Novel Mutants of the Human β1-Adrenergic Receptor Reveal Amino Acids Relevant for Receptor Activation*

Activation of G protein-coupled receptors like the β1-adrenergic receptor results in conformational changes that ultimately lead to signal propagation through a G protein to an effector like adenylyl cyclase. In this study we identified amino acids that seem to be critical for activation of the human β1-adrenergic receptor. Activation patterns of mutant receptors were analyzed using two structurally different ligands for β-adrenergic receptors that both are mixed agonist/antagonists. Broxaterol and terbutaline are agonists at β2- and β3-receptors; however, they act as antagonists at the β1-subtype. We reasoned that this functional selectivity may be reflected by a corresponding sequence pattern in the receptor subtypes. Therefore, we exchanged single amino acids of the β1-adrenergic receptor for residues that were identical in the β2- and β3-subtypes but different in the β1-receptor. Pharmacological characterization of such receptor mutants revealed that binding of a panel of agonists and antagonists including broxaterol and terbutaline was unaltered. However, two of the mutants (I185V and D212N) were activated by broxaterol and terbutaline, which acted as antagonists at the wild-type receptor. Two additional mutants (V120L and K253R) could be activated by terbutaline alone, which is structurally more closely related to endogenous catecholamines like epinephrine than to broxaterol. A model of the human β1-adrenergic receptor showed that the four gain-of-function mutations are outside of the putative ligand-binding domain substantiating the lack of an effect of the mutations on binding characteristics. These results support the notion that Val-120, Ile-185, Asp-212, and Lys-253 are critically involved in conformational changes occurring during receptor activation.

G protein-coupled receptors (GPCRs) 3 comprise a large number of structurally related membrane receptors, many of which are important drug targets. Among the receptors targeted in established therapies ␤-adrenergic receptors comprise one of the most important subgroups. The use of ␤-blockers is indicated in virtually all major cardiovascular diseases, whereas ␤ 2 -selective agonists are a mainstay in the treatment of asthma. Activation of GPCRs is a complex process that results in a state that propagates a signal to the corresponding G protein. It is thought that more than one activated state may exist (1, 2) providing a basis for transduction of different ligand-specific signals via a given receptor subtype (3). The notion that receptor activation is a multistep process whereby the receptor assumes a series of conformational intermediates provides an explanation for such functionally distinct states (4). Such functionally defined states could serve as distinct mediators of more specific drug actions.
The investigation of conformational intermediates should benefit from ligands that are able to distinguish their functional identity. Alternatively it would be interesting to study distinct activation patterns at closely related receptor subtypes that might be different for a given ligand. Recently we have discovered that some ligands exhibit distinct functional properties at the three subtypes of ␤-adrenergic receptors. In the course of a previous project for the development of subtype-selective ligands for human ␤-adrenergic receptors we noticed that the experimental compound broxaterol, which is thought to be a ␤ 2 -selective agonist, binds in fact with the same affinity to all three ␤-receptor subtypes (5,6). Interestingly this compound turned out to be an agonist at ␤ 2 -and ␤ 3 -receptors but an antagonist at the ␤ 1 -receptor (5, 6) making it a ␤ 2 -/␤ 3 -selective agonist in functional terms, although it does not bind selectively to these subtypes. In a recent study we found similar functional ␤ 2 selectivity for clinically used compounds including terbutaline and salbutamol (6). This discovery prompted us to undertake the present investigation where we generated several gain-of-function mutants of the ␤ 1 -adrenergic receptors that are activated by compounds like broxaterol and terbutaline. Based on the close sequence similarity between ␤-adrenergic receptor subtypes we reasoned that the functional selectivity of such compounds may have a corresponding sequence pattern in amino acids that are identical in ␤ 2 -and ␤ 3 -but different in the ␤ 1 -subtype. Sequence comparison of the entire receptor proteins revealed a total of 17 amino acids following this pattern with eight of these positions being located in transmembrane domains. We generated six mutants where a ␤ 1 amino acid was substituted for the corresponding ␤ 2 -/␤ 3 amino acid, and the functional characteristics of these mutants were analyzed. For mutation of such positions functionally different regions were selected. As the putative ligand-binding domain is embedded in the transmembrane domains (7,8) several mutations were introduced in transmembrane regions. The two mutants L154V and K253R are at the interface between transmembrane domains and intracellular loops two and three, respectively, which are thought to be involved in receptor-G protein coupling (9). Mutant D212N is located in the second extracellular (E2) loop in a position that has been shown to be relevant for ligand-receptor interaction in other GPCRs (10 -13). Fourgain-of-functionmutantswerefoundthatshoweddistinctliganddependent changes of activation patterns. Our data including a receptor model suggest that we have identified amino acid residues outside the ligand-binding domain that are critical for the activation of human ␤ 1 -adrenergic receptors.
