Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. 2008 Jan 22; 105(3): 877–882.
Published online 2008 Jan 11. doi: 10.1073/pnas.0711253105
PMCID: PMC2242685
PMID: 18192400
Biochemistry

Conformational thermostabilization of the β1-adrenergic receptor in a detergent-resistant form

Maria J. Serrano-Vega, Francesca Magnani, Yoko Shibata, and Christopher G. Tate*
Author information Article notes Copyright and License information Disclaimer

Associated Data

Supplementary Materials

Abstract

There are ≈350 non-odorant G protein-coupled receptors (GPCRs) encoded by the human genome, many of which are predicted to be potential therapeutic targets, but there are only two structures available to represent the whole of the family. We hypothesized that improving the detergent stability of these receptors and simultaneously locking them into one preferred conformation will greatly improve the chances of crystallization. We developed a generic strategy for the isolation of detergent-solubilized thermostable mutants of a GPCR, the β1-adrenergic receptor. The most stable mutant receptor, βAR-m23, contained six point mutations that led to an apparent Tm 21°C higher than the native protein, and, in the presence of bound antagonist, βAR-m23 was as stable as bovine rhodopsin. In addition, βAR-m23 was significantly more stable in a wide range of detergents ideal for crystallization and was preferentially in an antagonist conformation in the absence of ligand.

Keywords: G protein-coupled receptor, membrane protein, stabilization

Over the past 20 years the rate of determination of membrane protein structures has gradually increased, but most success has been in crystallizing membrane proteins from bacteria rather than from eukaryotes (1). Bacterial membrane proteins have been easier to overexpress using standard techniques in Escherichia coli than eukaryotic membrane proteins (2, 3), and the bacterial proteins are often far more stable in detergent, detergent stability being an essential prerequisite to purification and crystallization. However, of the 125 unique membrane protein structures that have been solved to date, there are only eight structures of mammalian integral membrane proteins; five of these membrane proteins were purified from natural sources and are stable in detergent solutions. Apart from the difficulties in overexpressing eukaryotic membrane proteins, they often have poor stability in detergent solutions, which severely restricts the range of crystallization conditions that can be explored without their immediate denaturation or precipitation. Ideally, membrane proteins should be stable for many days in any given detergent solution, but the detergents that are best suited to growing diffraction-quality crystals tend to be the most destabilizing detergents, i.e., those with short aliphatic chains and small or charged head groups. It is also the structures of human membrane proteins that we would like to solve, because these are required to help the development of therapeutic agents by the pharmaceutical industry; often there are substantial differences in the pharmacology of receptors, channels, and transporters from different mammals, whereas yeast and bacterial genomes may not include any homologous genes. There is thus an overwhelming need to develop a generic strategy that will allow the production of detergent-stable eukaryotic integral membrane proteins for crystallization and structure determination.

Membrane proteins have evolved to be sufficiently stable in the membrane to ensure cell viability, but they have not evolved to be stable in detergent solution, suggesting that membrane proteins could be artificially evolved, and detergent-stable mutants could be isolated (4). This was subsequently demonstrated for two bacterial proteins, diacylglycerol kinase (5, 6) and bacteriorhodopsin (7). Random mutagenesis of diacylglycerol kinase identified specific point mutations that increased thermostability, and, when combined, the effect was additive so that the optimally stable mutant had a half-life of 35 min at 80°C compared with a half-life of 6 min at 55°C for the native protein (6). A further study on bacteriorhodopsin by cysteine-scanning mutagenesis along helix B demonstrated that it was not possible to predict which amino acid residues would lead to thermostability upon mutation, nor, when studied in the context of the structure, was it clear why thermostabilization had occurred (7).

We decided to see whether stabilization of a eukaryotic integral membrane protein was possible and whether the effect was sufficiently great that it would significantly improve the chances of obtaining diffraction-quality crystals. The β1-adrenergic receptor (βAR) from turkey erythrocytes (8) was chosen as a test subject for this study for a number of reasons. The βAR is a G protein-coupled receptor (GPCR) that has well developed pharmacology (9), with many ligands commercially available and in a radiolabeled form. In addition, overexpression of βAR has been particularly successful using the baculovirus expression system, and it can be purified in milligram quantities in a functional form (10). There is, however, one other serious problem associated with trying to crystallize GPCRs, which is that the receptors have evolved to cycle through at least two distinct conformations, the agonist-bound form and the antagonist-bound form, and changes between these two conformations can occur spontaneously in the absence of ligand. It is thus likely that any purified receptors populate a mixture of conformations. To improve the likelihood of crystallization further, we also wanted the final optimally stable mutant to be preferentially in the antagonist-bound conformation.

