Random mutagenesis of the M3 muscarinic acetylcholine receptor expressed in yeast: identification of second-site mutations that restore function to a coupling-deficient mutant M3 receptor.

The M(3) muscarinic receptor is a prototypical member of the class A family of G protein-coupled receptors (GPCRs). To gain insight into the structural mechanisms governing agonist-mediated M(3) receptor activation, we recently developed a genetically modified yeast strain (Saccharomyces cerevisiae) which allows the efficient screening of large libraries of mutant M(3) receptors to identify mutant receptors with altered/novel functional properties. Class A GPCRs contain a highly conserved Asp residue located in transmembrane domain II (TM II; corresponding to Asp-113 in the rat M(3) muscarinic receptor) which is of fundamental importance for receptor activation. As observed previously with other GPCRs analyzed in mammalian expression systems, the D113N point mutation abolished agonist-induced receptor/G protein coupling in yeast. We then subjected the D113N mutant M(3) receptor to PCR-based random mutagenesis followed by a yeast genetic screen to recover point mutations that can restore G protein coupling to the D113N mutant receptor. A large scale screening effort led to the identification of three such second-site suppressor mutations, R165W, R165M, and Y250D. When expressed in the wild-type receptor background, these three point mutations did not lead to an increase in basal activity and reduced the efficiency of receptor/G protein coupling. Similar results were obtained when the various mutant receptors were expressed and analyzed in transfected mammalian cells (COS-7 cells). Interestingly, like Asp-113, Arg-165 and Tyr-250, which are located at the cytoplasmic ends of TM III and TM V, respectively, are also highly conserved among class A GPCRs. Our data suggest a conformational link between the highly conserved Asp-113, Arg-165, and Tyr-250 residues which is critical for receptor activation.

The superfamily of G protein-coupled receptors (GPCRs) 1 represents the largest group of cell surface receptors found in nature (1)(2)(3). All GPCRs contain a bundle of seven transmembrane (TM) helices (TM I-VII) that are connected by alternating intracellular and extracellular loops (4 -7; Fig. 1). Based on sequence similarity, mammalian GPCRs can be grouped into three major receptor subfamilies (A, B, and C). Family A contains by far the largest number of receptors including, for example, the receptors for light (rhodopsin), a very large number of odorant receptors, and the classical biogenic amine neurotransmitter receptors including the five muscarinic acetylcholine receptors (M 1 -M 5 ) (4 -7).
The hallmark of class A GPCRs is a set of about 20 amino acids that is highly conserved only within this GPCR subfamily (5,7,8). The majority of these amino acids are located within the cytoplasmic half of the TM receptor core. Mutagenesis data suggest that these highly conserved residues are required for protein stability and/or for mediating the conformational changes that accompany receptor activation (5,7,8).
At present, bovine rhodopsin, in its inactive state, is the only GPCR for which high resolution structural information is currently available (9 -11). Attempts to obtain a high resolution structure of a GPCR in its active form have not been successful so far. In the absence of this information, different mutagenesis techniques have been employed, often combined with molecular modeling approaches, to gain insight into the functional roles of specific receptor domains/amino acids in GPCR function (5,7,12).
During the past decade, we have used the M 3 muscarinic receptor as a model system to study the molecular mechanisms underlying the function of class A GPCRs activated by small diffusible ligands. The M 3 muscarinic receptor preferentially activates G proteins of the G q/11 family which mediate the activation of phospholipase C␤ (13).
To facilitate structure-function studies of the M 3 muscarinic receptor, we recently developed a heterologous expression system that allows the functional expression of the M 3 muscarinic receptor in yeast (Saccharomyces cerevisiae) (14,15). Specifically, we expressed the M 3 muscarinic receptor in a genetically modified yeast strain that requires agonist-dependent receptor/G protein coupling for cell growth. This strain, referred to as MPY578q5, harbors a mutant version of the GPA1 gene coding for a hybrid yeast/mammalian G protein ␣ subunit in which the last five amino acids of Gpa1p were replaced with the corresponding mammalian G␣ q residues (14). We demonstrated previously that the M 3 muscarinic receptor can activate this hybrid G protein with high efficiency and selectivity (14).
One major advantage of the yeast expression system is that powerful genetic screens can be employed to isolate mutant receptors endowed with novel phenotypes from large receptor libraries generated by random mutagenesis (15)(16)(17)(18)(19)(20)(21)(22). Another advantage is that the results obtained by the use of this strategy (receptor random mutagenesis followed by yeast genetic screens) do not rely on preconceived notions of GPCR function. Using this approach, we recently identified a point mutation (Q490L) in the M 3 muscarinic receptor which leads to robust agonist-independent receptor signaling in both yeast and mammalian cells (15). We then applied a secondary yeast genetic screen to isolate second-site mutations that were able to suppress the activating effects of the Q490L mutation. This screen led to the identification of 12 amino acids predicted to play key roles in M 3 receptor activation and/or receptor/G protein coupling (15).
Previous studies with different classes of GPCR have shown that the identification of second-site suppressor mutations can provide important new insights into the structural and functional roles of specific amino acids (20,(23)(24)(25)(26)(27). The present study was designed to learn more about the functional role of a TM II Asp residue that is highly conserved among class A GPCRs (Asp 2.50 according to the Ballesteros/Weinstein amino acid numbering system (28); corresponding to Asp-113 2.50 in the rat M 3 muscarinic receptor; Fig. 1). Studies with many different class A GPCRs have shown that replacement of this residue by other amino acids, such as Gln, abolishes or drastically reduces receptor-mediated G protein activation (29 -44). Consistent with these previous reports, we initially demonstrated in the present study that the D113N mutant M 3 receptor was unable to interact productively with G proteins in yeast or mammalian cells .
To learn more about the functional role of the conserved Asp-113 residue in M 3 receptor function, we decided to use yeast expression technology to identify second-site suppressor mutations that are able to restore function to the D113N mutant receptor. Specifically, we subjected the Asp-113 mutant receptor to random mutagenesis and then used a yeast genetic screen to recover second-site suppressor mutations. This screen led to the identification of three point mutations, R165W 3.50 , R165M 3.50 , and Y250D 5.58 , which were able to restore function to the D113N mutant receptor when expressed in yeast. To examine whether these results could be reproduced in a mammalian expression system, we also characterized the recovered mutant M 3 receptors in transfected COS-7 cells. Moreover, we carried out additional site-directed mutagenesis studies to examine the allele specificity of the recovered second-site suppressor mutations and the ability of other amino acid substitutions at positions Arg-165 and Tyr-250 to rescue the function of the D113N receptor.
