Mutational and Computational Analysis of the (cid:1) 1b -Adrenergic Receptor INVOLVEMENT OF BASIC AND HYDROPHOBIC RESIDUES IN RECEPTOR ACTIVATION AND G PROTEIN COUPLING* □ S

To investigate their role in receptor coupling to G q , we mutated all basic amino acids and some conserved hydrophobic residues of the cytosolic surface of the (cid:1) 1b - adrenergic receptor (AR). The wild type and mutated receptors were expressed in COS-7 cells and characterized for their ligand binding properties and ability to increase inositol phosphate accumulation. The experimental results have been interpreted in the context of both an ab initio model of the (cid:1) 1b -AR and of a new homology model built on the recently solved crystal structure of rhodopsin. Among the twenty-three basic amino acids mutated only mutations of three, Arg 254 and Lys 258 in the third intracellular loop and Lys 291 at the cytosolic extension of helix 6, markedly impaired the receptor-mediated inositol phosphate production. Addi-tionally, mutations of two conserved hydrophobic residues, Val 147 and Leu 151 in the second intracellular loop had significant effects on receptor function. The functional analysis of the receptor mutants in conjunction with the predictions of molecular modeling supports the hypothesis that Arg 254 , Lys 258 , as well as Leu 151 are directly involved in receptor-G protein interaction and/or receptor-mediated activation of the

To investigate their role in receptor coupling to G q , we mutated all basic amino acids and some conserved hydrophobic residues of the cytosolic surface of the ␣ 1badrenergic receptor (AR). The wild type and mutated receptors were expressed in COS-7 cells and characterized for their ligand binding properties and ability to increase inositol phosphate accumulation. The experimental results have been interpreted in the context of both an ab initio model of the ␣ 1b -AR and of a new homology model built on the recently solved crystal structure of rhodopsin. Among the twenty-three basic amino acids mutated only mutations of three, Arg 254 and Lys 258 in the third intracellular loop and Lys 291 at the cytosolic extension of helix 6, markedly impaired the receptor-mediated inositol phosphate production. Additionally, mutations of two conserved hydrophobic residues, Val 147 and Leu 151 in the second intracellular loop had significant effects on receptor function. The functional analysis of the receptor mutants in conjunction with the predictions of molecular modeling supports the hypothesis that Arg 254 , Lys 258 , as well as Leu 151 are directly involved in receptor-G protein interaction and/or receptor-mediated activation of the G protein. In contrast, the residues belonging to the cytosolic extensions of helices 3 and 6 play a predominant role in the activation process of the ␣ 1b -AR. These findings contribute to the delineation of the molecular determinants of the ␣ 1b -AR/G q interface.
The ␣ 1b -adrenergic receptor (␣ 1b -AR) 1 belongs to the superfamily of G protein-coupled receptors (GPCRs) that transmit a variety of signals across the cell membrane. Stimulation of the ␣ 1b -AR by catecholamines activates proteins of the G q family, resulting in the production of inositol phosphates (IP) via the activation of phospholipase C (PLC) (1).
GPCRs are structurally characterized by seven transmembrane ␣-helices connected by alternating extracellular (e) and intracellular (i) loops. While the extracellular portion of these receptors is primarily involved in ligand binding, the cytosolic loops mediate the interaction of the receptors with a number of signaling and regulatory proteins, including G proteins, arrestins, and G protein-coupled receptor kinases (reviewed in Ref. 2).
Evidence suggests that a conformational adjustment within the helical bundle of the receptor underlies the process of agonist-induced activation of GPCRs (reviewed in Ref. 3). The current hypothesis is that the transition from the inactive (R) to active (R*) state of a GPCR results in receptor interaction with, and activation of, a G protein. Thus a GPCR-mediated biological response involves a series of events (i.e. receptor activation, receptor-G protein interaction, and receptor-induced G protein activation) for which a detailed mechanism still remains elusive at the molecular level. Although residues located in the helical bundle and at the boundary between the membrane and the cytosol may play a role in the "conformational switch" underlying receptor activation, amino acids in the intracellular loops are believed to be more directly involved in receptor-G protein interaction and/or receptor-induced G protein activation. The combination of these two latter events, which cannot be unequivocally separated experimentally, is generally indicated with the term of receptor-G protein coupling.
We have previously provided evidence that the negatively and positively charged amino acids of the conserved DRY motif at the cytosolic end of helix 3 play a key role in the activation process of the ␣ 1b -AR (4 -6). Following a combination of experimental and computer-simulated mutagenesis of the ␣ 1b -AR, we have hypothesized that protonation of the aspartate (Asp 142 ) and a shift of the arginine (Arg 143 ) out of a conserved "polar pocket" are crucial steps in the transition of the receptor from the inactive (R) to active (R*) state (4 -6).
Several studies have tried to identify the amino acids of different GPCRs involved in G protein coupling at both experimental (as reviewed in Ref. 2) and theoretical levels (7)(8)(9)(10). The majority of these studies indicate that sequences in the i2 loop as well as in the N and C termini of the i3 loop play an important role in the efficiency of receptor-G protein coupling and/or in the selectivity of receptor-G protein recognition. BBXXB or BBXB motifs located in different cytosolic loops (where B is any basic amino acid and X is any residue) have been implicated in the coupling of a number of GPCRs to G proteins (11)(12)(13). However, this motif has not been found to be universally important for all GPCRs. Other studies have identified hydrophobic amino acids as contributing to the receptor-G protein interface (14 -16). In conclusion, what has become abundantly clear is that there is no simple sequence determinant that can be attributed to receptor-G protein coupling.