Results

Selection of Single Mutations That Increase the Thermostability of the βAR.

βAR from turkey erythrocytes is an ideal target for structural studies because it is well characterized and is expressed at high levels in insect cells using the baculovirus expression system (11, 12). The best overexpression of βAR is obtained by using a truncated version of the receptor containing residues 34–424 (βAR34–424) (10), and this was used as the starting point for this work. Alanine scanning mutagenesis was used to define amino acid residues in βAR34–424 that, when mutated, altered the thermostability of the receptor; if an alanine was present in the sequence it was mutated to a leucine residue. A total of 318 mutations were made to amino acid residues 37–369, a region that encompasses all seven transmembrane domains and 23 amino acid residues at the C terminus; mutations at 15 amino acid residues were not obtained because of strong secondary structure in the DNA template. After sequencing each mutant to ensure the presence of only the desired mutation, the receptors were functionally expressed in E. coli and assayed for stability.

The assay for thermostability was performed on unpurified detergent-solubilized receptors by heating the receptors at 32°C for 30 min, quenching the reaction on ice, and then performing a radioligand binding assay, using the antagonist [3H]dihydroalprenolol, to determine the number of remaining functional βAR34–424 molecules compared with the unheated control. Heating the unmutated βAR34–424 at 32°C for 30 min before the assay reduced binding to ≈50% of the unheated control [supporting information (SI) Fig. 6], and we have defined 32°C as the apparent Tm for βAR34–424; all of the data for the mutants were normalized by including the unmutated βAR34–424 as a control in every assay performed. In the first round of screening, 18 mutants showed an apparent increase in stability, maintaining >75% of antagonist binding after heating and being expressed in E. coli to at least 50% of the native βAR34–424 levels. In view of the possibility of increasing further the stability of these mutants, each of the 18 residues was mutated to two to five alternative amino acid residues of varying size or charge (Fig. 1). Of these 18 mutants, 12 were not improved by further changes, five had better thermostability if another amino acid was present, and one mutation from the first screen turned out to be a false positive; we estimate that the error in the measurements on the first round was ≈20%, so the false positive arose simply because of measurement inaccuracy. In addition, three residues that were not stabilized upon mutation to alanine (V89, S151, and L221) were mutated to a range of other amino acid residues; the two positions that, when mutated to alanine, did not affect thermostability were also unaffected by other changes. In contrast, V89 showed less thermostability when mutated to alanine, but thermostability increased when it was mutated to Leu. Thus, the initial alanine scanning successfully gave two-thirds of the best amino acid residues of those tested for any given position.

An external file that holds a picture, illustration, etc. Object name is zpq0010889990001.jpg

Amino acid changes in βAR that lead to thermostability. (a) Stability quotient indicates the percent stability of the mutants after heating the sample for 30 min at 32°C. All values are normalized to βAR34–424 (50%, shown as a discontinuous line) to remove any experimental variability between assays. Bars show the stability for each mutant. The letters on the x axis indicate the amino acid present in the mutant; an asterisk indicates that the amino acid residue is identical to the residue at the equivalent position in rhodopsin. The original amino acid and its position in βAR34–424 are indicated below with the equivalent positions in rhodopsin shown in gray; the secondary structural elements in rhodopsin where these residues occur are shown above the bar graph in corresponding colors. H, helix; CL, cytoplasmic loop; EL, extracellular loop. Bars corresponding to the same amino acid in βAR34–424 are in the same color. Errors were calculated from duplicate measurements; the best mutants were subsequently reassayed to determine the Tm for each individual mutation and to give an accurate rank order of stability for each mutant. (b) Side chains in rhodopsin that are at equivalent positions to the thermostable mutations in βAR34–424 are shown based on an alignment among rhodopsin, βAR, neurotensin receptor, and adenosine A2a receptor (data not shown). Side chains in the same transmembrane helix are shown as space-filling models in the same color. The name and position of the amino acid residues are those in rhodopsin, and the equivalent positions in the turkey β1 receptor are shown in a.