Interestingly, the three recovered second-site suppressor mutations involved the mutational modification of two amino acids which, like Asp-113, are highly conserved among class A GPCRs. It is likely that these three amino acids (Asp-113 2.50 , Arg-165 3.50 , and Tyr-250 5.58 ) participate in a network of interactions that is critical for converting the inactive state of the M 3 receptor into its active conformation. Given the conserved nature of the amino acids targeted in the present study, our results should be of broad general relevance. Construction of Plasmids-For yeast expression studies, all mutations were introduced into a modified version of the rat M 3 muscarinic receptor lacking the central portion of the i3 loop (Ala-274 to Lys-469) and containing an N-terminal, nine-amino acid hemagglutinin epitope tag (YPYDVPDYA) inserted after the initiating methionine codon (14,15). We demonstrated previously that these modifications have little effect on the ligand binding and G protein coupling properties of the M 3 receptor (14,(45)(46)(47). For the sake of simplicity, this i3 loop-deleted, epitope-tagged version of the M 3 receptor is referred to as 'WT' M 3 receptor. The coding sequences of the 'WT' M 3 receptor or 'WT' M 3 receptor-based mutant constructs were inserted into the polylinker of the yeast expression plasmid, p416GPD, as described previously (14). Site-directed mutagenesis was performed by using standard PCR mutagenesis techniques or the QuikChange site-directed mutagenesis kit (Stratagene), according to the manufacturer's instructions. The identity of all constructs was verified by DNA sequencing.

Materials-Media
Yeast Strains, Growth, and Transformation-The haploid yeast strain MPY578q5 (MATa gpa1::G␣ q 5 far1::LYS2 fus1::FUS1-HIS3 FIG. 1. Two-dimensional model of the rat M 3 muscarinic receptor highlighting functionally critical amino acids. For expression studies in yeast, all mutations were introduced into a modified version of the rat M 3 muscarinic receptor ('WT'-M 3 ) lacking the central portion of the i3 loop (Ala-274 to Lys-469) (14). Amino acid numbers refer to positions in the full-length rat M 3 muscarinic receptor sequence (70). The D113N point mutation virtually abolished receptor/G protein coupling in yeast and transfected COS-7 cells. To isolate second-site suppressor mutations that can restore function to the D113N mutant receptor, the receptor region ranging from Leu-114 to Gln-587 was subjected to PCR-based random mutagenesis followed by a yeast genetic screen. The recovered second-site suppressor mutations involved specific substitutions of Arg-165 and Tyr-250. As reported previously (55), a mutant M 3 receptor in which the positions of Ile-253 and Tyr-254 had been exchanged virtually lost the ability to couple to G proteins.
sst2::SST2-G418 r ste2::LEU2 fus2::FUS2-CAN1 ura3 lys2 ade2 his3 leu2 trp1 can1) was used as a host for the expression of the 'WT' M 3 receptor and all 'WT' M 3 receptor-based mutant constructs (14,15). The specific features of this strain have been described previously (14,15). In brief, the MPY578q5 strain contains inactive versions of the FAR1, SST2, and STE2 genes. Moreover, it harbors a mutant version of GPA1 coding for a G protein ␣ subunit in which the last five amino acids of Gpa1p were replaced with the corresponding sequence derived from mammalian ␣ q . Importantly, the genomic incorporation of a FUS1-HIS3 reporter construct makes the production of His3 protein dependent on receptor-mediated activation of the yeast pheromone pathway, allowing auxotrophic (his3) yeast strains expressing coupling-competent receptors to grow in histidine-deficient media (14,15).
Yeast cells were grown at 30°C in synthetic complete medium (SC) (48) unless noted otherwise. The lithium acetate method was used to transform yeast with plasmid DNAs coding for the different receptor constructs (49). Transformants were selected and maintained in SC medium lacking uracil (SC-Ura).
Yeast Liquid Bioassays-Yeast liquid bioassays were performed essentially as described previously (14,15). In brief, mid-log phase cell cultures (1-4 ϫ 10 7 cells/ml) were washed with phosphate-buffered saline and diluted to 10 5 cells/ml in SC medium lacking uracil and histidine (pH 7). 3-Amino-1,2,4-triazole (5 mM) was added to the medium to suppress background growth. Cell suspensions were incubated in 96-well microtiter dishes at 25°C for 72 h in the presence of increasing concentrations of carbachol (10 Ϫ13 to 10 Ϫ4 M). Receptor-mediated yeast growth was assessed by measuring increases in the optical density of the yeast cultures at 630 nm. Assays were conducted in triplicate, using three independent transformants. Carbachol concentration-response curves were analyzed using the nonlinear curve fitting program Prism 3.0 (GraphPad).

Construction of a Yeast Library Expressing Randomly Mutagenized D113N Mutant M 3 Muscarinic Receptors-
In an effort to identify secondsite suppressor mutations that can restore function to the activation-defective D113N M 3 mutant receptor, we subjected the DNA sequence coding the regions from Leu-114 to Gln-587 ( Fig. 1) to PCR-based random mutagenesis (50). PCRs were conducted in a total volume of 50 l in the presence of 10 mM Tris-HCl, pH 8. We utilized a gap-repaired protocol (51,52) to generate a library of yeast clones expressing randomly mutagenized M 3 muscarinic receptors containing the D113N point mutation. Specifically, the yeast expression plasmid, 'WT'-M 3 (D113N)-p416GPD, was "gapped" by digestion with MscI (contained within codons Val-149 to Ser-151) and NheI (contained within codons Leu-496 to Ala-498), thus removing a 455-bp restriction fragment. The resulting gapped plasmid (32 g; total size: 6,949 bp) and the mutagenized PCR fragment (7.2 g) were cotransformed into MPY578q5. After transformation into yeast, the ends of the mutagenized PCR fragments recombined with the homologous ends of the gapped receptor plasmid to regenerate circular receptor expression plasmids (51,52).