In a recent modeling study (8), docking simulations between active forms of the ␣ 1b -AR and a G q heterotrimer led us to suggest that the positive surface of the cytosolic portion of GPCRs could complement a negative surface found on different G protein ␣ subunits and thereby play a role in receptor-G protein coupling. However, the docking simulations also suggested that, despite the large number of cationic amino acids, only some might interact with anionic residues in the G q ␣-subunit. To investigate the role of cationic residues in receptor-G protein coupling, we have mutated all the basic amino acids located in the i1, i2, and i3 loops of the ␣ 1b -AR and investigated the effect of these mutations on receptor-mediated production of IP. In addition, we have also characterized the effects resulting from mutations of conserved hydrophobic residues in the cytosolic portion of the receptor.
Our findings demonstrate that mutations of the basic residues in the cytosolic portion of the receptor have wide ranging phenotypes. Only mutations of three (Arg 254 , Lys 258 , and Lys 291 ) out of the twenty-three basic amino acids studied impaired the receptor-mediated signaling response. We also demonstrate an important role for two highly conserved hydrophobic residues in receptor function. The effect of these mutations has been evaluated in the context of both the ab initio model previously described (17) and a new model of the ␣ 1b -AR built on the recently solved 2.8-Å crystal structure of rhodopsin (18 Mutagenesis of the ␣ 1b -AR-The cDNA of the hamster ␣ 1b -AR (19) was mutated using PCR-mediated mutagenesis and Pwo DNA polymerase. The constructs were subcloned in the pRK5 expression vector, and mutations were confirmed by automated DNA sequencing (Microsynth GmbH, Switzerland).
Cell Culture and Transfection-COS-7 cells were grown in DMEM supplemented with 10% fetal bovine serum and gentamicin (100 g/ml) and transfected using the DEAE-dextran method. For inositol phosphate determination, COS-7 cells (0.15 ϫ 10 6 ) were seeded in 12-well plates. The quantity of transfected receptor encoding DNA was 0.3-3 g/10 6 cells.
Ligand Binding-Membrane preparations derived from cells expressing the ␣ 1b -AR or its mutants and ligand binding assays using [ 125 I]HEAT were performed as previously described (19). Prazosin (10 Ϫ6 M) was used to determine nonspecific binding. [ 125 I]HEAT at a concentration of 250 pM was used for measuring receptor expression at a single concentration and 80 pM for competition binding analysis. Saturation analysis and competition curves were analyzed using Prism 3.02 (GraphPad Software Inc., San Diego, CA).
Inositol Phosphate Measurements-Transfected cells were labeled for 12 h with myo-[ 3 H]inositol at 4 Ci/ml in inositol-free DMEM supplemented with 1% fetal bovine serum. Cells were preincubated for 10 min in phosphate-buffered saline containing 20 mM LiCl and then stimulated for 45 min with different concentrations of epinephrine from 10 Ϫ10 to 10 Ϫ4 M. Total inositol phosphates were extracted and separated as described previously (19).
Ab Initio Modeling of the ␣ 1b -AR and Its Mutants-Ab initio modeling of the ␣ 1b -AR receptor was achieved following the iterative procedure previously described (17). The wild type ␣ 1b -AR input structure was selected from among over 200 tested input arrangements according to both internal and external consistency criteria and was used to produce the input structures for the receptor mutants. These structures were obtained by substituting the mutated residue in the wild type input structure by means of the molecular graphics package QUANTA (release 98; Molecular Simulations Inc., Waltham, MA). Minimization and molecular dynamics (MD) simulations of the receptor models were performed using the program CHARMm (Molecular Simulations Inc), following the computational protocol previously described (17). In a previous study MD runs of 1050 ps were performed to compare the dynamic features of the wild type receptor with those of the constitutively active mutants (8). Because the first 100 ps of the equilibrated trajectory were sufficiently representative of the whole trajectory and given the high number of mutants considered in this study, MD runs of 150 ps were generally performed following the same heating and equilibration set-up as that employed for the longer MD simulations. The results reported were collected every 0.5 ps during the last 100 ps of the equilibrated MD trajectory. Finally, for each mutant the structure averaged over the 200 structures stored during the production phase were used for the comparative analysis.
The average minimized structure of the wild type ␣ 1b -AR showed a root mean square deviation (r.m.s.d.) of 3.94 Å from the rhodopsin structure, the deviation being larger (r.m.s.d. Comparative Modeling of ␣ 1b -AR and Its Mutants-Another model of the ␣ 1b -AR was built by comparative modeling (20) using the recently determined 2.8-Å x-ray structure of rhodopsin (18) as a template. Eight different chimeric ␣ 1b -AR/rhodopsin templates (shown in the supplementary material) were constructed in which the e2, the i3, and in some cases only the i2 loop were extracted from the input structure of the ab initio model of the ␣ 1b -AR. Furthermore, in the chimeras helix 5 has been elongated by 10 amino acids using the ␣ 1b -AR sequence after deleting the 226 -235 rhodopsin segment. Finally, an ␣-helical segment of 6 amino acids using the ␣ 1b -AR sequence has been added to the N terminus of the helix 6 of rhodopsin after deleting the 240 -248 rhodopsin segment. For each of the eight different templates, MODELLER generated 25 models. Among the 200 models finally obtained, 20 models were selected showing low restraint violations and low numbers of main-chain and side-chain bad conformations or close contacts. These models were completed by the addition of the polar hydrogen's and subjected to automatic and manual rotation of the side-chain torsion angles when in bad conformations, as well as to energy minimization and MD simulations according to the computational protocol employed for simulating the ab initio ␣ 1b -AR model. Different combinations of intra-helix distance constraints were also probed. About 450 MD trial runs were done to select the proper input structure for the wild type ␣1b-AR. The final input structure selected that was obtained using the alignment (see the supplementary material) was then used for generating the input structures for the receptor mutants. The structures of the wild type receptor and its mutants averaged over the last 100 ps of the 150 ps MD trajectory were finally minimized and considered for the comparative analysis.