The position and environment predicted for each of the 16 amino acid residues that gave the best increases in thermostability when mutated were determined by aligning the βAR sequence with that of rhodopsin, which was the only GPCR structure known at that time (Fig. 1). Fourteen of these residues were predicted to be present in transmembrane α-helices, with five of the residues predicted to be lipid-facing, four being deeply buried; the remainder were predicted to be at the interfaces between the helices. Some of these residues would be expected to interact with each other in the βAR structure, such as the consecutive amino acids G67 and R68 (V63 and Q64 in rhodopsin) or the amino acids within the cluster Y227, R229, V230, and A234 in helix 5 (Y223, Q225, L226, and V230 in rhodopsin). Other amino acid residues that could interact in βAR were Q194A in external loop 2 and D322A in external loop 3 (G182 and P285 in rhodopsin, respectively).

The increase in stability that each individual mutation gave to βAR34–424 was determined by measuring the apparent Tm for each mutant (results not shown); Tm in this context is the temperature that gave a 50% decrease in functional binding after heating the receptor for 30 min. Each mutation increased the Tm of βAR34–424 by 1–3°C, with the exception of Y227A, which increased the Tm by 8°C.

Combining Mutations to Make an Optimally Stable Receptor.

Initially, mutations that improved thermostability that were adjacent to one another in the primary amino sequence of βAR were combined (G67A and R68S; Y227A, R229Q, V230A, and A234L). However, the Tm values (results not shown) were only 1–3°C higher than the Tm for βAR34–424, suggesting that combining mutations adjacent to one another in the primary amino acid sequence does not greatly improve thermostability. Subsequently, mutations predicted to be distant from one another in the structure were combined. PCRs were performed by using various mixes of primers to combine up to five different mutations in a random manner and then tested for thermostability (SI Table 1). The best of these combinations increased the Tm >10°C compared with the Tm of βAR34–424. In some cases there was a clear additive effect on the Tm with the sequential incorporation of individual mutations. The most thermostable mutants obtained, which were still expressed at high levels in E. coli, were m6-10, m7-7, and m10-8. These mutants contained collectively a total of 10 different mutations, with eight mutations occurring in at least two of the mutants. A second round of mutagenesis was performed by using m10-8 as the template and adding or replacing mutations present in m6-10 and m7-7 (Fig. 2); many of these mutations improved the Tm further (SI Table 2). For example, exchanging two mutations in m10-8 to create m18 raised the Tm to 49.6°C, and adding A282L to make m23 increased the Tm a further 3°C to 52.8°C. This produced the most thermostable βAR34–424 mutant so far, and it will be referred to as βAR-m23.

An external file that holds a picture, illustration, etc. Object name is zpq0010889990002.jpg

Evolution of thermostability in βAR. Starting from βAR-m10-8, combinations of mutations were rearranged systematically to find the optimum combination of mutations (see also SI Table 2). The numbers in brackets refer to the apparent Tm (in °C) for each mutant combination. The numbers above or below the arrows indicate the change in individual βAR thermostabilizing mutations to generate the combination indicated. The residues and the mutations involved are the same as those in Fig. 1. Thus, m23 contains the mutations R68S, M90V, Y227A, A282L, F327A, and F338M.

The thermostability assays used to develop βAR34–424 mutants were performed by heating the receptor in the absence of the antagonist, but it is well known that bound ligand stabilizes receptors. Therefore, stability assays for βAR34–424 and βAR-m23 were repeated with antagonist bound to the receptors during the heating step (Fig. 3). As expected, the Tm of the receptor that contained bound antagonist during the incubation was higher than that for the receptor without antagonist. For βAR34–424 the Tm was 6°C higher with bound antagonist, and for βAR-m23 the Tm increased by 2°C to 55°C; the smaller increase in thermostability observed for βAR-m23 when antagonist binds suggests that the receptor is already in a more stable conformation, similar to the antagonist-bound state than βAR34–424 (see also below). The Tm of βAR-m23 with antagonist bound is very similar to the Tm of dark-state rhodopsin in dodecylmaltoside (DDM) (13), whose structure has been solved by two independent laboratories (14, 15). This suggested that βAR-m23 is sufficiently stable for crystallization.

An external file that holds a picture, illustration, etc. Object name is zpq0010889990003.jpg

Stability of βAR-m23 and βAR34–424 in the apo state or containing the bound antagonist [3H]DHA. To determine the apparent Tm in the absence of ligand (apo state, discontinuous lines), detergent-solubilized receptors were incubated for 30 min at the temperatures indicated before carrying out the binding assay. For the apparent Tm determination of the antagonist-bound form (continuous lines), detergent-solubilized receptors were preincubated with [3H]DHA, followed by incubation at the temperatures indicated. Circles, βAR-m23; squares, βAR34–424. Data points are from duplicate measurements in a representative experiment.