Yeast Screen to Recover Second-site Mutations That Restore Function to the D113N Mutant M 3 Muscarinic Receptor-Initially, the MPY578q5-based yeast expression library expressing mutant M 3 muscarinic receptors (see previous paragraph) was plated onto uracil-deficient SC medium to select for Uraϩ (plasmid-containing) transformants (plating density ϳ10,000 clones/150-mm plate). After incubation of plates for 2 days at 30°C, Uraϩ colonies were transferred, via replica plating, onto plates containing SC medium lacking both uracil and histidine (SC-Ura/His medium), 1 mM muscarinic agonist, carbachol, and 8 mM 3-amino-1,2,4-triazole. For control purposes, the primary Uraϩ transformants were also replica-plated onto plates containing the identical selection medium but lacking carbachol. Under these conditions, a MPY578q5-based yeast strain expressing the D113N mutant receptor was unable to grow, either in the absence or presence of carbachol. However, under the same conditions, a MPY578q5-based yeast strain expressing the 'WT' M 3 muscarinic receptor showed robust, agonist-dependent growth. Second-site mutations that restore function to the D113N mutant receptor were therefore predicted to give rise to yeast colonies displaying carbachol-dependent growth on SC-Ura/His plates. A total of ϳ200,000 yeast clones expressing mutant M 3 musca-rinic receptors were screened. Plasmids were isolated from yeast colonies that were able to grow in a carbachol-dependent fashion on SC-Ura/His medium, amplified in Escherichia coli (DH5␣), and retransformed into MPY578q5 to confirm the plasmid linkage of the observed growth phenotype. The mutant receptor plasmids were then subjected to sequencing.
Isolation of Yeast Crude Membranes-Crude yeast membrane preparations were obtained from fresh 2-liter yeast cultures, using a glass bead procedure to break up the yeast cell wall. The protocol used has been described in detail by Schmidt et al. (15). The protein concentration of yeast membrane preparations was determined by using the BCA protein assay kit.
Transient Expression of Muscarinic Receptors in COS-7 Cells-COS-7 cells were grown to 80% confluence in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin at 37°C in a humidified 5% CO 2 incubator. Approximately 24 h prior to transfection, 1 ϫ 10 6 cells were seeded into 100-mm dishes and subsequently transfected with receptor plasmid DNA (4 g/dish), using the Lipofectamine Plus kit (Invitrogen) according to the manufacturer's instructions.
Receptor-mediated Stimulation of Phosphatidylinositol (PI) Hydrolysis-COS-7 cells were transferred to 6-well plates (ϳ5 ϫ 10 5 cells/well) 24 h after transfection. After transfer of the cells, 3 Ci/ml myo-[ 3 H] inositol was added to the growth medium. Approximately 20 -24 h later, cells were washed twice with Hanks' balanced salt solution followed by a 20-min incubation at room temperature with Hanks' balanced salt solution containing 10 mM LiCl. Subsequently, cells were incubated in the same medium for 1 h at 37°C with increasing concentrations of the muscarinic agonist, carbachol (final concentrations 10 Ϫ9 to 10 Ϫ3 M). After this incubation step, the growth medium was removed, and cells were incubated for 30 min at 4°C with 20 mM formic acid. This solution was then neutralized by the addition of 60 mM ammonium hydroxide. The inositol monophosphate (IP 1 ) fraction was isolated by anion exchange chromatography as described previously (53). Carbachol concentration-response curves were analyzed by using the computer program Prism 3.0.
Preparation of Membranes from Transfected COS-7 Cells-Approximately 48 h after transfection, COS-7 cells membranes were prepared as described in detail elsewhere (15). Aliquots of membrane homogenates were stored at Ϫ80°C. Protein concentrations were determined by using the BCA protein assay kit.
Radioligand Binding Assays-[ 3 H]NMS binding studies were carried out with membrane homogenates prepared from yeast (ϳ250 g/tube) or COS-7 cells (ϳ20 g/tube). The binding buffer consisted of 25 mM phosphate buffer, pH 7.4, containing 5 mM MgCl 2 . Binding reactions were carried out for 3 h at 25°C in a 1-ml final volume. In saturation binding assays, six different concentrations of [ 3 H]NMS, ranging from 13 to 2,800 pM were used. In competition binding assays, incubations were carried out in the presence of 500 pM (yeast) or 250 pM (COS-7 cells) [ 3 H]NMS and 10 different concentrations of the competing ligand, carbachol (0.14 M to 10 mM). Nonspecific binding was determined in the presence of 10 M atropine. Bound and free ligand were separated by vacuum filtration over GF/B filters (Whatman) that had been pretreated with 0.3% polyethyleneimine for 3 h. The filters were washed three times with 5 ml of ice-cold phosphate buffer, dried, and placed in vials with 7 ml of Biosafe II scintillation mixture (RPI Corporation). Radioactivity bound to the filters was determined after 18 h of extraction. Radioligand binding data were analyzed by using the computer program Prism 3.0.

RESULTS
We reported previously that a modified version of the rat M 3 muscarinic receptor lacking the central portion of the i3 loop (Ala-274 to Lys-469) can be functionally expressed in the haploid yeast (S. cerevisiae) strain MPY578q5 (14,15). For the sake of simplicity, this i3 loop-shortened version of the M 3 receptor is referred to as 'WT' M 3 receptor throughout this study. The MPY578q5 yeast strain was modified genetically such that it required productive receptor/G protein coupling for growth in histidine-deficient medium (for details, see "Experimental Procedures"). Importantly, the MPY578q5 strain harbors a mutant version of the GPA1 gene coding for a hybrid yeast/mammalian G protein ␣ subunit in which the last five amino acids of Gpa1p were replaced with the corresponding mammalian G␣ q residues (14). For yeast expression studies, all receptor mutations were introduced into the 'WT'-M 3 -p416GPD centromeric yeast expression plasmid (14,54), thus placing receptor expression under the control of the strong yeast GPD promoter.

Characterization of the D113N Mutant M 3 Muscarinic Receptor Expressed in Yeast-
The TM II Asp 2.50 residue (corresponding to Asp 2.50 in the rat M 3 muscarinic receptor; Fig. 1) is highly conserved among class A GPCRs and is known to be critically involved in receptor activation (29 -44). To explore further the role of this residue in M 3 receptor function, we expressed the 'WT' M 3 receptor containing the D113N 2.50 point mutation in the MPY578q5 yeast strain. To study the ability of the D113N mutant M 3 receptor to interact with the coexpressed hybrid Gpa1p/␣ q G protein, we monitored yeast growth in histidine-deficient liquid medium in the presence of increasing concentrations of the muscarinic agonist, carbachol. Under these conditions, carbachol (10 Ϫ12 to 10 Ϫ3 M) stimulated the growth of the 'WT' M 3 receptor-expressing yeast strain in a concentration-dependent fashion (Fig. 2). In contrast, the D113N point mutation virtually abolished M 3 receptor function (Fig. 2).