The input structure of the wild type ␣ 1b -AR showed an r.m.s.d. of 0.17 Å from the rhodopsin structure (r.m.s.d. was computed by employing the matching criteria described above for the ab initio model). This deviation increases to 2.07 Å upon energy minimization and MD simulations, becoming quite close to the value that would be expected given a sequence identity of 22.4% between the transmembrane segments of the ␣ 1b -AR and rhodopsin (21).

Expression of Receptor Mutants-
The wild type and mutated ␣ 1b -ARs were expressed in COS-7 cells and tested for their ability to bind the radioligand [ 125 I]HEAT and epinephrine. Saturation binding experiments indicated that the K D of [ 125 I]HEAT was ϳ80 pM for all the receptors studied (results not shown), whereas the IC 50 values for epinephrine varied as indicated in Table I. The affinity of prazosin for the different receptor mutants was similar to that for the wild type ␣ 1b -AR (results not shown). Receptor coupling to the G q /PLC pathway was assessed as the ability of the receptor mutants to mediate epinephrine-stimulated IP accumulation (Table I). Transfections using 3 g of DNA per 1 ϫ 10 6 cells resulted in the expression of all receptor mutants at levels ranging from 60 to over 250 fmol/well. In each experiment, the wild type ␣ 1b -AR was expressed using varying quantities of DNA (0.3, 1.3, and 3 g of DNA/1 ϫ 10 6 cells) resulting in low (between 60 and 100 fmol/well), medium (between 100 and 200 fmol/well), and high (between 200 and 300 fmol/well) levels of expression. This allowed us to always be able to directly compare the properties of the mutated receptors with those of the wild type ␣ 1b -AR expressed at comparable levels within the same experiment. Fig. 1 shows the localization of the amino acids mentioned in this study within a simplified topographical scheme of the

-AR and its mutants
The wild type ␣ 1b -AR (WT) and its mutants were expressed in COS-7 cells. Receptor expression was measured using 250 pm of [ 125 I]HEAT on membrane preparations derived from transfected cells from one well of a six-well dish (approximately 150 g of protein). Inositol phosphate (IP) accumulation was measured following incubation in the absence (Basal) or presence of 100 M epinephrine (Epi-stimulated) for 45 min. The IP accumulation is expressed as the percentage increase in IP levels above those of mock transfected cells. Results for receptor expression and IP accumulation are the mean Ϯ S.E. of at least three independent experiments. The IC 50 for epinephrine was assessed in competition binding experiments using 80 pm of [ 125 I]HEAT. The IC 50 values are from thirty and three independent experiments for the wild type and mutated receptors, respectively. The EC 50 values are from fifteen and two independent experiments for the wild type and mutated receptors, respectively.
␣ 1b -AR based on its sequence alignment with bovine rhodopsin. The mutated amino acids are colored according to the functional effects induced upon their mutation.

Mutagenesis of Basic Residues in the i1
Loop of the ␣ 1b -AR-The first intracellular loop of the ␣ 1b -AR as with most GPCRs is short, being predicted to consist of just six amino acids (Fig.  1). Within this region there are three basic residues (Arg 74 , His 75 , and Arg 77 ) forming a BBXXB/BBXB motif that has been described as important in the coupling of some receptors to G proteins. The individual mutations of Arg 74 , His 75 , and Arg 77 into Glu did not result in any significant change in the ligand binding properties of the receptors or in their ability to mediate epinephrine-induced IP accumulation (Table I).
To investigate whether the loss of more than a single basic residue had a greater effect than the single mutations, we generated the double mutants R74E/H75E and R74E/R77E. Both mutants displayed decreased levels of expression. However, their ability to mediate an agonist-induced IP response did not significantly differ from that of the wild type ␣ 1b -AR expressed at similar levels (Table I). It may therefore be concluded that the basic residues forming the BBXB motif in the i1 loop of the ␣ 1b -AR do not play a significant role in receptor-G protein coupling. These findings are in agreement with those from other studies on various GPCRs indicating that amino acids in the i1 loop are not important (22,23) or only play a modest role (24,25) in receptor-G protein coupling.
Mutagenesis of Basic Residues in the i2 Loop and Cytosolic Extension of Helix 4 of the ␣ 1b -AR-The 13 amino acids that constitute the i2 loop of the ␣ 1b -AR and the cytosolic extension of helix 4 contain four cationic amino acids, Arg 148 , Arg 159 , Arg 160 , and Lys 161 . Within this region is found the DRYXX(V/ I)XXXL motif identified as a common feature in the rhodopsin family of GPCRs and an essential part of the receptor activation mechanism (5,6). Mutations of the four positively charged residues, Arg 148 , Arg 159 , Arg 160 , and Lys 161 , did not result in any change in the ligand binding properties of the receptor mutants (Table I). Only the triple mutant R148E/R159E/ R160E displayed a 9-fold increase in affinity for epinephrine (Table I).
Significantly, all the mutations resulted in an increased maximal epinephrine-stimulated activity of the receptor. However, the EC 50 values of epinephrine for all the receptor mutants were similar to that of the wild type ␣ 1b -AR (Table I). The R148E mutant also displayed a significant 6-fold increase in its constitutive activity (Table I). Interestingly, when the sequence of the ␣ 1b -AR is aligned with those of the muscarinic M1, M3, and M5 receptors a homologous arginine is similarly located. Mutation of this arginine in the M5 muscarinic receptor to either Asp or Glu also produced constitutive activity (26).
The mutations of Arg 148 and Arg 159 to Ala resulted in a 2-fold increase in the efficacy of epinephrine, whereas the mutation of Arg 160 into Ala did not (results not shown). This suggests that the effect on agonist efficacy is linked to the loss of the positive charge at positions 148 and 159, rather than to the introduction of the anion. In contrast, at position 160 the introduction of the anion is responsible for the effects seen, rather than the loss of the positive charge.