Characterization of βAR-m23.

The two characteristic activities measured for βAR-m23 and βAR34–424 to identify the effect of the six mutations were the affinity of antagonist binding and the relative order of agonist binding. Saturation binding experiments to membranes using the antagonist [3H]dihydroalprenolol (SI Fig. 7) showed that the affinity for βAR-m23 (KD = 6.5 ± 0.2 nM, n = 2) was similar to that for βAR34–424 (KD = 2.8 ± 0.1 nM, n = 2), suggesting that there are no large perturbations in the structure of βAR-m23 in the antagonist-bound conformation. This is consistent with the observation that none of the mutations in βAR-m23 corresponds with amino acids that are essential for ligand binding (16). In addition, a competition curve shows that binding of the antagonist cyanopindolol to βAR-m23 and βAR34–424 is also similar (Fig. 4). In contrast to antagonist binding, βAR-m23 binds the agonists norepinephrine and isoprenaline 2,470 times and 650 times, respectively, more weakly than βAR34–424 (Fig. 4). The affinity of the agonist isoprenaline is consistently lower in βAR-m23 and βAR34–424 than for the native agonist norepinephrine, indicating that the agonist-bound conformation for the two receptors is likely to be similar. However, the fact that βAR-m23 binds agonists three orders of magnitude more weakly than βAR34–424 indicates that the six mutations in βAR-m23 have locked the receptor preferentially in an antagonist-bound conformation. From a crystallization perspective this is an added bonus to thermostabilization because it is essential to have a conformationally homogeneous protein population for the production of diffraction-quality crystals.

An external file that holds a picture, illustration, etc. Object name is zpq0010889990004.jpg

Competition binding of agonists to βAR-m23 and βAR34–424. Binding assays were performed on receptors partially purified in DDM. Triangles, βAR-m23; squares, βAR34–424. [3H]DHA was used at a concentration three times greater than the KD of partially purified receptor (see Methods). [3H]DHA binding was competed with increasing concentrations of the agonists norepinephrine (a) and isoprenaline (b) or with the antagonists alprenolol (c) and cyanopindolol (d). LogEC50 and corresponding EC50 values for the different ligands were calculated by nonlinear regression using GraphPad Prism software, and the error for logEC50 values was <10%. The EC50 values for ligand binding to βAR34–424 and βAR-m23 are as follows: 1.5 μM norepinephrine for βAR34–424 and 3.7 mM norepinephrine for βAR-m23; 315 nM isoprenaline for βAR34–424 and 205 μM isoprenaline for βAR-m23; 78 nM alprenolol for βAR and 112 nM alprenolol for βAR-m23; and 2 nM cyanopindolol for βAR and 1 nM cyanopindolol for βAR-m23.

All of the thermostability assays used to derive βAR-m23 were performed on receptors solubilized in DDM. The aim of the thermostabilization process was to produce a receptor that is ideal for crystallography, which means being stable in a variety of different detergents and not just DDM. We therefore tested the stability of βAR-m23 and βAR in a variety of different detergents, concentrating on small detergents that are preferentially used in crystallizing integral membrane proteins. Membranes prepared from E. coli expressing βAR-m23 or βAR34–424 were solubilized in DDM, bound to Ni-NTA agarose, and then washed with DDM, decylmaltoside (DM), octyl glucoside (OG), lauryldimethylamine oxide (LDAO), or nonylglucoside (NG). Stability assays were performed on the receptors in each of the different detergents (Fig. 5). βAR34–424 was stable only in DDM and DM, with no active receptors eluting from the resin washed with OG, NG, or LDAO. In contrast, functional βAR-m23 was still present in all detergents, and the Tm could be determined. As expected, the smaller detergents were considerably more denaturing than either DDM (Tm = 52°C) or DM (Tm = 48°C), with Tm values of 25°C (NG), 23°C (LDAO), and 17°C (OG). The difference in Tm between βAR-m23 and βAR34–424 is ≈20°C, irrespective of whether the receptors were solubilized in either DDM or DM; it is therefore not surprising that no active βAR34–424 could be found in even NG, because the predicted Tm would be ≈5°C, thus resulting in rapid inactivation of the receptor under the conditions used for purification. The selection strategy used for the generation of βAR-m23 was chosen deliberately to be based on thermostability, because it is far simpler to apply than selecting for stability in detergents of increasing harshness. However, it is clear that increasing the thermostability of βAR34–424 also resulted in increasing tolerance to small detergents ideal for crystallizing integral membrane proteins.