Radioligand binding studies showed that the D113N mutant receptor displayed [ 3 H]NMS and carbachol binding affinities that differed from the corresponding 'WT' receptor values by less than 3-fold (Table I). The D113N point mutation led to a ϳ50% reduction in the number of detectable [ 3 H]NMS binding sites (B max ; Table I). However, previous studies using the same experimental conditions and the same yeast expression system have shown that the reduction of 'WT' M 3 receptor expression levels by more than 50% had little effect on receptor function (15), indicating that the reduction in B max values observed with the D113N mutant receptor cannot be responsible for the G protein coupling defect observed with this mutant receptor.
We recently found that the Q490L 6.35 point mutation (Gln-490 is located at the cytoplasmic end of TM VI; Fig. 1) led to robust agonist-independent M 3 receptor signaling in both yeast and mammalian cells (15). Introduction of the activating Q490L point mutation into a series of functionally impaired mutant M 3 receptors led to mutant receptors that showed 'WT' receptor-like G protein coupling properties (15). To examine whether the Q490L point mutation was also able to restore function to the D113N mutant receptor, we characterized the D113N/Q490L double mutant M 3 receptor in yeast growth and radioligand binding assays. Yeast growth assays showed that the Q490L point mutation did not lead to increased basal signaling and failed to restore agonist-mediated G protein coupling to the D113N mutant receptor (data not shown). In radioligand binding studies, the D113N/Q490L double mutant M 3 receptor showed [ 3 H]NMS and carbachol binding properties similar to those of the D113N mutant receptor ( Table I).

Identification of Second-site Suppressor Mutations That Restore Function to the D113N Mutant Receptor in Yeast-
In an effort to learn more about the molecular mechanisms by which the D113N point mutation interferes with receptor activation, we subjected the D113N mutant M 3 receptor to random mutagenesis. Our specific goal was to identify secondary point mutations (so-called second-site suppressor mutations) that were able to restore function to the D113N mutant receptor. We anticipated that the structure and location of such second-site suppressor mutations might yield new information about the functional role of the highly conserved Asp-113 residue.
Specifically, we subjected a region of the 'WT' M 3 receptor ranging from the middle of TM II to the C-terminal tail (Leu-114 to Gln-587; Fig. 1) to PCR-based random mutagenesis. We then used a gap-repair method (51, 52) that involved cotransformation of the MPY578q5 strain with a gapped version of the 'WT'-M 3 (D113N)p416GPD plasmid and a PCR fragment containing the random point mutations (for details, see "Experimental Procedures"). In vivo recombination events then led to the reformation of circular plasmids coding for D113N mutant M 3 muscarinic receptors containing additional point mutations (51,52).
To evaluate the quality of the generated mutant receptor library, we recovered and sequenced plasmids from 20 randomly picked colonies grown on SC-Ura plates. We found that all sequenced clones contained the D113N point mutation and an average of only ϳone additional nucleotide change/receptor, reducing the likelihood that the yeast genetic screen would yield primarily mutant receptors with multiple point mutations. The sequencing results also showed that nucleotide substitutions were distributed evenly throughout the targeted receptor sequence (Leu-114 to Gln-587).
In the next step, ϳ200,000 yeast transformants (containing randomly mutated D113N mutant M 3 receptors) grown on plates containing uracil-deficient medium were replicaplated onto plates containing medium lacking uracil and histidine (SC-Ura/His). The transformants were replicaplated on plates that did not contain any carbachol as well as on plates that contained 1 mM carbachol. We then selected  (14), and liquid bioassays were carried out as described under "Experimental Procedures." All mutations were introduced into 'WT'-M 3 -p416GPD. Yeast growth was measured by determining the absorbance at 630 nm, either in the absence or in the presence of carbachol (for carbachol EC 50 values, see Table III). The curves shown are representative of five independent experiments, each carried out in triplicate. Data are given as the means Ϯ S.D. (note that error bars are smaller than the size of the symbols).  (14). Amino acid numbers refer to positions in the full-length rat M 3 muscarinic receptor sequence (70). ͓ 3 H͔NMS saturation and carbachol competition binding studies were carried out using yeast membrane homogenates, as described under "Experimental Procedures." In the carbachol competition binding experiments, ͓ 3 H͔NMS was used at a concentration of 500 pM. Carbachol IC 50 values were converted to K i values by using the Cheng-Prusoff equation (71). Binding data were evaluated using the nonlinear curve fitting program Prism 3.0. Data are given as the means Ϯ S.D. of three independent experiments, each carried out in duplicate.
Receptor those colonies that did not grow in the absence of carbachol but grew in its presence. This selection procedure greatly reduced the recovery of false positive yeast clones caused by other, nonreceptor-related yeast mutations. This screen led to the identification of 30 yeast colonies that were able to grow in a carbachol-dependent fashion.
To confirm further that the observed growth phenotype was indeed plasmid-dependent, receptor plasmids were recovered from these clones, amplified in E. coli, and retransformed into the MPY578q5 yeast strain. Liquid bioassays of the resulting transformants demonstrated that 19 of the resulting transformants displayed carbachol-dependent growth, in a fashion similar to the 'WT' M 3 receptor. Sequencing of the 19 corresponding receptor plasmids led to the identification of eight different mutant receptors (Table II). Strikingly, two of the recovered mutant receptors contained the Y250D point mutation, and six of them contained a mutation of Arg-165, either a Met or a Trp substitution, either alone or in combination with other point mutations. Because the R165W point mutation was recovered 12 times, we concluded that the yeast genetic screen was performed under saturating conditions. Interestingly, Tyr-250 5.58 and Arg-165 3.50 are among the ϳ20 residues that are highly conserved among class A GPCRs (8).