To further explore the respective role of these residues, the mutations R148E, R159E, and R160E were combined. The triple mutant R148E/R159E/R160E, despite being expressed at a lower level as compared with the single mutants, displayed both a significantly increased constitutive and epinephrinestimulated activity ( Table I).
Mutagenesis of Basic Residues in the i3 Loop and Cytosolic Extensions of Helices 5 and 6 of the ␣ 1b -AR-The region including the i3 loop and the cytosolic extensions of helices 5 and 6 of the ␣ 1b -AR is rich in basic amino acids, containing 16 Arg and Lys residues (Fig. 1). Herein we have generated point mutations of each of these residues, and the mutations fall into three groups (Table I).
The first group contains the 6 basic residues, Lys 235 , Lys 243 , Lys 269 , Lys 282 , Arg 288 , and Lys 294 that, when mutated into Glu, did not significantly change the ligand binding or G proteincoupling properties of the receptor (Table I). We therefore did not perform other mutations at these positions. The second group concerns the basic amino acids Lys 231 , Arg 232 , Lys 249 , Lys 271 , Arg 276 , Lys 285 , and Lys 290 . Mutation of all these residues into Glu significantly increased the efficacy of epinephrine without any change in its binding affinity (Table I). Interestingly, the mutation of Lys 231 also resulted in a 3.5-fold increase in constitutive activity.
The third group of mutations concern the three residues, Arg 254 , Lys 258 , and Lys 291 , whose mutations resulted in a significant impairment of the receptor-mediated IP response without changing the ligand binding properties of the receptor ( Table I). Mutations of Arg 254 and Lys 258 into Glu resulted in a 65% and a 45% decrease in epinephrine-stimulated IP response, respectively ( Fig. 2A). When these two mutations were combined to make the R254E/K258E mutant, the receptormediated IP response was almost completely abolished, suggesting an important role for these two residues in receptor G protein coupling (Table I and Fig. 2A).
To determine whether the effects observed were due to either the loss of the positive charge or to the introduction of the anions, Arg 254 and Lys 258 were also mutated into alanine, both individually and in combination. Although the maximal ago-FIG. 1. Topographical representation of the ␣ 1b -AR. The sequence of the hamster ␣ 1b -AR is topographically arranged according to its alignment with bovine rhodopsin (see supplementary material). The amino acid residues mutated in this study are circled in boldface and some key amino acids are numbered. The background color for the mutated amino acids depicts the effect of the mutations at that residue with white representing no effect, green being constitutively activating, yellow increased efficacy for epinephrine, red impaired receptor-mediated signaling, and violet being either impaired signaling or constitutively activating depending upon the substituent amino acid.
nist-stimulated IP response of the R254A mutant remained impaired, that of the K258A mutant was similar to the wild type ␣1b-AR (Table I). However, when the two mutations were combined to make the R254A/K258A mutant, the maximal epinephrine-stimulated activity was profoundly impaired (75% lower than the WT average), suggesting that the integrity of both Arg 254 and Lys 258 is important for receptor-mediated signaling (Table I). It is noteworthy that combining the mutations of Arg 254 and Lys 258 (whether replaced by Ala or Glu) had a greater effect than either of the two individual mutations. These findings suggest that, due to their relative locations, Arg 254 and Lys 258 may partially substitute for each other.
We made the hypothesis that, if the integrity of both Arg 254 and Lys 258 was essential for receptor function, then their mutation should abolish both the constitutive and agonist-induced activity of the constitutively active mutants (CAMs) D142A and A293E previously described (5,27). Thus, the double mutation of Arg 254 and Lys 258 into Glu was combined with the constitutively active D142A and A293E mutants to generate the triple mutants R254E/K258E/D142A and R254E/K258E/ A293E, respectively. In agreement with our hypothesis, the mutation of Arg 254 and Lys 258 in the context of the CAMs abolished both the constitutive and the epinephrine-stimulated activities of the receptor ( Fig. 2A and Table I). The binding affinity of epinephrine for the triple mutants was increased by about 30-fold as for the constitutively active receptors D142A and A293E (5,27). However, although the triple mutants maintained the ligand binding properties of the CAMs, they had lost their signaling ability.
The third basic residue found to be important for receptormediated signaling was Lys 291 found at the cytosolic extension of helix 6 (Fig. 1). The mutation of Lys 291 into Ala and Glu resulted in 65 and 75% decrease in epinephrine-stimulated IP production, respectively, without any significant effect on the ligand binding properties of the receptor (Table I). The EC 50 value for epinephrine was not measured for the K291E mutant due to its low activity.
Because Arg 288 and Lys 291 are predicted in our receptor models to be one helix turn apart on the cytosolic extension of helix 6 ( Fig. 1), we combined the mutations of these two residues so as to assess the effect of a greater loss of cationic charge at this location. The R288A/K291A and R288E/K291E mutants each displayed properties similar to those of the single mutants K291A and K291E, respectively ( Fig. 2B and Table I).
To further assess the importance of Lys 291 in receptor function, the double mutation R288E/K291E was combined with the constitutively active D142A and A293E mutants to generate the triple mutants R288E/K291E/D142A and R288E/ K291E/A293E, respectively. Interestingly, the triple mutants did not display a significantly altered constitutive activity as compared with the wild type ␣1b-AR. However, the R288E/ K291E/D142A receptor displayed some epinephrine-stimulated activity whereas the R288E/K291E/A293E mutant was as good as the wild type receptor in its response to epinephrine ( Fig. 1B and Table I). Both triple mutants displayed a 30-fold increase in binding affinity for epinephrine as found for the constitutively active receptors D142A and A293E (5,27). The double mutants K291E/D142A and K291E/A293E displayed ligand binding and G protein-coupling properties similar to those of the triple mutants R288E/K291E/D142A and R288/K291E/ A293E, respectively (results not shown).
