.

Insights into G-protein signaling by Rs1
Our study yielded insights into the G-protein selectivity and functional selectivity (differential effects of ligands on the same receptor) of Rs1. Of the 12 Rs1-5-HT 2C chimeras that are expressed on the cell surface, none of the i2 or i3 chimeras showed any G s or G q signaling. Evidently, these intracellular loops of Rs1 are crucial for signaling via those pathways. The importance of i2 and i3 for G s signaling of Rs1 is further supported by the lack of G s signaling by the Rs1-i2-5-HT 1A and Rs1-i3-5-HT 1A chimeras. This is the first study showing the importance of i2 and i3 in both G s and G q signaling of the human 5-HT 4 -D 100 A receptor.
We also found that i3 domain swapping abolished all G s signaling and enabled Rs1 to stimulate G i signaling of 5-HT 1A . The role of i2 and i3 in the G i signaling of 5-HT 1A receptor has been extensively reported. The entire N-terminus [37] and Cterminus of i2 [38] of 5-HT 1A are thought to be sufficient to support G-protein coupling, but not signaling. On the other hand, the N-terminus [39] and C-terminus of i3 of 5-HT 1A [40,41] seem to be essential for the G i signaling of 5-HT 1A . In fact, replacing the N-terminus of the i3 of the a 2 -adrenergic receptor with that of 5-HT 1A resulted in a chimera that signals like a 5-HT 1A receptor when stimulated by a a 2 -adrenergic receptor agonist [42]. Since Rs1-5HT 1A chimeras with multiple internal domains replaced are not significantly expressed on the cell surface (data not shown), it may be difficult to further improve the potency of the Rs1.3 using our current approach. We hypothesize that replacing the N-and C-terminal portions of i2 and i3 instead of the whole i2 and i3 loops may increase the potency of Rs1.
We also showed that various drugs can differentially activate Gprotein signaling of Rs1. Functional selectivity has been reported for many receptors. It led to divergent fates of internalization for the dopamine D 1 receptor [25], various binding specificities for gonadotropin-releasing hormone receptors, and different levels of activation of G proteins for the b 2 -adrenergic [26], mu-opioid [27], dopamine D 2 [28], and human 5-HT 2A [29] receptors.
The indoleamine derivatives GR113808, GR125487, and RO110-0235 did not fully activate G q signaling of Rs1, Rs1.1, or the 5-HT 4 receptor. On the other hand, the benzamide derivatives cisapride, zacopride, RS23597, RS39604, and RS67333 activated the G q signaling of Rs1. These findings may reflect distinct conformational changes caused by indoleamine and benzamide derivatives.
The possibility of functional selectivity is further supported by the results obtained with Rs1-C-5HT 2c and Rs1 point mutants (Rs1-D 66 A and Rs1-D 66 N). The D 100 A mutation and replacement of the C-terminus amplified G q signaling by the 5-HT 4 receptor. The addition of D 66 A and D 66 N abolished G q signaling. Since D 100 A is located in the binding pocket of the 5-HT 4 receptor, this mutation in Rs1 may have changed the configuration of the binding pocket, making the receptor more susceptible to G q activation by zacopride and RS23597. This response was even more pronounced when the D 100 A mutation was combined with domain swapping of the C-terminus with that of 5-HT 2C . Thus, it is reasonable to hypothesize that these changes modified the ligand-selective receptor conformation [5], changing the receptor susceptibility to functional selectivity.

Conclusions
Our studies with Rs1 provide a proof-of-concept for making a series of RASSLs with different signaling properties. Recently, Armbruster et al. made a series of RASSLs based on the muscarinic M3 and M4 receptors, which have low constitutive activity. These RASSLs each couple different G-protein signaling pathways and can be activated by clozapine-N-oxide, an inert ligand with high bioavailability [16]. These RASSLs nicely complement our Rs1 RASSLs with varying constitutive activity. In addition, we predict that some RASSLs with the same canonical G-protein signaling (G s , G i , or G q ) will have different in vivo phenotypes due to noncanonical signaling. This growing collection of RASSLs will greatly facilitate our efforts to understand the physiological significance of the inherent signaling diversity of GPCRs.
An ideal series of RASSLs would have receptors with different combinations of low and high basal signaling, with robust ligandinduced effects for each major pathway, and potent inverse agonists. Although we have not achieved this goal with the Rs1 series, we are hopeful that it can be achieved with other receptors in the future. Indeed, RASSLs based on the muscarinic receptors [16] show great promise, as there are naturally occurring, or published mutants of the muscarinic reports that activate each of the major G protein signaling pathways.