Functional Characterization of Second-site Suppressor Mutations in Yeast-Comparison of carbachol concentration-response curves obtained with the eight recovered mutant receptors indicated that the Y250D, R165W, and R165M point mutations were responsible for the ability of the recovered mutant receptors to restore function to the D113N mutant receptor (data not shown; Fig. 2). Strikingly, in the presence of the Y250D and R165M point mutations, the D113N mutant receptor gained the ability to function in a fashion similar to the 'WT' receptor, as shown in yeast liquid bioassays ( Fig. 2 and Table III). The R165W point mutation also restored function to the D113N mutant receptor, although with significantly reduced (ϳ250-fold) carbachol potency compared with the 'WT' receptor ( Fig. 2 and Table III). The three second-site suppressor mutations had no significant effect on the basal activity of the D113 mutant receptor, measured in the absence of carbachol (Fig. 2).
Functional Effect of Introducing the Y250D, R165M, and R165W Point Mutations into the 'WT' Receptor Background in Yeast-To gain insight into the mechanism by which the Y250D, R165M, and R165W point mutations restored function to the coupling-deficient D113N mutant receptor, we introduced each of these three point mutations into the 'WT' receptor background. The resulting mutant receptors were expressed in the MPY578q5 yeast strain and then analyzed in liquid bioassays (Fig. 3 and Table III). All three mutant receptors showed little basal activity (yeast growth measured in the absence of carbachol), similar to the 'WT' receptor (Fig. 3). Interestingly, all three mutant receptors showed greatly reduced carbachol potencies compared with the 'WT' receptor (-fold increase in carbachol EC 50 values: Y250D, ϳ12-fold; R165M, ϳ160-fold; R165W, ϳ90,000-fold) (Fig. 3 and Table  III). This observation indicated that the Y250D, R165M, and R165W point mutations that were able to restore function to the D113N mutant receptor led to impaired receptor/G protein coupling when introduced into the 'WT' receptor.
Analysis of the Three Identified Second-site Suppressor Mutations in Mammalian Cells-We next examined whether the key results obtained with the yeast expression system involv-  (14). Amino acid numbers refer to positions in the full-length rat M 3 muscarinic receptor sequence (70).
b Positions according to Ballesteros and Weinstein (28).  (14), as described under "Experimental Procedures." All mutations were introduced into a modified version of the rat M 3 muscarinic receptor (referred to as 'WT') lacking the central portion of the i3 loop (Ala-274 to Lys-469) (14). Amino acid numbers refer to positions in the full-length rat M 3 muscarinic receptor sequence (70). Yeast growth assays (liquid bioassays) were carried out as described under "Experimental Procedures." Yeast growth was measured in the absence or the presence of increasing concentrations of carbachol by determining the absorbance at 630 nm. The extent of basal growth measured in the absence of carbachol was not significantly affected by the different single and double point mutations (see Figs. 2 and 3). Data were evaluated using the nonlinear curve fitting program Prism 3.0. Data are given as the means Ϯ S.D. of three independent experiments, each carried out in triplicate.   Table III. ing the use of a chimeric yeast/mammalian ␣ q G protein ␣ subunit could be reproduced in a mammalian expression system. Initially, we introduced the D113N point mutation, either alone or together with the Y250D, R165M, or R165W point mutations, into a pCD-based mammalian expression plasmid coding for the full-length WT M 3 receptor. The resulting mutant receptors were transiently expressed in COS-7 cells and then characterized in functional studies (carbachol-induced PI hydrolysis) and radioligand binding assays ( Fig. 4 and Table  IV). In cells expressing the WT M 3 receptor, carbachol elicited a robust PI response characterized by a carbachol EC 50 value of 93.1 Ϯ 2.8 nM (Fig. 4 and Table IV). As observed in the yeast expression system, the D113N mutant receptor was virtually devoid of functional activity ( Fig. 4 and Table IV). Moreover, as found in yeast, the Y250D, R165M, and R165W point mutations restored functional activity to the D113N mutant receptor ( Fig. 4 and Table IV). The three double mutant receptors (D113N/R165M, D113N/R165W, and D113/Y250D) showed carbachol EC 50 values that were similar to the corresponding WT receptor value (Fig. 4 and Table IV). Whereas the D113N/ R165M and D113N/R165W receptors displayed E max values similar to the WT receptor, the D113/Y250D construct showed a ϳ50% reduction in E max (Fig. 4 and Table IV). Interestingly, the D113N/R165M mutant receptor also showed a robust increase in basal activity (ϳ6-fold compared with the WT receptor; Fig. 4 and Table IV). In contrast, the degree of basal signaling remained unchanged in the D113N/R165W and D113/Y250D receptors ( Fig. 4 Table IV).

Characterization of Mutant Receptors Containing the Y250D, R165M, or R165W Point Mutation in the WT M 3 Receptor
Background in Mammalian Cells-PI assays showed that introduction of the Y250D, R165M, or R165W point mutation into the WT M 3 receptor background resulted in mutant receptors that displayed a decrease in the efficiency of receptor/G protein coupling in transfected COS-7 cells (Fig. 5 and Table IV), similar to our observations made using the yeast expression system (Fig. 3). As shown in Fig. 5 and Table IV, the three mutant receptors showed significantly reduced carbachol potencies and E max values. The R165W receptor showed the strongest degree of functional impairment, displaying an average E max value of only ϳ27% (WT ϭ 100%). Basal signaling, studied in the absence of carbachol, was not increased in the Y250D, R165M, and R165W mutant constructs compared with the WT receptor ( Fig. 5 and Table IV).
Functional Effects of Replacing Arg-165 and Tyr-250 with Other Amino Acids Studied in Mammalian Cells-As described in the previous paragraphs, the R165M and R165W point mutations restored robust functional activity to the D113N mutant M 3 receptor in yeast and mammalian cells. To examine further to what extent the degree of functional rescue was dependent on the identity of the amino acid replacing Arg-165, we used site-directed mutagenesis to generate several additional mutant receptors in which Arg-165 was replaced with various other amino acids differing in size, polarity, charge, or hydrophobicity (Ala, Gln, Leu, or Glu). PI assays with transfected COS-7 cells showed that the R165A, R165Q, and R165L substitutions were partially able to restore function to the inactive D113N mutant M 3 receptor ( Fig. 6 and Table IV). The D113N/R165A receptor showed the highest degree of functional activity, displaying an E max value of ϳ67%. In contrast, the D113N/R165E receptor was virtually inactive, similar to the D113N mutant receptor ( Fig. 6 and Table IV). On the other hand, when introduced into the WT receptor background, all four point mutations (R165A, R165Q, R165L, and R165E) led to strong reductions in receptor/G protein coupling efficiency, as indicated by significantly decreased E max values and pronounced rightward shifts of carbachol concentration-response curves ( Fig. 7 and Table IV). In fact, the R165L mutant receptor was nearly devoid of functional activity (Fig. 7). None of the R165-derived mutant receptors, either in the D113N or in the WT receptor background, showed any increases in basal receptor activity (Figs. 6 and Table IV).