In conclusion, these results suggest that Lys 291 plays an important role in receptor function. However, the finding that its mutation does not completely abolish the agonistinduced activity of the CAMs suggests that its integrity is not essential.
Mutagenesis of Conserved Hydrophobic Residues of the ␣ 1b -AR-We have previously investigated the role of Asp 142 and Arg 143 belonging to the DRYXX(V/I)XXXL motif that has been identified as an essential part of the activation mechanism in GPCRs. Herein we have mutated the other conserved residues of this motif, Tyr 144 , Val 147 , and Leu 151 .
The Y144A mutant displayed ligand binding properties similar to those of the wild type ␣ 1b -AR. However, it was characterized by a small but significant increase in its constitutive activity and a 5-fold increase in epinephrine-induced IP response ( Table I).
The replacement of Val 147 by alanine resulted in a marked increase in the constitutive activity as well as in the epinephrine-induced IP response of the receptor. In contrast, the mutation V147E resulted in a complete loss of receptor-mediated signaling. Interestingly, the affinity of both the V147A and V147E mutants for epinephrine was increased by more than 100-fold. The introduction of the V147E mutation into the constitutively active receptor A293E abolished both its constitutive and epinephrine-induced activity ( Fig. 2C and Table I).
Altogether the features of the V147E mutant are similar to those of the previously described R143E mutant that displayed high affinity for epinephrine, despite being completely impaired in its signaling properties (6). Mutations of the homologous valine in other GPCRs has also resulted in a profound impairment of receptor-G protein coupling (26,28), whereas increased constitutive activity induced by its mutation has not been reported to date.
Mutation of Leu 151 in the ␣ 1b -AR into both Ala and Asp resulted in 62 and 83% impairment of epinephrine-induced IP response, respectively, without any significant change in the ligand binding properties of the receptor (Fig. 2C and Table I). Previous studies on other GPCRs have reported that mutations of a conserved leucine homologous to Leu 151 decreased the signaling properties of the receptors (14,15,29). In each case mutation to any amino acid other than one that is large and hydrophobic in character results in significant impairment of the receptor-mediated response.
The results obtained from the double mutants V147E/A293E and L151D/A293E (Fig. 2C and Table I) also support an important role for Val 147 and Leu 151 in receptor function. The introduction of the mutations V147E or L151D into the constitutively active mutant A293E almost entirely abolished the constitutive as well as the agonist-induced activity. However, both double mutants displayed increased binding affinity for epinephrine as seen for the constitutively active receptor A293E.
In conclusion, the high degree of conservation of Val 147 and Leu 151 among the rhodopsin-like GPCRs and the similarity of the effects found when they are mutated in the ␣ 1b -AR and in other receptors, supports the importance of their role in receptor-mediated signaling.
Analysis of the ␣ 1b -AR and Its Mutants by Molecular Modeling-The initial hypothesis, upon which the mutational analysis of the amino acids located in the cytosolic portion of the receptor was based, originated from our previous study on the docking between the ␣ 1b -AR and a modeled G q heterotrimer (8). The model of the ␣ 1b -AR used in those studies was built following an ab initio approach (17). However, because the crystal structure of rhodopsin has recently been determined at 2.8-Å resolution (18) we have also generated a homology model of the ␣ 1b -AR based on the rhodopsin structure. In this work, the experimental data have been interpreted in the context of both the ab initio and the homology models following molecular dynamics (MD) simulations of the majority of the ␣ 1b -AR mutants. FIG. 3. Location of mutated amino acids within the ab initio and homology models of the ␣ 1b -AR. Each figure shows three different views of a "super average" structure for the ab initio (A) and homology (B) models of the wild type ␣ 1b -AR. The models represent an average structure derived from mutant receptors displaying the functional properties of the wild type receptor. In the right hand and lower left views, the receptor is seen from a direction parallel to the membrane surface, whereas in the upper left view the helical bundle is seen from the intracellular side in a direction perpendicular to the membrane surface. In the upper left panel, the intracellular loops are omitted so as to permit the location of amino acids studied on the helical extensions to be clearly visible. The amino acids mutated in this study are represented by spheres centered on the ␤-carbon of the side chain of the amino acid. The effect of the mutations at each residue is depicted by the color of the sphere, with white representing no effect, green being constitutively activating, yellow increased efficacy for epinephrine, red impaired receptor-mediated signaling, and violet being either impairing or constitutively activating depending upon the substituent amino acid. Fig. 3 shows three different views of the models representing the inactive state of the ␣ 1b -AR obtained following either the ab initio (Fig. 3A) or homology (Fig. 3B) modeling approaches. Each model was obtained by averaging the structures of the wild type ␣ 1b -AR and those of all mutants with functional properties equivalent to the wild type receptor. We consider this "super average" structure to be more representative of the inactive state of the receptor than that of the wild type ␣ 1b -AR alone.
The ab initio and homology models each displayed a high degree of similarity in the arrangement of the transmembrane helices. In addition, the i3 loop in both models contains an ␣-helical segment (ranging from Met 242 to Arg 254 ) within its N-terminal half and a less structured C-terminal portion. Most importantly, the i2 and i3 loops in both models contribute to the formation of a cytosolic crevice that potentially represents a site with electrostatic and shape complementarity with G proteins as previously described (8).
However, there are also some important differences between the two models that result in different mechanistic hypotheses for receptor function. In the homology model, helices 2 and 3, respectively, begin three and five amino acids earlier than in the ab initio model. Furthermore, helix 3 in the homology model displays a greater degree of tilt such that its cytosolic extension is closer to helix 5 than in the ab inito model. In addition, the orientation of the cytosolic extension of helix 6 is slightly different in the two models (Fig. 3). As a result, important differences in the amino acids contributing to the helix 3/helix 6 interface as well as in the environment of Arg 143 of the DRY motif are found (Fig. 3).