An external file that holds a picture, illustration, etc. Object name is zpq0010889990005.jpg

Stability of βAR-m23 and βAR34–424 in five different detergents. Samples of βAR34–424 (a) and βAR-m23 (b) solubilized in DDM were partially purified on Ni-NTA agarose columns, allowing the exchange into various different detergents: DDM (pale blue squares), DM (dark blue triangles), OG (red inverted triangles), LDAO (brown diamonds), and NG (green circles). βAR is so unstable in OG, NG, and LDAO that it was not possible to measure any activity after purification at 6°C. Assays were carried out as described in Methods, and the results are duplicate measurements from a representative experiment performed in parallel.

Discussion

The evolution of βAR-m23 from βAR34–424 by a combination of alanine scanning mutagenesis and the selection of thermostable mutants has resulted in a GPCR that is ideal for crystallography. The apparent Tm for βAR-m23 is 21°C higher than for βAR34–424, and, in the presence of antagonist, βAR-m23 has a similar stability to rhodopsin. In comparison, the binding of the antagonist [3H]DHA improves the thermostability of βAR34–424 by 6°C for βAR34–424 and 2°C for βAR-m23 (Fig. 3). The increased apparent Tm of βAR-m23 has resulted in an increased stability in a variety of small detergents that inactivate βAR34–424. In addition, the selection strategy used resulted in a receptor that is preferentially in the antagonist-bound conformation, which will also improve the chances of obtaining crystals because the population of receptor conformations will be more homogeneous than for wild-type βAR34–424. The evidence for βAR-m23's being preferentially in an antagonist conformation is based on the competition curves where two different antagonists bound with similar affinity to wild-type receptor, but the two different agonists bound 2–3 orders of magnitude more weakly. The simplest explanation of this is that the mutant has the equilibrium shifted toward the antagonist-bound form and away from the agonist conformation. A full pharmacological analysis of βAR-m23 is currently in progress, including its ability to couple to G proteins.

The βAR construct used to develop βAR-m23 is ideally suited to the thermostabilization procedure by having good functional expression in E. coli. However, it is unlikely that this construct will produce protein that is suitable for crystallization because of the fusion of βAR via a flexible linker to maltose binding protein. Extensive previous work by Warne et al. (10) has shown that βAR can be overexpressed and engineered to facilitate crystallization. Improving the crystallizability of a membrane protein by locking it into a single conformation was graphically demonstrated by the structure determination of lactose permease (LacY), which contained a single point mutation (Cys154Gly) that locked it into the inward-facing conformation (17). However, bacterial membrane proteins are invariably far more stable than their mammalian counterparts, but here we demonstrate that a single selection procedure can simultaneously lock the membrane protein in a defined conformation and increase its thermostability; we have called this process conformational thermostabilization. This approach has also been used on two other GPCRs and has produced 16°C and 19°C increases in apparent Tm to date; therefore, we are sure that this methodology is likely to be broadly applicable.

It is not clear why the particular mutations we have introduced lead to the thermostabilization of the receptor. Given that the agonist conformation of GPCRs is considerably less stable than the antagonist conformation, it is likely that a proportion of the thermostabilization is due to mutations that prevent the formation of the agonist conformation. Other mutations could be functioning in either altering the energy balance between unfolded and folded states in favor of the folded state or changing the kinetic barriers between them. After the completion of this work, the structure of the β2 adrenergic receptor was determined (18, 19), which is 59% identical to the turkey β1 receptor, but with a distinctly different pharmacological profile (20, 21). The equivalent positions in the β2 receptor corresponding to the thermostabilizing mutations in βAR-m23 show that five of them are at the protein-solvent boundary, with one residue being at a kink in an α-helix and pointing toward the substrate binding pocket (SI Fig. 8). Given the difficulties in trying to understand the complexities of the thermostabilization of soluble proteins (22), it seems unlikely that membrane proteins will be any easier to comprehend; indeed, we found that, when the full set of mutations was considered, the basis for thermostabilization was not immediately obvious. However, because nearly 5% of the mutants produced were more stable than the native receptor, alanine scanning mutagenesis represents an efficient strategy to rapidly identify thermostable mutants.