Radioligand binding studies showed that all Arg-165-derived mutant receptors were highly expressed in transfected COS-7 cells and showed [ 3 H]NMS binding affinities similar to the WT receptor (Table IV). The four point mutations (R165A, R165Q, R165L, and R165E), when introduced into either the WT or the D113N mutant receptor background, had relatively little effect on carbachol binding affinities (Table IV).
As outlined above (Figs. 2 and 4), the Y250D point mutation was able to functionally rescue the coupling-deficient D113N mutant receptor in both yeast and mammalian cells. To study the importance of the Asp-250 side chain for this activity, we also introduced the Y250A point mutation into the D113N mutant receptor background. Functional studies with trans- FIG. 4. Functional rescue by second-site mutations of the D113N mutant M 3 muscarinic receptor expressed in mammalian cells. The indicated point mutations were introduced into the mammalian expression plasmid, WT-M 3 -pCD, and the resulting mutant M 3 receptors were transiently expressed in COS-7 cells. Carbachol-mediated increases in intracellular [ 3 H]IP 1 levels were determined using 6-well plates, as described under "Experimental Procedures." The curves shown are representative of three independent experiments, each carried out in duplicate. Data are given as the means Ϯ S.D. (for a summary of carbachol EC 50 and E max values, see Table IV).
fected COS-7 cells showed that the D113N/Y250A receptor, in contrast to the D113N/Y250D construct, was functionally inactive, similar to the D113N mutant receptor (Fig. 8 and Table  IV). However, in radioligand binding studies, the D113N/ Y250A and D113N/Y250D receptors displayed comparable B max values and similar [ 3 H]NMS and carbachol binding affinities (Table IV). When introduced into the WT M 3 receptor, the Y250A point mutation had relatively little effect on G protein coupling and ligand binding properties (Fig. 8 and Table IV).
Allele Specificity of the Y250D, R165M, and R165W Secondsite Suppressor Mutations-We reported previously (55) that a mutant M 3 receptor in which the positions of Ile-253 and Tyr-254 had been exchanged (referred to as I253Y/Y254I or simply as EX) was unable to activate G proteins, similar to the D113N receptor. Moreover, the present study led to the identification of other coupling-deficient mutant M 3 receptors, such as R165L and R165E (Fig. 7). To examine whether the second-site suppressor mutations identified in the present study were functionally able to rescue the EX mutant receptor, we generated two additional mutant receptors in which we combined the EX mutation with the R165M or R165W point mutations. Similarly, we generated a mutant receptor in which we combined the R165E substitution with the Y250D point mutation. Functional studies with transfected COS-7 cells showed that all three resulting double mutant receptors remained functionally inactive ( Fig. 9 and Table IV). The Y250D, R165M, and R165W second-site suppressor mutations therefore displayed allele specificity in that they specifically suppressed the functionally detrimental effects of the D113N point mutation.
Radioligand binding studies indicated that the EX/R165W receptor was expressed at very low levels (B max Ͻ 0.1 pmol/mg; Table IV). The EX, EX/R165M, and R165E/Y250D mutant receptors exhibited considerably higher B max values, ranging from 0.54 to 1.84 pmol/mg (Table IV). The EX, EX/R165M, and R165E/Y250D mutant receptors displayed [ 3 H]NMS and carbachol binding affinities similar to the WT receptor, indicating that these three mutant receptors were folded properly (Table IV). DISCUSSION The TM II Asp 2.50 residue is highly conserved among class A GPCRs (8). In the x-ray structure of the inactive state of bovine  Table IV).

TABLE IV Ligand binding and functional properties of mutant M 3 muscarinic receptors expressed in COS-7 cells
The indicated point mutations were introduced into the full-length (WT) rat M 3 muscarinic receptor. All receptor constructs were transiently expressed in COS-7 cells. Radioligand binding studies and PI assays were carried out as described under "Experimental Procedures." Carbachol binding data were corrected for the Cheng-Prusoff shift (71). Radioligand binding and functional data were evaluated using the nonlinear curve fitting program Prism 3.0. In the case of the WT receptor, basal inositol monophosphate (IP 1 ) levels were 4,178 Ϯ 520 dpm, and maximal carbachol-stimulated IP 1 levels were 63,531 Ϯ 9,725 dpm, respectively. In each individual experiment, the basal IP 1 accumulation displayed by the WT receptor was set equal to 1, and the maximum carbachol-stimulated response mediated by the WT receptor was set equal to 100%. Data are presented as the means Ϯ S.D. of three (binding assays) or three to five independent experiments (PI assays), each performed in duplicate. Many studies using different GPCR subtypes have shown that mutational modification of the conserved TM II Asp 2.50 residue causes profound defects in the efficiency of receptor/G protein coupling (29 -44). Similarly, we demonstrated in the present study that the D113N 2.50 mutant M 3 receptor is unable to interact productively with G proteins in yeast or transfected COS-7 cells. Taken together, these findings suggest that Asp 2.50 plays a key role in mediating the conformational changes that convert an inactive class A GPCR into its activated state. Consistent with this concept, spectroscopic studies with bovine rhodopsin have shown that Asp-83 2.50 undergoes changes in its H-bonding pattern during the formation of the activated receptor state (metarhodopsin II) (56,57). In the present study, we replaced Asp 2.50 with Asn which, in contrast to Asp, can act as both H-bond donor and acceptor. The resulting, as yet undefined changes in the interhelical network of H-bond interactions are predicted to prevent the coordinated structural changes necessary for agonist-induced receptor activation.

Receptor
To learn more about the functional role of the Asp 2.50 residue, we subjected the D113N 2.50 mutant M 3 muscarinic receptor to PCR-based random mutagenesis and employed a yeast genetic screen to identify second-site suppressor mutations that can restore function to the D113N mutant receptor. This screen led to the recovery of three point mutations, R165W 3.50 , R165M 3.50 , and Y250D 5.58 , which enabled the D113N mutant receptor to interact productively with G proteins when expressed in yeast (note, however, that the D113N/R165W double mutant receptor showed a pronounced decrease in carbachol potency in yeast; Fig. 2). Interestingly, like Asp-113 2.50 , Arg-165 3.50 and Tyr-250 5.58 are also highly conserved among class A GPCRs (8).
To examine whether similar results could be obtained in a mammalian expression system expressing native G proteins (rather than a chimeric yeast Gpa1p/mammalian ␣ q G␣ subunit as is the case in yeast), we also characterized the D113N/ R165M, D113N/R165W, and D113/Y250D double mutant receptors in transiently transfected COS-7 cells. Whereas the D113N mutant receptor was unable to stimulate agonist-dependent PI hydrolysis to an appreciable extent, the three double mutant receptors all regained the ability to stimulate PI hydrolysis with high carbachol potency ( Fig. 4 and Table IV). The D113N/R165M and D113N/R165W receptors showed E max values similar to the WT receptor, whereas the D113/Y250D construct showed a ϳ50% reduction in E max (Fig. 4). Interestingly, the D113N/R165M mutant receptor displayed pronounced functional activity even in the absence of agonist (Fig.  4), indicating that this receptor is constitutively active. Interestingly, PI assays with transfected COS-7 cells also showed that introduction of the R165W, R165M, and Y250D point mutations into the WT receptor background resulted in mutant receptors that displayed significant reductions in carbachol potencies and E max values (basal activities remained unchanged; Fig. 5 and Table IV). Similar functional impairments were also observed in the yeast expression system (Fig.  3). These observations rule out the possibility that the ability of the R165W, R165M, and Y250D point mutations to restore function to the D113N mutant receptor is simply because these three point mutations render the WT M 3 receptor hyperactive.
In a recent study, we showed that the Q490L 6.35 point mutation leads to robust agonist-independent M 3 receptor signaling in both yeast and mammalian cells (15). Introduction of this point mutation into a series of functionally impaired mutant M 3 receptors led to receptors that were able to interact with G proteins with high efficiency (15). However, the Q490L/D113N double mutant M 3 receptor remained functionally inactive, indicating that the Q490L 6.35 point mutation is not a general activator mutation.
Although the R165W, R165M, and Y250D point mutations restored function to the D113N mutant receptor in both yeast and mammalian cells, the three resulting double mutant receptors showed somewhat different functional properties in the two expression systems. For example, whereas all double mutant receptors gave maximum functional responses in yeast (Fig. 2), E max values ranged from 51 to 118% in transfected COS-7 cells (Fig. 4). Also, whereas the D113N/Y250D mutant receptor showed 'WT' receptor-like coupling properties in yeast (Fig. 2), it displayed a ϳ50% reduction in E max in COS-7 cells (Fig. 4). Possible reasons for these discrepancies include the presence of different G proteins in the two expression systems and differences in receptor levels and the nature of the functional assays employed (carbachol-induced cell growth over a 3-day period in yeast versus carbachol-induced short term accumulation of IP 1 in COS-7 cells).
Based on its location deep in the TM receptor core, Asp 2.50 is unlikely to be in direct contact with acetylcholine (carbachol) or other biogenic amine neurotransmitters bound to their target receptors (13,58). In transfected COS-7 cells, the D113N 2.50 point mutation increased the affinity of carbachol, a stable acetylcholine derivative, by ϳ8-fold (Table IV), indicating that the chemical nature of the amino acid present at position 2.50 has a pronounced effect on the configuration of the agonist binding pocket. Interestingly, the D113N/R165M, D113N/ R165W, and D113N/Y250D mutant receptors showed additional increases in carbachol binding affinities (ϳ25-240-fold, compared with the WT receptor). These findings support the view (see discussion below) that the R165W, R165M, and Y250D point mutations, all of which are located on the cytoplasmic receptor surface (Fig. 10), exert indirect conformational effects on the extracellular portion of the TM helical bundle where the binding of biogenic amine ligands is predicted to occur.
Because the presence of the R165W, R165M, and Y250D point mutations in the 'WT'/WT M 3 muscarinic receptor background impaired receptor/G protein coupling, it is rather surprising that these three point mutations were able to restore function, at least partially, to the coupling-deficient D113N mutant receptor. These results clearly indicate that the Asp- and Tyr-250 5.58 within the TM receptor core are shown in Fig.  10. Arg-165 is located at the bottom of TM III and is part of the conserved D/ERY, motif which is known to be critically involved in receptor activation (5,7). Tyr-250 is located at the cytoplasmic end of TM V, close to the N terminus of the i3 loop, which plays a critical role in determining the selectivity and efficacy of receptor/G protein interactions (7). Consistent with the high resolution structure of bovine rhodopsin, Arg-165 and Tyr-250 are not located adjacent to Asp-113 (Fig. 10). Also, Arg-165 and Tyr-250 are unlikely to interact directly with each other in the inactive state of the receptor (9 -11; Fig. 10). The most likely scenario therefore is that Arg-165, Tyr-250, and Asp-113 participate in a shared network of interhelical interactions which is critical for agonist-induced M 3 receptor activation. Our data support the concept that the R165W, R165M, and Y250D suppressor mutations indirectly affect this network of interhelical interactions, thus allowing the formation of a coupling-competent receptor conformation despite the presence of the primary D113N mutation. The availability of a high resolution structure of the active form of a class A GPCR is expected to shed light on the detailed structural mechanisms underlying the observed rescue phenomenon. Given the conserved nature of Asp-113 2.50 , Arg-165 3.50 , Tyr-250 5.58 , our results should be of general relevance for class A GPCRs.
As has been observed with many different GPCRs, the Asp/ Asn 2.50 point mutation also functionally inactivates the 5-HT 2A receptor (26). Interestingly, the resulting mutant 5-HT 2A receptor could be functionally rescued, at least partially, by simultaneously mutating Asn 7.49 to Asp (26). Similar results have been obtained with the tachykinin NK 2 receptor (25) and an ␣ 2A adrenoreceptor/G i1 -␣ fusion protein (24). The high resolution structure of bovine rhodopsin suggests that the side chains of Asp 2.50 and Asn 7.49 are linked directly by a water molecule (11). Our findings therefore differ from these previous rescue experiments in that the recovered second-site suppressor mutations are not located adjacent to the primary inactivating mutation and are therefore predicted to rescue receptor function through indirect conformational effects.
In our yeast genetic screen, the Asn/Asp 7.49 point mutation was not found among the recovered second-site suppressor mutations that could restore function to the D113N mutant M 3 receptor. One possible explanation for this observation is that our yeast genetic screen was carried out under very stringent conditions designed to recover mutant receptors showing maximum functional activity.
Mutagenesis studies suggest that the highly conserved Arg 3.50 residue plays a key role in G protein recognition and activation (for review, see Ref. 7). For example, Prossnitz et al. (41) showed that an N-formyl peptide mutant receptor lacking the conserved Arg 3.50 residue was unable to associate physically with G i . Essentially similar findings were obtained with a mutant version of rhodopsin in which the positions of Arg 3.50 and Glu 3.49 were reversed (59,60). However, the results of the present study clearly demonstrate that the presence of Arg 3.50 (or another basic amino acid instead; 61) is not an absolute requirement for efficient receptor/G protein coupling. The high resolution x-ray structure of the inactive state of bovine rhodopsin indicates that Arg 3.50 is involved in a network of interactions involving Glu 3.49 and two polar/charged residues located at the cytoplasmic end of TM VI (9 -11). Molecular modeling studies suggest that a similar set of interactions exists in the M 3 muscarinic receptor. 2 Taken together, our results support the concept that Arg 3.50 does not directly interact with G proteins but plays a role in the conformational rearrangement of the interhelical interactions that allow productive receptor/G protein coupling.
Site-directed mutagenesis studies showed that the degree of functional rescue of the D113N mutant M 3 receptor was critically dependent on the identity of the amino acid replacing Arg-165 3.50 (rank order of E max values: Table IV). This finding suggests that the observed rescue phenomenon is not simply the result of the loss of interactions in which Arg-165 3.50 normally participates but that the newly introduced amino acids themselves can critically affect receptor structure. Consistent with this concept, the Y250A point mutation, in contrast to Y250D, was unable to restore function to the D113N mutant receptor (Fig. 8).
The high resolution x-ray structure of the inactive state of 2 S.-K. Kim, K. A. Jacobson, and J. Wess, unpublished results.

FIG. 10. Predicted location of key M 3 muscarinic receptor residues targeted in the present study.
A three-dimensional model of the inactive state of the rat M 3 muscarinic receptor was built via homology modeling using the high resolution x-ray structure of bovine rhodopsin as a template (Ref. 9 and Footnote 2). A, side view of the TM helical bundle of the M 3 muscarinic receptor. B, cytoplasmic view of the TM receptor core of the M 3 muscarinic receptor. Note that Asp-113 2.50 , which is situated in the middle of TM II, is not located adjacent to Arg-165 3.50 and Tyr-250 5.58 , which are located at the cytoplasmic ends of TM III and V, respectively. Although Arg-165 3.50 and Tyr-250 5.58 are located relatively close to each other on the cytoplasmic surface of the receptor protein, the model predicts that their amino acid side chains do not interact with each other in the inactive state of the receptor. bovine rhodopsin indicates that Tyr-223 5.58 (corresponding to Tyr-250 5.58 in the M 3 muscarinic receptor) partially covers the interhelical region between TM V and VI near the lipid interface (9 -11). Interestingly, a recent model (62) of the activated state of bovine rhodopsin suggests that the side chains of Tyr-223 5.58 and Arg-135 3.50 do not lie adjacent to each other in the inactive state of rhodopsin but face each other after receptor activation. Assuming that a similar structural change occurs in the M 3 muscarinic receptor, one possibility therefore is that the Y250D 5.58 point mutation promotes receptor activation by stabilizing the activated state of the M 3 receptor through an electrostatic interaction between the Arg-165 3.50 and Asp-250 5.58 side chains.
Although the Y250D, R165M, and R165W point mutations were able to restore function to the Asp-113 mutant M 3 receptor, additional studies showed that these three point mutations were unable functionally to rescue other mutant M 3 receptors containing different inactivating mutations. For example, we demonstrated that the R165M substitution was unable to restore function to a mutant receptor containing an inactivating mutation at the cytoplasmic end of TM V (I253Y/Y254I ϭ EX; Fig. 9). Although the resulting mutant receptor was expressed at considerably lower levels than the WT M 3 or the D113N mutant receptors (B max ϳ0.54 pmol/mg), previous studies with transiently transfected COS-7 cells have shown that robust carbachol-mediated PI hydrolysis can be observed at much lower M 3 receptor densities (B max Յ 0.1 pmol/mg; 45,55). This observation suggests that the inability of the R165M substitution to restore function to the EX mutant receptor is not simply the result of reduced receptor expression levels. Similarly, the Y250D substitution failed to improve the G protein coupling efficiency of the R165E mutant receptor (Fig. 9). These data indicate that the second-site suppressor mutations identified in the present study are not general activator mutations but can specifically overcome the detrimental structural effects of the D113N mutation.
We demonstrated recently (65) that M 3 muscarinic receptors, like many other GPCRs (63,64), are able to form dimers or oligomers, at least under certain experimental conditions. Accumulating evidence suggests that this process is critically involved in many aspects of GPCR function (63). Recent work indicates that multiple TM domains are likely to be involved in the oligomerization of class A GPCRs (64). Several studies suggest, for example, that residues located on TM II may contribute to the formation of GPCR dimers or oligomers (66 -69). The possibility therefore exists, at least theoretically, that the D113N point mutation interferes with M 3 receptor dimerization or oligomerization and that the recovered second-site suppressor mutations allow the D113N mutant receptor to regain the ability to form receptor complexes. However, preliminary Western blotting studies using yeast membranes, analogous to those carried out previously with membranes derived from transfected COS-7 cells (65), demonstrated that the D113N M 3 mutant receptor, similar to the 'WT' M 3 receptor, was able to form SDS-resistant dimers (data not shown). It is therefore unlikely that the ability of the recovered second-site suppressor mutations to rescue the D113N M 3 mutant receptor functionally involves changes in receptor dimerization patterns. However, more sophisticated techniques, including, for example, the use of FRET or BRET technology in living cells, are needed to rule out conclusively the possibility that the mutations described in the present study affect the ability of the M 3 receptor to form dimers or higher order oligomers.
In conclusion, the strategy described here (receptor random mutagenesis followed by yeast genetic screens) allows the isolation of rare mutant receptors with novel coupling properties.
From an evolutionary point of view, the mutant receptors isolated in this study provide a good example of how functionally detrimental primary GPCR mutations can be compensated for by second-site substitutions. The approach described here should be useful to shed new light on the functional roles of the many other residues that are highly conserved among class A GPCRs.