In particular, Arg 143 in the ab initio model (Fig. 3A) is directed toward helix 2, and we have previously predicted that its interaction with Asp 91 was an important constraint in maintaining the receptor in its inactive state (30). It is worth noting that in the "super average" structure shown in Fig. 3A, the interaction between Arg 143 and Asp 91 is not as apparent as in the original model of the wild type receptor described in our previous studies (17). In fact the Asp 91 to Arg 143 distance in the super average structure is 7.86 Å in comparison to 4.82 Å in the previously described wild type model. This suggests that the Arg 143 to Asp 91 interaction in the ab initio model is not a feature common to all the inactive conformations of the receptor as would have been expected when looking at the model of the wild type ␣ 1b -AR alone.
In the homology model, Asp 91 in helix 2 is not in the proximity of Arg 143 . Instead, Arg 143 makes a salt bridge with both the adjacent Asp 142 and Glu 289 on helix 6 (Fig. 3B). The latter interaction introduces a link between helices 3 and 6 that potentially represents an important constraint keeping the ␣ 1b -AR in the inactive state. This hypothesis seems to be supported by our preliminary findings, which show that mutations of Glu 289 in helix 6 can increase the constitutive activity of the receptor (results not shown).
We cannot say at present which receptor model is more representative of the actual structure of the ␣ 1b -AR. Although the arrangement of the seven helices in the homology model may be more reliable, the lack of sequence and length similarity in the extra-and intracellular portions of ␣ 1b -AR and rhodopsin does not favor the homology model in preference to the ab initio model. In this study we have therefore used both models of the ␣ 1b -AR to interpret the results of the mutagenesis experiments. The modeling analysis was mainly used to locate the mutated residues in the inactive state of the ␣ 1b -AR and to highlight potential relationships between the effect of the mutations on receptor function and their position in the receptor structure.
Molecular dynamics analysis of the basic amino acids in the three intracellular loops revealed a number of structural features consistent with the experimental findings. In both models the majority of the basic amino acids of the i1 (Arg 74 and Arg 77 ) and i2 loops (Arg 159 and Arg 160 ) as well as of the cytosolic extension of helix 4 (Lys 161 ), lie at the putative lipid/water interface (Fig. 1). Modeling the mutations of these residues does not predict dramatic functional changes for the receptor in agreement with the experimental effects observed for these mutations However, the mutations of Arg 159 , Arg 160 , and Lys 161 resulted in a significant increase in the efficacy for epinephrine. This effect may be in part due to the influence of these mutations on the pK a of Asp 142 of the (E/D)RY motif that is in their proximity. Mutation of Arg 159 , Arg 160 , or Lys 161 to Glu could potentially increase the pK a of Asp 142 , thereby favoring its protonation. Computer simulations (5), as well as experimental studies on rhodopsin (31), have suggested that protonation of this residue is one of the key events in the activation process of these receptors.
The majority of the basic amino acids in the i3 loop are not directed toward the solvent-accessible crevice formed between the i2 and i3 loops but instead are involved in intra-loop interactions. In particular, most of the cationic amino acids of the i3 loop that when mutated did not alter receptor function (Lys 235 , Lys 243 , Lys 269 , Lys 282 ) or increased the efficacy of the agonist (Lys 249 and Lys 271 ), were found to form salt bridges with anionic amino acids within the loop. Thus, both the ab initio and homology models suggest that the main role of the majority of the cationic amino acids in the i3 loop is to stabilize its structure.
Interestingly, the few basic amino acids in the i3 loop that are directed toward the cytosolic crevice of the receptor include Arg 254 and Lys 258 , both of which are fully exposed to the solvent in both models (Fig. 3). The potential key position of Arg 254 and Lys 258 has been previously highlighted by the results of our study on docking between the ␣ 1b -AR and G q (8). Docking solutions between active forms of the ␣ 1b -AR and a G q heterotrimer identified a number of cationic residues (Arg 148 , Arg 160 , Arg 232 , Arg 243 , Arg 254 , Lys 258 , Lys 282 , and Arg 288 ) on the cytosolic surface of the receptor as being available to make contact with anionic amino acids in the ␣ q subunit. Interestingly, Arg 254 and Lys 258 were among the few residues shared by all the docking solutions proposed.
The position of residues whose mutations resulted in important functional effects, either constitutively activating the receptor (Tyr 144 , Arg 148 ), impairing receptor-G protein coupling (Lys 291 ) or both (Val 147 ), have given insight into their structural functional role. In particular, both models demonstrate that the receptor sites susceptible to activating mutations (Asp 142 , Tyr 144 , Val 147 , Arg 148 , Glu 289 , and Arg 293 ) reported in this and previous studies (5,27) belong to, or are close by, the helix 3/helix 6 interface. As a result, all are in close proximity of the highly conserved arginine, Arg 143 , of the DRY motif (Fig.  3). In both models, these constitutively activating mutations all perturb the intramolecular interactions involving Arg 143 , which represents an important constraint that stabilizes the inactive state (R) of the ␣ 1b -AR (results not shown). These findings strongly support a rearrangement between helix 3 and helix 6 as a fundamental step in the activation process of GPCRs (32,33).
The amino acid Lys 291 at the cytosolic extension of helix 6 belongs to the helix 3/helix 6 interface in the ab initio model, whereas it is directed toward the outer face of helix 7 in the homology model (Fig. 3). Despite these topographical differences, the lack of solvent accessibility of Lys 291 in the ab initio model and its orientation in the homology model appear to exclude that this residue is directly involved in G protein interaction and/or activation. Mutations of Lys 291 therefore probably indirectly impair receptor-G protein coupling by inducing a structural perturbation in the helix 6/helix 3 or helix 6/helix 7 packing. The experimental results support the hypothesis that the integrity of Lys 291 is important for productive receptor-mediated signaling, but it is not an essential mediator of receptor-G protein coupling. In fact, mutations of Lys 291 impaired the agonist-induced response of the wild type ␣ 1b -AR but not that of the constitutively active D142A and A293E mutants (Fig. 2B).