The procedure we have used to generate βAR-m23 is equally applicable to any membrane protein that has a convenient assay for detecting activity in the detergent-solubilized form. The assay could detect the binding of a radiolabeled ligand to an allosteric, orthosteric, or substrate binding site, an antibody binding to a conformational epitope, or, in fact, any other compound (protein, RNA, DNA, ion, or molecule) that specifically binds somewhere to the membrane protein only when it is correctly folded. The advantage of these techniques is that they can all be applied to picomole quantities of unpurified protein in a high-throughput manner, whereas more traditional techniques for determining protein stability such as differential scanning calorimetry or CD spectroscopy rely on having purified protein in reasonable quantities. The latter techniques also have the disadvantage of not being able to select for functional mutants. Although we have selected for stability as a function of temperature as the most convenient primary parameter, the procedure can easily be extended to test primarily for stability, for example, in a harsh detergent, in an extreme of pH, or in the presence of chaotropic salts. Conformational thermostabilization of a variety of human receptors, channels, and transporters will make them far more amenable to crystallography and will also allow the improvement in resolution of membrane proteins that have already been crystallized. It is to be hoped that conformational thermostabilization will allow membrane protein crystallization to become a far more tractable problem with a greater probability of rapid success than is currently the case. This should allow routine crystallization of human membrane proteins in the pharmaceutical industry, resulting in valuable structural insights into drug development.

Methods

Materials.

The truncated βAR from turkey (βAR34–424) (10) was kindly provided by Tony Warne (Laboratory of Molecular Biology, Medical Research Council). This βAR construct encoding residues 34–424 contains the mutation C116L to improve expression (12) and a C-terminal tag of 10 histidines for purification. 1-[4,6-propyl-3H]dihydroalprenolol ([3H]DHA) was supplied by Amersham Bioscience, (+)l-norepinephrine bitartrate salt, (−)isoprenaline hydrochloride, (−)alprenolol tartrate salt, and s-propranolol hydrochloride were from Sigma.

Mutagenesis of βAR.

The βAR cDNA was ligated into pRGIII to allow the functional expression of βAR in E. coli as a MalE fusion protein (23). Mutants were generated by PCR using the expression plasmid as template and using the QuikChange II methodology (Stratagene). PCRs were transformed into XL10-Gold ultracompetent cells (Stratagene), and individual clones were fully sequenced to check that only the desired mutation was present. Different mutations were combined randomly by PCR by including all of the pairs of primers that introduced the following mutations: Mut4, G67A, G068A, V230A, D322A, and F327A; Mut6, R068S, Y227A, A234L, A282L, and A334L; Mut7, M90V, I129V, Y227A, A282L, and F338M; Mut10, R68S, M90V, V230A, F327A, and A334L. The PCR mixes were transformed, and the clones were sequenced to determine exactly which mutations were introduced.

Protein Expression and Membrane Preparations.

Expression of βAR and the mutants was performed in XL10 cells (Stratagene). Cultures of 50 ml of 2× TY medium containing ampicillin (100 μg/ml) were grown at 37°C with shaking until OD600 = 3 and then induced with 0.4 mM IPTG. Induced cultures were incubated at 25°C for 4 h, and then cells were harvested by centrifugation at 13,000 × g for 1 min (aliquots of 2 ml) and stored at −20°C. For the assays, cells were broken by freeze–thaw (five cycles) and resuspended in 500 μl of buffer [20 mM Tris, pH 8/0.4 M NaCl/1 mM EDTA/protease inhibitors (Complete; Roche)]. After an incubation for 1 h at 4°C with 100 μg/ml lysozyme and DNase I (Sigma), samples were solubilized with 2% DDM on ice for 30 min. Insoluble material was removed by centrifugation (15,000 × g for 2 min at 4°C), and the supernatant was used directly in radioligand binding assays.

For large-scale membrane preparations, 2 liters and 6 liters of E. coli culture of βAR and Mut23, respectively, were grown as described above. Cells were harvested by centrifugation at 5,000 × g for 20 min, frozen in liquid nitrogen, and stored at −80°C. Pellets were resuspended in 10 ml of 20 mM Tris (pH 7.5) containing 1× protease inhibitor mixture (Complete EDTA-free; Roche); 1 mg of DNase I (Sigma) was added, and the final volume was made to 100 ml. Cells were broken by a French press (two passages, 20,000 psi) and centrifuged at 12,000 × g for 45 min at 4°C to remove cell debris. The supernatant (membranes) was centrifuged at 200,000 × g for 30 min at 4°C; the membrane pellet was resuspended in 15 ml of 20 mM Tris (pH 7.5) and stored in 1-ml aliquots at −80°C after flash-freezing in liquid nitrogen. The protein concentration was determined by the amido black method (24). These samples were used in radioligand binding assays after thawing and being solubilized in 2% DDM as above.