HEK293 maintenance and electroporation
Early-passage (#20) HEK293 cells were maintained in highglucose DMEM (Invitrogen, Carlsbad, CA) supplemented with sodium pyruvate (Invitrogen) and 10% Fetalplex (Gemini Bio-Products, West Sacramento, CA). Receptors were electroporated into HEK293 cells as described [44]. The electroporated cells were reconstituted into a suspension using DMEM with 10% heatinactivated, dialyzed fetal bovine serum (Thermo-Fisher Scientific, Logan, UT). The transfection efficiency was monitored by flow cytometry, and the cell-surface expression of the receptor was determined by FLAG ELISA (enzyme-linked immunosorbent assay) the next day.

Measuring cell-surface expression by FLAG ELISA
Cell-surface receptor expression was measured with a FLAG ELISA as described [45]. Cells seeded in poly-D-lysine-coated 96well plates were fixed with 100 ml of 4% paraformaldehyde for 10 min at room temperature, washed, and stained with 100 ml of staining buffer (DMEM, 10% FBS, and 1 mM CaCl 2 ) containing anti-FLAG M1 antibody (1:1000; Sigma-Aldrich) for 1 h at 25uC. The samples were washed three times with wash buffer (PBS and 1 mM CaCl 2 ) and stained with 100 ml of staining buffer with rat anti-mouse IgG antibody conjugated with horseradish peroxidase (1:1000; Bio-Rad Laboratories, Hercules, CA). After 30 min, the samples were washed with wash buffer, placed on a rocker for 10 min, and washed again. This process was repeated two more times. Then, 2,2-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) liquid substrate (200 ml; Sigma-Aldrich) was added to the samples. After rocking for 30 min, 200 ml of the substrate was transferred to new 96-well plates, and optical density was measured with Victor 3 (PerkinElmer, Waltham, MA) at 405 nm. All samples contain three replicates, and all experiments were repeated at least three times.

Measurement of cAMP production in intact cells
To improve assay consistency and minimize pipetting error in the 384-well plates, we modified the high-range HTRF assay (CisBio International, Bagnols-sur-Cèze, France) by seeding, stimulating, and lysing the cells in 96-well plates and using the lysate instead of live cells to determine cAMP production. The remainder of the analysis was performed according to the manufacturer's instructions G i Assay G i signaling was examined in cells transfected with 1.5 mg of receptor cDNA, 0.6 mg of human dopamine 1 receptor cDNA, and pcDNA3 (up to 6 mg). The co-transfectants were stimulated first with 100 ml of KRBG buffer containing IBMX for 10 min at room temperature and then with 50 ml of PBS containing 10 mM apomorphine (agonist for the dopamine 1 receptor) and 10 mM zacopride for 10 min at 37uC. The cells were lysed in 50 ml of lysis buffer, and 5 ml of lysate was used in the HiRange HTRF assay.
Fluorometric imaging plate reader assay to measure calcium mobilization To measure calcium mobilization, 4.8 mg of receptor cDNA, 0.6 mg of DsRed plasmid, and 0.6 mg of human bombesin receptor cDNA were electroporated into 5610 6 HEK293 cells as above [44]. Hank's balanced salt solution (10 ml) with 20 mM HEPES, 0.25 mM probenecid acid (Sigma-Aldrich), and 2% pluronic acid (Sigma-Aldrich) was added to each bottle of Calcium 4 (Molecular Devices, Sunnyvale, CA), and 100 ml of the resulting solution was added to each well for 1 h at 37uC before measurement. Assays were performed with a FLEX Station (Molecular Devices), with excitation of 485 nm, emission of 525 nm, and cut-off of 515 nm, as recommended by the manufacturer.

Determination of IP1 production in intact cells
A modified version of the IP1 protocol was used (CisBio International). HEK293 cells were washed once with calcium-free PBS and dissociated from flasks with cell dissociation buffer (Invitrogen). Cells (5610 6 ) were electroporated as described above. Then, 10 5 cells were placed in DMEM supplemented with 10% decomplemented, dialyzed against FBS, and seeded onto 96-well plates coated with poly-D-lysine. The next day, the cells were stimulated with agonists in 50 ml of 16 stimulation buffer for 30 min at 37uC and lysed for 10 min with 9 ml of lysis/detection buffer. Then, 14 ml of lysate was added to 384-well plates and subjected to High-Range HTRF assay as described above, except that 3 ml of cAMP-d2 and anti-cAMP-cryptate solution were added to each well.

Data analysis
cAMP and IP1 values were analyzed with GraphPad Prism 4 (GraphPad Software, San Diego, CA). Calcium mobilization results were analyzed with SoftMax Pro v5 (Molecular Devices). Statistical significance was determined with paired Student's t tests.