Mutations of Val 147 at the cytosolic extension of helix 3 and of Leu 151 in the i2 loop resulted in marked effects on receptor function. Both models are consistent with mutations of Val 147 introducing structural perturbations in the helix 3/helix 6 packing. According to the ab initio model, the V147A mutation that results in an increased constitutive activity of the receptor changes the interaction pattern of Arg 143 and increases the solvent accessibility of the cytosolic crevice between the i2 and i3 loops, which is a feature of the active receptor state (R*). In contrast, the mutation V147E introduces a link between the replacing glutamate and Lys 291 on helix 6. Consistent with this additional constraint between helices 3 and 6, this mutation almost completely abolished the receptor-mediated signaling response. The homology model suggests that the activating mutation of Val 147 into Ala destabilizes the interaction found in the inactive state of the wild type receptor between Arg 143 and Glu 289 . In contrast, the V147E mutation reinforces the link between the two helices, by introducing a new inter-helical interaction between the replacing glutamate and Lys-290 on helix 6. In addition, the replacing glutamate at position 147 can also interact with the neighboring Arg 143 . Thus, both models highlight Val 147 as a crucial residue by its close proximity to Arg 143 , and its inactivating mutation V147E introduces additional constraints into the receptor. We therefore propose that the integrity of Val 147 is important because a supports the mechanistic role of Arg 143 in the activation process of the ␣ 1b -AR.
According to the ab initio model, Leu 151 in the i2 loop is directed toward helix 6 and buried with respect to the cytosol. Mutation of Leu 151 to Asp is predicted to trigger the formation of a salt bridge across the core of the helical bundle between the replacing aspartate and Arg 288 on helix 6. This new constraint may be expected to impair receptor function. In contrast, Leu 151 in the homology model is accessible to the cytosol. Its mutation to Asp may introduce an intra-loop salt bridge with Arg 148 , thereby changing the conformation and solvent accessibility of the loop. Thus, the homology model suggests that Leu 151 may either be important for maintaining the conformation and orientation of the i2 loop and/or play a direct role in receptor-G protein coupling. DISCUSSION In this study we have applied a systematic mutagenesis approach upon the ␣ 1b -AR to investigate the role in receptor-G protein coupling of all the basic amino acids as well as of some conserved hydrophobic residues located in the cytosolic portion of the receptor. Interpreting the effects of the mutations in conjunction with results from molecular modeling analysis has provided some insight on the structure function role of several amino acids.
The ␣ 1b -AR-mediated IP response in cells can be mediated by different members of the G q family (34). However, the results of this study have only been interpreted in the context of the ␣ 1b -AR coupling to G q , because our previous modeling study investigated the docking of a ␣ 1b -AR model with a modeled G q heterotrimer (8).
The Majority of the Cationic Amino Acids in the Intracellular Loops Are Not Directly Involved in Receptor-G Protein Coupling-An important finding of this study is that mutating the majority of basic residues in the cytosolic loops and extensions of helices 4, 5, and 6 of the ␣ 1b -AR did not impair the receptormediated IP response. This result is in good agreement with the predictions made in our previous modeling work docking the ␣ 1b -AR with G q (8). In that study we suggested that, although the majority of cationic residues on the cytosolic surface of the ␣ 1b -AR contribute to reciprocal electrostatic properties between the receptor and the ␣ q subunit, only a selected number of cationic residues could be contact sites on the receptor for the G protein.
It is noteworthy that several mutations in the i3 loop of the ␣ 1b -AR (colored yellow in Figs. 1 and 3) increased the agonistinduced response of the receptor. This in conjunction with our molecular modeling studies suggests that the i3 loop is highly constrained by a number of intramolecular interactions, thereby limiting its propensity to interact with and/or activate the G protein. Mutations of these residues would reduce these constraints leading to increased receptor-G protein coupling.
Arg 254 and Lys 258 in the i3 Loop and Leu 151 in the i2 Loop Are Directly Involved in Receptor-G Protein Coupling-We have found that, among all the basic residues of the cytosolic surface of the receptor, only the combined mutations of Arg 254 and Lys 258 in the i3 loop almost totally impaired the IP response of the receptor as well as of the constitutively active mutants D142A and A293E.
In both the ab initio and homology models of the ␣ 1b -AR, Arg 254 and Lys 258 are the only cationic amino acids in the i3 loop that are solvent-accessible and directed toward the cytosolic crevice of the receptor. Thus, we hypothesize that Arg 254 and Lys 258 are among the contact sites on the receptor for the ␣ q subunit and may therefore play a direct role in receptor-G protein coupling. This is also supported by the results of a previous study showing that Arg 254 and Lys 258 belonged to the only stretch of residues identified in the ␣ 1b -AR that could confer to ␤ 2 -AR the ability to activate the G q /PLC pathway (19). It will be interesting to assess whether mutations of the amino acids predicted to be the partners to Arg 254 and Lys 258 in the ␣ q subunit also impair receptor G protein coupling.
Our experiments, however, do not directly assess the functional role played by Arg 254 and Lys 258 at a mechanistic level. The impairment of the ␣ 1b -AR-mediated IP response induced by mutations of these residues could result from their effect on any of the steps leading to a receptor-mediated response, i.e. receptor activation, receptor-G protein interaction or receptorinduced G protein activation. Arg 254 and Lys 258 are located far from the transmembrane helical bundle and from the membrane/cytosol boundary where constitutively activating mutations have principally been found in GPCRs (3). This location appears to exclude Arg 254 and Lys 258 from being involved in the process of receptor activation, i.e. the transition of the receptor from its inactive (R) to active (R*) state.