Mutagenesis and Cell Transfection-The cDNA encoding for human ␤ 1 -adrenergic receptor (15) (GenBank TM accession number J03019) was mutated by a PCR-mediated mutagenesis technique using VENT DNA polymerase (New England Biolabs). After confirmation of the mutation, the PCR products were digested with the appropriate enzymes and cloned into the expression vector pcDNA3 containing the wild-type ␤ 1 -adrenergic receptor cDNA to obtain full-length mutated ␤ 1 -adrenergic receptors. The sequences of all resultant cDNAs were verified by automated sequencing. The cDNA of the ␤ 2 -adrenergic receptor (16) (GenBank TM accession number Y00106) was cloned into the pcDNA3 expression vector as described earlier (6).
Cell Culture and Membrane Preparation-Chinese hamster ovary cells stably transfected with human ␤-adrenergic receptor subtypes and different mutants of the ␤ 1 -adrenergic receptor, respectively, were grown and split in Dulbecco's modified Eagle's medium with nutrient mixture F12 as described before (6). Before cells were harvested the culture medium was removed, and cells were washed twice with phosphate-buffered saline. Then membranes were prepared, or cells were frozen on the dishes for later preparation of membranes. Crude membrane fractions were prepared from fresh (measurement of adenylyl cyclase) or frozen cells (radioligand binding) according to two different protocols, which have been described recently (14). The resulting membrane pellets were resuspended in 50 mM Tris/HCl, pH 7.4, at a final protein concentration of 1-2 mg/ml. Protein concentration was determined by the method of Bradford (17) with bovine serum albumin (Sigma) as a standard.
Adenylyl Cyclase Activity and Radioligand Binding Studies-Determination of adenylyl cyclase activity in cell membranes was based on the method originally described by Jakobs et al. (18); for details see Ref. 6. Accumulation of [␣-32 P]cAMP was linear over at least 20 min under all conditions. The basal and isoproterenol-stimulated adenylyl cyclase activity, respectively, in the different cell clones characterized in Table 3 was (all values in pmol/mg of membrane protein/min): 13.5 Ϯ 1. 26  The radioligand binding experiments were performed with membranes prepared as described above and followed the procedure as outlined previously (6). For competition binding ϳ50 pM [ 125 I]CYP was used, and assays were done in the presence of 100 M GTP to ensure monophasic competition curves for agonists. Membranes with ␤ 1 wildtype and mutant receptors used in competition binding experiments showed comparable receptor expression (Table 1). K D values for [ 125 I]CYP and K i values from competition experiments were calculated from saturation experiments by nonlinear curve fitting with the program SCTFIT (19).
Receptor Homology Modeling-The alignment of the primary sequences (Swiss-Prot accession numbers P02699 (rhodopsin) and P08588 (␤ 1adrenergic receptor)) was generated with the program ClustalW. To ensure that residues in corresponding secondary structure elements were properly classified, the aligned structures were examined using the program Jpred (20). For every receptor 15 regions were identified according to the topology of rhodopsin (21,22): seven transmembrane helices, three cytoplasmic loops, three extracellular loops, the extracellular N terminus, and the cytoplasmic C terminus. The amino acid sequences of these areas flanked by four additional amino acids were submitted to the WU-Blast server to search in the Research Collaboratory for Structural Bioinformatics Protein Data Bank (23) for additional fragments with a known three-dimensional structure. The structurebased multiple sequence alignments of all the fragments plus the rhodopsin data were merged as input for the homology modeling to obtain three-dimensional models of the ␤ 1 -adrenergic receptor (24 -26). The program MODELLER (24 -26) was used to generate the three-dimensional structures of ␤ 1 -receptor, and during this process the models were refined by molecular dynamics routines with incremental increases in simulation temperature from 150 to 1000 K followed by incremental temperature decreases from 1000 to 300 K. Several slightly different models were calculated by varying the initial parameters, and the variability and energy among these models was used to estimate the lower limit of the error in the corresponding regions of the fold. The models obtained were submitted to the PROCHECK package for the structure evaluation (27).
Ligand Docking-The AutoDock 3.0 program was used to perform an automatic docking exploration for different conformations of the ligand in the ␤ 1 models (28). This automated docking program uses a gridbased method for energy calculation of the flexible ligand in complex with a rigid protein. Points on the three-dimensional grid are placed such that they cover the entire inner cavity of the ␤ 1 -adrenergic receptor and are probed with the atoms that constitute the ligand. The docking experiments explored the interaction of terbutaline with the binding region of ␤ 1 -subtype.
The simulation was carried out within a 22-Å cube using a 0.375-Å grid spacing. AutoDock uses an adaptive global-local search method based on Lamarckian genetics (LGA) in conjunction with an empirical force field that allows the prediction of binding free energies for docked ligand. In our setup we used a starting population of 50 ligand conformations with a stopping criterion of a maximum of 10 million energy evaluations. The number of dockings was set to 250 to get good statistics of the docked complex, and the resulting bound states were clustered in groups on the basis of a root mean square deviation of 0.5 Å relative to the initial starting position of the ligand. These numbers together with the default parameters of AutoDock have been shown to be a useful setting for blind docking (29). From the 250 simulations performed the binding modes with the most populated clusters was selected. Fig. 1 shows a sequence comparison documenting the high sequence identity of the human ␤-adrenergic receptor subtypes in particular in the transmembrane domains. Broxaterol and terbutaline activate only ␤ 2 -and ␤ 3 -receptors as opposed to agonists like isoproterenol and fenoterol, which activate all three subtypes of ␤-receptors. Therefore, amino acids that are shared between the ␤ 2 -and ␤ 3 -adrenergic receptors but are different in the ␤ 1 -subtype were investigated to determine their importance for the specific activation pattern of broxaterol and terbutaline. In Fig. 1 respective amino acids are emphasized with a box. Mutation of such amino acids in the ␤ 1 -receptor to the corresponding amino acid of the ␤ 2 /␤ 3 -subtypes was accomplished by a PCR approach (see "Experimental Procedures"). A selection of the sites following this pattern was chosen for mutation ( Fig. 1, arrows, and Table 1).

RESULTS
The functional characteristics of the mutated receptors were investigated in Chinese hamster ovary cells stably transfected with mutant and wild-type receptors. Fig. 2A shows that broxaterol did not stimulate adenylyl cyclase activity via ␤ 1 -receptors but had partial agonistic activity at the ␤ 2 -subtype as has been shown before (5, 6). The ␤ 1 -receptor mutants V120L, L154V, K253R, and F362L behaved like the wild-type receptor as broxaterol did not mediate a stimulation of adenylyl cyclase. In contrast, broxaterol did stimulate adenylyl cyclase via mutants I185V located in TM4 and D212N in the second extracellular loop (Fig. 2A). The stimulation of adenylyl cyclase activity mediated by these gain-offunction mutants was comparable to the stimulation achieved by the full agonist isoproterenol.
In addition to broxaterol we tested the efficacy of terbutaline, which is another compound that has shown a mixed ␤ 1 -antagonist/␤ 2 -/␤ 3 -agonist profile (6). Fig. 2B confirms the minimal efficacy of terbutaline at the ␤ 1 -adrenergic receptor (19% compared with isoproterenol) and near full efficacy at the ␤ 2 -receptor. Terbutaline behaved similarly to broxaterol at the ␤ 1 -receptor mutants L154V and F362L as it showed an identical minimal functional response at these mutants as well as at the wild-type receptor (Fig. 2B). The mutants I185V and D212N mediated a stimulation of adenylyl cyclase by terbutaline; this is in correspondence to the agonistic activity shown for broxaterol at these mutants. In contrast to the structurally different compound broxaterol, terbutaline activated adenylyl cyclase also via the mutants V120L, which is at the extracellular end of TM2, and K253R, which is located at the cytoplasmic face of TM5 (Fig. 2B). Again the efficacy of terbutaline at these mutants was similar to isoproterenol. All gain-of-function mutants showed the same or a slightly lower basal activity compared with the ␤ 1 wild type (see "Experimental Procedures").
The mutant ␤ 1 -adrenergic receptors were tested for their binding characteristics and expression levels. Table 1 shows that wild-type ␤ 1and ␤ 2 -receptors and all mutant ␤ 1 -receptors bound the nonselective antagonist [ 125 I]CYP with similar K D values. Also the expression levels of the cell lines studied were in the same order of magnitude excluding  that the observed gain of function may be a result of massive overexpression of the respective mutant receptor. As a next step it was confirmed that the mutations leading to activation by broxaterol or terbutaline did not cause a pharmacological conversion of ␤ 1 -into ␤ 2 -(or ␤ 3 -) adrenergic receptors. One of the main characteristics of ␤ 2 -adrenergic receptors is the low affinity for norepinephrine compared with epinephrine, whereas the ␤ 1 -subtype does not distinguish between these two endogenous catecholamines. In contrast, the ␤ 3 -subtype shows a marked preference for norepinephrine over epinephrine. Table 2 shows that all mutants clearly maintained the pharmacological identity of a wild-type ␤ 1 -adrenergic receptor.
In addition, more detailed competition studies were performed with a panel of agonists that confirmed that the mutants share all their pharmacological characteristics with the wild-type ␤ 1 -receptor ( Table 3). The affinity of some prototypical subtype-selective antagonists was also not affected by any of the mutations introduced into the ␤ 1 sequence ( Table 3).
The three-dimensional model of the ␤ 1 -adrenergic receptor in Fig.  3A shows the location of the mutations that changed the functional responses to broxaterol and terbutaline. The model is based on the crystal structure of rhodopsin (21) and was generated as described under "Experimental Procedures." In Fig. 3B some amino acids thought to be involved in agonist recognition are shown. These amino acids were identified previously as critical for agonist binding in models for both ␤ 1 -(30) and ␤ 2 -receptors (7,8). In addition, terbutaline is docked into the putative binding site of the receptor. The docking of terbutaline was simulated with various starting positions, but independent of the starting conditions about a quarter of the obtained structures fitted to the same position (root mean square deviation, 0.5 Å). The model suggests that all mutated amino acids are too far away from the binding pocket to be directly implicated in ligand binding. The closest distance between the ligand and a mutated amino acid in the transmembrane region of 7.7 Å is found between terbutaline and Ile-185.

DISCUSSION
Recently it was found that not only the experimental drug broxaterol but, surprisingly, also compounds used clinically such as terbutaline bind nonselectively to human ␤-receptors. However, they exhibit functional selectivity as they act as agonists only at ␤ 2 -and ␤ 3 -adrenergic receptors, whereas they are antagonists at the ␤ 1 -subtype (5,6). The lack of binding selectivity of such compounds suggests that their interaction with the ligand-binding domain does not provide a sufficient explanation for the observed functional differences at ␤-adrenergic receptor subtypes. There seem to exist common features of the ligand-binding domains that dictate the binding affinity. However, it may be reasonable to assume that receptor activation is additionally controlled by subtypespecific sequence patterns of amino acids outside the ligand-binding domain. Therefore, nonselective ligands that are mixed agonist/ antagonists like broxaterol or terbutaline should be interesting tools to study receptor activation. Based on the high sequence identity between the subtypes of ␤-adrenergic receptors it seemed possible to determine amino acids that are responsible for the subtype-specific activation patterns of broxaterol and terbutaline. Such positions should play a prominent role in receptor activation in general. We reasoned that the observed functional pattern for mixed agonists/ antagonists might be reflected by a corresponding sequence pattern in the three ␤-adrenergic receptor subtypes. The gain-of-function mutants of the ␤ 1 -receptor that were found in the course of this study confirm this initial hypothesis.
Several of the generated mutants showed altered functional characteristics compared with the wild-type ␤ 1 -adrenergic receptor. It can be excluded that these changes are caused by a decrease in structural constraints as all gain-of-function mutants showed the same or a slightly lower basal activity compared with the ␤ 1 wild type. Interestingly we found two ␤ 1 mutants at which broxaterol turned into an agonist, whereas for terbutaline two additional gain-of-function mutations were identified. Broxaterol is an unusual ␤-agonist as it does not share a catechol or related structural motif with the endogenous ligands epinephrine and norepinephrine or most typical ␤-adrenergic agonists. It seems reasonable, therefore, that more gain-of-function mutants were found for terbutaline than for broxaterol as the catechol and related structures are important for receptor activation (31).
One obvious explanation for the functional effect of exchanging amino acids in the ␤ 1 -receptor for ␤ 2 or ␤ 3 residues would be a change of the pharmacological characteristics and a concomitantly altered activation pattern. Our data clearly showed that the mutated ␤ 1 -receptors are pharmacologically absolutely identical with the wild-type receptor. Closer investigation with a number of subtype-selective agonists and  antagonists confirmed that all mutants independent of their activation by broxaterol or terbutaline are pharmacologically indistinguishable from wild-type ␤ 1 -receptors. The receptor model shown in Fig. 3 also suggests that the mutations did not alter the ligand binding pocket. Therefore, our data provide evidence that single amino acids in various positions that do not seem to be involved in ligand recognition are decisive for the agonistic properties of a receptor ligand.