For competition assays, as well as testing different detergents, DDM-solubilized βAR was partially purified with Ni-NTA agarose (Qiagen). A total of 200 μl of Ni-NTA agarose was added to 2 ml of solubilized samples (10 mg/ml membrane protein) in 20 mM Tris (pH 8)/0.4 M NaCl/20 mM imidazole (pH 8) and incubated for 1 h at 4°C. After incubation, samples were centrifuged at 13,000 × g for 30 s and washed twice with 250 μl of buffer (20 mM Tris, pH 8/0.4 M NaCl/20 mM imidazole) containing detergent (0.1% DDM, 0.1% DM, 0.1% LDAO, 0.3% NG, or 0.7% OG). Receptors were eluted in 2 × 100 μl of buffer (0.4 M NaCl/1 mM EDTA/250 mM imidazole, pH 8, plus the relevant detergent). The KD for [3H]DHA binding to semipurified βAR34–424 and βAR-m23 was, respectively 3.7 nM and 12.5 nM, and the final concentration of [3H]DHA used in the competition assays was three times the KD, i.e., 12 nM for βAR34–424 and 40 nM for βAR-m23.

Radioligand Binding and Thermostability Assays.

Single point binding assays contained 20 mM Tris (pH 8), 0.4 M NaCl, 1 mM EDTA, and 0.1% DDM (or corresponding detergent) with 50 nM [3H]DHA and 20–100 μg of membrane protein in a final volume of 120 μl; equilibration was for 1 h at 4°C. Thermostability was assessed by incubating the binding assay mix with or without [3H]DHA at the specified temperature for 30 min; reactions were placed on ice, and [3H]DHA added as necessary and equilibrated for a further hour. Receptor-bound and free radioligand were separated by gel filtration as described previously (10). Nonspecific binding was determined in the presence of 1 μM s-propranolol. Saturation curves were obtained by using a range of [3H]DHA concentrations from 0.4 nM to 100 nM. Competition assays were performed by using a concentration of [3H]DHA of 12 nM for βAR34–424 and 40 nM for βAR-m23 (i.e., three times the KD) and various concentrations of unlabeled ligands (0–100 mM). Radioactivity was counted on a Beckman LS6000 liquid scintillation counter, and data were analyzed by nonlinear regression using Prism software (GraphPad).

Location of βAR-m23 Thermostable Mutations in Rhodopsin Structure.

The Protein Data Bank file for the rhodopsin structure, ID code 1GZM (15), was downloaded from the Protein Data Bank (www.pdb.org) and displayed in the program PyMOLX11Hybrid (DeLano Scientific). The equivalent amino acid residues in rhodopsin for the thermostable mutations in βAR were located in the rhodopsin structure based on an alignment among the four GPCRs with which we are most familiar, namely rhodopsin, βAR, neurotensin receptor, and adenosine A2a receptor (25).

Supplementary Material

Supporting Information:

ACKNOWLEDGMENTS.

We are grateful to to R. Henderson for comments and support. We thank G. Schertler (Medical Research Council Laboratory of Molecular Biology, Cambridge, U.K.), T. Warne, and R. Grisshammer (National Institute of Neurological Disorders and Stroke, National Institutes of Health), for plasmid constructs and informal discussions about previously published work. This work was supported by the Medical Research Council Technology Development Gap Fund and Pfizer.

Footnotes

The authors declare no conflict of interest.

This article contains supporting information online at www.pnas.org/cgi/content/full/0711253105/DC1.