For most GPCRs, receptor-G protein interaction cannot be conclusively distinguished from receptor-mediated G protein activation at the experimental level. The guanine nucleotidesensitive high affinity binding of agonists to the ␤ 2 -AR has been interpreted as a measure of its physical interaction with G s and some of its mutants (35). Unfortunately, this experimental tool cannot be applied to the ␣ 1b -AR which, like other GPCRs coupled to the G q /PLC pathway, displays monophasic binding isotherms for agonists that are insensitive to GTP analogues. We have also explored the possibility of assaying the ability of different receptor mutants to co-immunoprecipitate the ␣ q sub-unit as a tool to potentially measure the receptor-G protein physical interaction in COS-7 cells. However, the inability to detect a significant agonist-dependent regulation of the ␣ 1b -AR-␣ q subunit association (results not shown), combined with the difficulty in finding a convincing theoretical interpretation of the co-immunoprecipitation experiments, discouraged us from further pursuit of this assay.
In conclusion, despite the lack of direct evidence, both the functional analysis of the receptor mutants and the predictions of molecular modeling support the hypothesis that Arg 254 and Lys 258 are directly involved in the receptor-mediated activation of the G protein.
The experimental results indicate that Leu 151 also plays an important role in receptor-mediated signaling. Despite the fact that the ab initio and homology models of the ␣ 1b -AR do not provide a consistent interpretation of the structural functional role of Leu 151 , the fact that its mutation into Asp almost completely abolished the IP response mediated by the wild type ␣ 1b -AR as well as by the constitutively active mutant A293E suggests that this residue may be part of the receptor-G protein interface. The high degree of conservation of this residue, in conjunction with the similar effects found upon its mutation in other receptors, suggests that its role in receptor-G protein coupling is conserved among different GPCRs.
The Cytosolic Extensions of Helices 3 and 6 Play a Crucial Role in Receptor Activation-Both the ab initio and homology models of the ␣ 1b -AR predict that the majority of amino acids susceptible to activating mutations identified in this study (Tyr 144 , Val 147 , Arg 148 , and Glu 289 ) or in our previous work (Asp 142 (5) and Arg 293 (27)) belong to or are close to the interface between the cytosolic extensions of helices 3 and 6 ( Fig. 3). Altogether, our results suggest that the residues located in the environment of the interface between the cytosolic extensions of helices 3 and 6 are mainly involved in the activation process of the ␣ 1b -AR, i.e. its transition between the inactive (R) and active (R*) states. This is in agreement with the conclusions of other studies on rhodopsin (36,37) and ␤ 2 -AR (38), suggesting that a rearrangement in the relative positioning of helices 3 and 6 is a fundamental step in receptor activation.
The findings of this study exclude the basic amino acids in the cytosolic extension of helix 6 of the ␣ 1b -AR from playing a direct role in receptor-G protein coupling. In fact, mutating most of these residues (Arg 288 , Lys 290 , and Lys 294 ) had no significant effect on receptor function. Only mutations of Lys 291 profoundly impaired the ␣ 1b -AR-mediated IP response. The functional as well as modeling analysis of these mutations suggest that Lys 291 is not a contact site on the receptor for the G protein, but rather it plays a structural role in helix 3/helix 6 or helix 6/helix 7 packing, thereby allowing productive receptor-G protein coupling.
Previous studies on muscarinic cholinergic receptors have provided evidence that it is mainly hydrophobic residues in the cytosolic extension of helix 6 that dictate receptor-G protein coupling selectivity (reviewed in Ref. 2). Future mutagenesis studies of the ␣ 1b -AR targeting other amino acids that have not been considered in this study will further investigate the functional role of this portion of the receptor.
Conclusions-The findings of this study significantly improve our knowledge of the molecular determinants of the ␣ 1b -AR/G q protein interface. The role of cationic as well as hydrophobic residues in receptor-G protein coupling has been suggested for several GPCRs (2). However, a systematic mutational analysis of these residues has been lacking. It is therefore difficult to build a complete map of the different amino acids involved in receptor-G protein coupling for various receptors and to compare the positions of the basic residues found functionally important among different receptors. Our findings do not support the hypothesis that simple motifs like the BBXXB or BBXB sequences found in different cytosolic loops of GPCRs can predict receptor-G protein coupling. In fact, among the three BBXXB or BBXB motifs found in the ␣ 1b -AR, in the i1 loop, the N-terminal portion of the i3 loop, and in the cytosolic extension of helix 6, only mutations of a single residue (Lys 291 ) in the later motif ( 290 KKAAK 294 ) impaired the receptor-mediated response. This suggests that the effect of this mutation is linked to an important structural functional role of Lys 291 rather than to the disruption of the motif.
The results of the mutagenic analysis of the ␣ 1b -AR are in agreement with the conclusions of docking simulations between the ␣ 1b -AR and G q models (8). It is noteworthy that the complementary areas of charge and shape driving the docking between the ␣ 1b -AR and G q display similarities with that recently described between the rhodopsin structure and transducin (39). The involvement of the i2 and i3 loops of rhodopsin in G protein interaction was recently demonstrated by elegant studies, in which different sites in these loops were cross-linked to transducin (40,41).
Despite the large number of experimental studies on GPCRs, our knowledge on how agonist binding to receptors results in G protein activation still remains unclear. The crystal structure of rhodopsin in its ground state has represented a significant breakthrough in GPCR research (18). However, a better understanding of how the active conformation of rhodopsin interacts with and activates transducin awaits the resolution of the active structure of the receptor. New structural approaches like those recently described by the group of Khorana (40 -42) will be extremely useful in elucidating the architecture of the receptor-G protein interface. However, the results from systematic mutational analysis of different receptors, such as those presented herein, represent an important step in determining the role of individual amino acids in GPCR function.