Overall the number of mutations of the ␤ 1 -adrenergic receptor reported in the literature is rather limited compared with numerous mutations of the ␤ 2 -receptor. A number of mutations turning ␤ 1 -receptors into constitutively active receptors have been described (32). A polymorphism of the human ␤ 1 -subtype with functional consequences is known in position 389 (33). It turned out that an Arg in position 389 is more common than the Gly originally identified as the amino acid in the wild-type receptor. The variant with an Arg-389 is functionally different from the wild type as it shows both higher basal and isoproterenol-stimulated adenylyl cyclase activity based on enhanced G s coupling (33). To our knowledge, so far there were no ␤ 1 -receptor mutants known to turn an antagonist into an agonist.
Activation of a GPCR is thought to be associated with conformational changes in the receptor protein. Current concepts suggest that after initial contact between an agonist and the ligand-binding domain a multistep process guides the receptor through a series of conformational intermediates (4,34). Different conformational states may be responsible for the selective activation of signaling cascades mediated through different G proteins as in the case of the ␤ 2 -adrenergic receptor (35). They may also be the basis for functionally distinct responses to different concentrations of a ligand as has been shown for CGP 12177 and other compounds (3). It is clear that any motion in the receptor protein resulting in conformational changes requires the existence of hinges or pivots to allow for structural changes to occur. In addition to amino acids responsible for ligand-receptor interaction other critical amino acids should exist that act as key points for such motion. It is not surprising, therefore, that mutations of amino acids that are not involved in ligand recognition may have effects on receptor function. Although we do not know the exact mechanism that underlies the functional changes caused by mutation it is obvious, however, that residues Val-120, Ile-185, Asp-212, and Lys-253 play a critical role in the activation of the human ␤ 1 -adrenergic receptor.
Isogaya et al. (8) describe a number of mutations including mutation of Val-120, which is one of the positions that led to a gain-of-function mutation in our study. The mutation of Val-120 to Ala shows an indirect contribution to subtype selectivity of selected agonists (8). Although ligand binding was not affected in our study we also found that activation by terbutaline, but not by broxaterol, was changed by mutation of Val-120 to Leu in our case, confirming a significant role of this position.
A number of studies suggest that the E2 loop plays an important role in ligand binding and receptor activation at least in some GPCRs (10 -13). Our data support this notion as the D212N mutant showed a functional change compared with the wild-type receptor as it could be activated by both broxaterol and terbutaline. A number of mutations in the E2 loop of the C5a receptor result in constitutive activity; therefore, a role as a negative regulator of receptor activation has been proposed for the E2 loop (13). Position Asp-212 in the human ␤ 1 -adrenergic receptor does not seem to function in such a way as the mutation to Asn did not result in a change in basal receptor activity. In the case of the dopamine D 2 receptor a contribution of the E2 loop to the ligand binding site has been suggested (12). In particular our binding data led us to conclude that Asp-212 is not directly involved in ligand binding as the D212N mutation like all other mutations presented in this study did not affect agonist or antagonist binding. All functionally changed mutants including the D212N mutation support the concept that amino acids outside the ligand-binding domain contribute relevant structural elements for receptor activation and thus for ligand efficacy. The model of the human ␤ 1 -adrenergic receptor shows that the gainof-function mutants described here are very unlikely to be close enough to the docked ligand terbutaline to directly interfere with ligand-receptor interaction. Further confirmation of this notion comes from the absolutely unaffected pharmacological characteristics of all functionally significant mutants investigated in our study. The mutations affect residues outside the ligand-binding domain and reveal, therefore, that the respective amino acids contribute to the control of the activation process independently of ligand-receptor recognition. It turned out that the ligands broxaterol and terbutaline, which are nonselective in binding but functionally selective, are ideal pharmacological tools to analyze the activation process in conjunction with point mutation and receptor modeling.
In summary, we present data revealing an important role of Val-120, Ile-185, Asp-212, and Lys-253 for the activation of the human ␤ 1 -adrenergic receptor. The mutation of these positions to the corresponding amino acids of the ␤ 2 -/␤ 3 -subtype resulted in a gain of function as the ␤ 1 -antagonists broxaterol and terbutaline turned into agonists at these mutants. We conclude that the mutated positions represent critical residues for conformational changes that occur during receptor activation. This notion is supported by a receptor model suggesting that the amino acids mutated in this study are not involved in direct ligandreceptor recognition.