References

1. White SH. The progress of membrane protein structure determination. Protein Sci. 2004;13:1948–1949. [PMC free article] [PubMed] [Google Scholar]
2. Tate CG. Overexpression of mammalian integral membrane proteins for structural studies. FEBS Lett. 2001;504:94–98. [PubMed] [Google Scholar]
3. Grisshammer R, Tate CG. Overexpression of integral membrane proteins for structural studies. Q Rev Biophys. 1995;28:315–422. [PubMed] [Google Scholar]
4. Bowie JU. Stabilizing membrane proteins. Curr Opin Struct Biol. 2001;11:397–402. [PubMed] [Google Scholar]
5. Lau FW, Nauli S, Zhou Y, Bowie JU. Changing single side-chains can greatly enhance the resistance of a membrane protein to irreversible inactivation. J Mol Biol. 1999;290:559–564. [PubMed] [Google Scholar]
6. Zhou Y, Bowie JU. Building a thermostable membrane protein. J Biol Chem. 2000;275:6975–6979. [PubMed] [Google Scholar]
7. Faham S, et al. Side-chain contributions to membrane protein structure and stability. J Mol Biol. 2004;335:297–305. [PubMed] [Google Scholar]
8. Yarden Y, et al. Proc Natl Acad Sci USA. 1986;83:6795–6799. [PMC free article] [PubMed] [Google Scholar]
9. Hertel C, et al. A truncation mutation in the avian beta-adrenergic receptor causes agonist-induced internalization and GTP-sensitive agonist binding characteristic of mammalian receptors. J Biol Chem. 1990;265:17988–17994. [PubMed] [Google Scholar]
10. Warne T, Chirnside J, Schertler GF. Expression and purification of truncated, non-glycosylated turkey beta-adrenergic receptors for crystallization. Biochim Biophys Acta. 2003;1610:133–140. [PubMed] [Google Scholar]
11. Parker EM, Ross EM. Truncation of the extended carboxyl-terminal domain increases the expression and regulatory activity of the avian beta-adrenergic receptor. J Biol Chem. 1991;266:9987–9996. [PubMed] [Google Scholar]
12. Parker EM, Kameyama K, Higashijima T, Ross EM. Reconstitutively active G protein-coupled receptors purified from baculovirus-infected insect cells. J Biol Chem. 1991;266:519–527. [PubMed] [Google Scholar]
13. Degrip WJ. Thermal stability of rhodopsin and opsin in some novel detergents. Methods Enzymol. 1982;81:256–265. [PubMed] [Google Scholar]
14. Palczewski K, et al. Crystal structure of rhodopsin: A G protein-coupled receptor. Science. 2000;289:739–745. [PubMed] [Google Scholar]
15. Li J, Edwards PC, Burghammer M, Villa C, Schertler GF. Structure of bovine rhodopsin in a trigonal crystal form. J Mol Biol. 2004;343:1409–1438. [PubMed] [Google Scholar]
16. Rezmann-Vitti LA, et al. Role of Tyr(356(7.43)) and Ser(190(4.57)) in antagonist binding in the rat beta1-adrenergic receptor. J Med Chem. 2006;49:3467–3477. [PubMed] [Google Scholar]
17. Abramson J, et al. Structure and mechanism of the lactose permease of Escherichia coli. Science. 2003;301:610–615. [PubMed] [Google Scholar]
18. Rasmussen SG, et al. Crystal structure of the human beta(2) adrenergic G-protein-coupled receptor. Nature. 2007;15:383–387. [PubMed] [Google Scholar]
19. Cherezov V, et al. High-resolution crystal structure of an engineered human β2-adrenergic G protein coupled receptor. Science. 2007;318:1258–1265. [PMC free article] [PubMed] [Google Scholar]
20. Minneman KP, Weiland GA, Molinoff PB. A comparison of the beta-adrenergic receptor of the turkey erythrocyte with mammalian beta1 and beta2 receptors. Mol Pharmacol. 1980;17:1–7. [PubMed] [Google Scholar]
21. Parker EM, Swigart P, Nunnally MH, Perkins JP, Ross EM. Carboxyl-terminal domains in the avian beta 1-adrenergic receptor that regulate agonist-promoted endocytosis. J Biol Chem. 1995;270:6482–6487. [PubMed] [Google Scholar]
22. Jaenicke R, Bohm G. The stability of proteins in extreme environments. Curr Opin Struct Biol. 1998;8:738–748. [PubMed] [Google Scholar]
23. Tucker J, Grisshammer R. Purification of a rat neurotensin receptor expressed in Escherichia coli. Biochem J. 1996;317:891–899. [PMC free article] [PubMed] [Google Scholar]
24. Schaffner W, Weissmann C. Anal Biochem. 1973;56:502–514. [PubMed] [Google Scholar]
25. Weiss HM, Grisshammer R. Purification and characterization of the human adenosine A (2a) receptor functionally expressed in Escherichia coli. Eur J Biochem. 2002;269:82–92. [PubMed] [Google Scholar]
Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences