MUTANT G-PROTEIN COUPLED RECEPTORS AND METHODS FOR SELECTING THEM

Abstract

The invention relates to mutant G-protein coupled receptors with increased conformational stability, and methods of use thereof. In some aspects, polynucleotides encoding the mutant G-protein coupled receptors are provided. In some aspects, host cells comprising the polynucleotides are provided. In some aspects, the invention relates to crystallized forms of the mutant G-protein coupled receptors, and methods of preparing the same.

Description

RELATED APPLICATIONS

This application is a continuation of U.S. patent application Ser. No. 13/493,898, filed Jun. 11, 2012, entitled “Mutant G-Protein Coupled Receptors and Methods For Selecting Them,” which is a continuation of U.S. patent application Ser. No. 12/450,358, filed Mar. 19, 2010, entitled “Mutant G-Protein Coupled Receptors and Methods For Selecting Them,” which is a national stage filing under 35 U.S.C. § 371 of international application PCT/GB2008/000986, filed Mar. 20, 2008, which was published under PCT Article 21(2) in English and claims priority to UK Patent Application No. 0724052.6, filed Dec. 8, 2007, and UK Patent Application No. 0705450.5, filed Mar. 22, 2007, the entire disclosures of each of which are herein incorporated by reference in their entireties.

The present invention relates to mutant G protein coupled receptors (GPCRs) and methods for selecting those with increased stability. In particular, it relates to the selection and preparation of mutant GPCRs which have increased stability under a particular condition compared to their respective parent proteins. Such proteins are more likely to be crystallisable, and hence amenable to structure determination, than the parent proteins. They are also useful for drug discovery and development studies.

Over the past 20 years the rate of determination of membrane protein structures has gradually increased, but most success has been in crystallising 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 sometimes far more stable in detergent, detergent-stability being an essential prerequisite to purification and crystallisation. Genome sequencing projects have also allowed the cloning and expression of many homologues of a specific transporter or ion channel, which also greatly improves the chances of success during crystallisation. However, out of the 120 different membrane protein structures that have been solved to date, there are only seven structures of mammalian integral membrane proteins (blanco.biomol.uci.edu/); 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 crystallisation 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 destabilising detergents ie 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, whilst yeast and bacterial genomes may not include any homologous proteins. There is thus an overwhelming need to develop a generic strategy that will allow the production of detergent-stable eukaryotic integral membrane proteins for crystallisation and structure determination and potentially for other purposes such as drug screening, bioassay and biosensor applications.

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 isolated [4]. This was subsequently demonstrated for two bacterial proteins, diacylglycerol kinase (DGK) [5,6] and bacteriorhodopsin [7]. Random mutagenesis of DGK 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 minutes at 80° C. compared with a half-life of 6 minutes at 55° C. for the native protein [6]. It was shown that the trimer of the detergent-resistant DGK mutant had become stable in SDS and it is thus likely that stabilisation of the oligomeric state played a significant role in thermostabilisation. Although the aim of the mutagenesis was to produce a membrane protein suitable for crystallisation, the structure of DGK has yet to be determined and there have been no reports of successful crystallization. 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 thermostabilisation had occurred [7].

GPCRs constitute a very large family of proteins that control many physiological processes and are the targets of many effective drugs. Thus, they are of considerable pharmacological importance. A list of GPCRs is given in Foord et al (2005) Pharmacol Rev. 57, 279-288, which is incorporated herein by reference. GPCRs are generally unstable when isolated, and despite considerable efforts, it has not been possible to crystallise any except bovine rhodopsin, which naturally is exceptionally stable.

GPCRs are druggable targets, and reference is made particularly to Overington et al (2006) Nature Rev. Drug Discovery 5, 993-996 which indicates that over a quarter of present drugs have a GPCR as a target.

GPCRs are thought to exist in multiple distinct conformations which are associated with different pharmacological classes of ligand such as agonists and antagonists, and to cycle between these conformations in order to function (Kenakin T. (1997) Ann N Y Acad Sci 812, 116-125).

It will be appreciated that the methods of the invention do not include a method as described in D'Antona et al., including binding of [³H]CP55940 to a constitutively inactive mutant human cannabinoid receptor 1 (T210A) in which the Thr residue at position 210 is replaced with an Ala residue.

The listing or discussion of an apparently prior-published document in this specification should not necessarily be taken as an acknowledgement that the document is part of the state of the art or is common general knowledge.

We have realised that there are two serious problems associated with trying to crystallise GPCRs, namely their lack of stability in detergent and the fact that they exist in multiple conformations. In order to function GPCRs 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. Just adding ligands to GPCRs during crystallisation trials has not resulted in their structure determination. To improve the likelihood of crystallisation, we therefore selected mutations that improved the stability of the GPCR and, in addition, preferentially locked the receptor in a specific biologically relevant conformation.

We decided to see whether stabilisation of a GPCR in a particular, biologically relevant conformation was possible and whether the effect was sufficiently great that it would significantly improve the chances of obtaining diffraction-quality crystals. In Example 1, 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 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 [9]. In Example 2, a human adenosine receptor was used, and in Example 3, a rat neurotensin receptor was used.

Method for Selecting Mutant GPCRs with Increased Stability

A first aspect of the invention provides a method for selecting a mutant G-protein coupled receptor (GPCR) with increased stability, the method comprising

• • (a) providing one or more mutants of a parent GPCR,
  • (b) selecting a ligand, the ligand being one which binds to the parent GPCR when the GPCR is residing in a particular conformation,
  • (c) determining whether the or each mutant GPCR has increased stability with respect to binding the selected ligand compared to the stability of the parent GPCR with respect to binding that ligand, and
  • (d) selecting those mutants that have an increased stability compared to the parent GPCR with respect to binding of the selected ligand.

The inventors have appreciated that, in order to improve the likelihood of crystallisation of a GPCR in a biologically relevant form (which is therefore pharmacologically useful), it is desirable not only to increase the stability of the protein, but also for the protein to have this increased stability when in a particular conformation. The conformation is determined by a selected ligand, and is a biologically relevant conformation in particular a pharmacologically relevant conformation. Thus, the method of the invention may be considered to be a method for selecting mutants of a GPCR which have increased stability of a particular conformation, for example they may have increased conformational thermostability. The method may be used to create stable, conformationally locked GPCRs by mutagenesis. The selected mutant GPCRs are effectively purer forms of the parent molecules in that a much higher proportion of them occupies a particular conformational state. The deliberate selection of a chosen receptor conformation resolved from other conformations by use of a ligand (or ligands) that bind preferentially to this conformation is therefore an important feature of the invention. The method may also be considered to be a method for selecting mutant GPCRs which are more tractable to crystallisation.

Thus the invention includes a method for selecting a mutant G-protein coupled receptor (GPCR) with increased stability, the method comprising

• • (a) providing one or more mutants of a parent GPCR,
  • (b) selecting a ligand, the ligand being one which binds to the parent GPCR when the GPCR is residing in a particular conformation,
  • (c) determining whether the or each mutant GPCR when residing in the particular conformation has increased stability with respect to binding the selected ligand compared to the stability of the parent GPCR when residing in the same particular conformation with respect to binding that ligand, and
  • (d) selecting those mutants that have an increased stability compared to the parent GPCR with respect to binding of the selected ligand.

In a review of the druggable genome by Hopkins & Groom (2002) Nature Rev. Drug Discovery 1, 727-730, Table 1 contains a list of protein families many of which are GPCRs. Overington et al (2006) Nature Rev. Drug Discovery 5, 993-996 provides more details of drug targets, and FIG. 1 indicates that more than a quarter of current drugs target GPCRs. There are 52 GPCR targets for orally available drugs out of a total of 186 total targets in this category.

Suitable GPCRs for use in the practice of the invention include, but are not limited to β-adrenergic receptor, adenosine receptor, in particular adenosine A_2a receptor, and neurotensin receptor (NTR). Other suitable GPCRs are well known in the art and include those listed in Hopkins & Groom supra. In addition, the International Union of Pharmacology produce a list of GPCRs (Foord et al (2005) Pharmacol. Rev. 57, 279-288, incorporated herein by reference and this list is periodically updated at www.iuphar-db.org/GPCR/ReceptorFamiliesForward. It will be noted that GPCRs are divided into different classes, principally based on their amino acid sequence similarities. They are also divided into families by reference to the natural ligands to which they bind. All GPCRs are included in the scope of the invention.

The amino acid sequences (and the nucleotide sequences of the cDNAs which encode them) of many GPCRs are readily available, for example by reference to GenBank. In particular, Foord et al supra gives the human gene symbols and human, mouse and rat gene IDs from Entrez Gene (www.ncbi.nlm.nih.gov/entrez). It should be noted, also, that because the sequence of the human genome is substantially complete, the amino acid sequences of human GPCRs can be deduced therefrom.

Although the GPCR may be derived from any source, it is particularly preferred if it is from a eukaryotic source. It is particularly preferred if it is derived from a vertebrate source such as a mammal or a bird. It is particularly preferred if the GPCR is derived from rat, mouse, rabbit or dog or non-human primate or man, or from chicken or turkey. For the avoidance of doubt, we include within the meaning of “derived from” that a cDNA or gene was originally obtained using genetic material from the source, but that the protein may be expressed in any host cell subsequently. Thus, it will be plain that a eukaryotic GPCR (such as an avian or mammalian GPCR) may be expressed in a prokaryotic host cell, such as E. coli, but be considered to be avian- or mammalian-derived, as the case may be.

In some instances, the GPCR may be composed of more than one different subunit. For example, the calcitonin gene-related peptide receptor requires the binding of a single transmembrane helix protein (RAMP1) to acquire its physiological ligand binding characteristics. Effector, accessory, auxiliary or GPCR-interacting proteins which combine with the GPCR to form or modulate a functional complex are well known in the art and include, for example, receptor kinases, G-proteins and arrestins (Bockaert et al (2004) Curr Opinion Drug Discov and Dev 7, 649-657).

The mutants of the parent GPCR may be produced in any suitable way and provided in any suitable form. Thus, for example, a series of specific mutants of the parent protein may be made in which each amino acid residue in all or a part of the parent protein is independently changed to another amino acid residue. For example, it may be convenient to make mutations in those parts of the protein which are predicted to be membrane spanning. The three-dimensional structure of rhodopsin is known (Li et al (2004) J Mol Biol 343, 1409-1438; Palczewski et al (2000) Science 289, 739-745), and it is possible to model certain GPCRs using this structure. Thus, conveniently, parts of the GPCR to mutate may be based on modelling. Similarly, computer programs are available which model transmembrane regions of GPCRs based on hydrophobicity (Kyle & Dolittle (1982) J. Mol. Biol. 157, 105-132), and use can be made of such models when selecting parts of the protein to mutate. Conventional site-directed mutagenesis may be employed, or polymerase chain reaction-based procedures well known in the art may be used. It is possible, but less desirable, to use ribosome display methods in the selection of the mutant protein.

Typically, each selected amino acid is replaced by Ala (ie Ala-scanning mutagenesis), although it may be replaced by any other amino acid. If the selected amino acid is Ala, it may conveniently be replaced by Leu. Alternatively, the amino acid may be replaced by Gly (ie Gly-scanning mutagenesis), which may allow a closer packing of neighbouring helices that may lock the protein in a particular conformation. If the selected amino acid is Gly, it may conveniently be replaced by Ala.

Although the amino acid used to replace the given amino acid at a particular position is typically a naturally occurring amino acid, typically an “encodeable” amino acid, it may be a non-natural amino acid (in which case the protein is typically made by chemical synthesis or by use of non-natural amino-acyl tRNAs). An “encodeable” amino acid is one which is incorporated into a polypeptide by translation of mRNA. It is also possible to create non-natural amino acids or introduce non-peptide linkages at a given position by covalent chemical modification, for example by post-translational treatment of the protein or semisynthesis. These post-translational modifications may be natural, such as phosphorylation, glycosylation or palmitoylation, or synthetic or biosynthetic.

Alternatively, the mutants may be produced by a random mutagenesis procedure, which may be of the whole protein or of a selected portion thereof. Random mutagenesis procedures are well known in the art.

Conveniently, the mutant GPCR has one replaced amino acid compared to the parent protein (ie it is mutated at one amino acid position). In this way, the contribution to stability of a single amino acid replacement may be assessed. However, the mutant GPCR assayed for stability may have more than one replaced amino acid compared to the parent protein, such as 2 or 3 or 4 or 5 or 6 replacements.

As is discussed in more detail below, combinations of mutations may be made based on the results of the selection method. It has been found that in some specific cases combining mutations in a single mutant protein leads to a further increase in stability. Thus, it will be appreciated that the method of the invention can be used in an iterative way by, for example, carrying it out to identify single mutations which increase stability, combining those mutations in a single mutant GPCRs which is the GPCR then provided in part (a) of the method. Thus, multiply-mutated mutant proteins can be selected using the method.

The parent GPCR need not be the naturally occurring protein. Conveniently, it may be an engineered version which is capable of expression in a suitable host organism, such as Escherichia coli. For example, as described in Example 1, a convenient engineered version of the turkey β-adrenergic receptor is one which is truncated and lacks residues 1-33 of the amino acid sequence (ie βAR_34-424). The parent GPCR may be a truncated form of the naturally occurring protein (truncated at either or both ends), or it may be a fusion, either to the naturally occurring protein or to a fragment thereof. Alternatively or additionally, the parent GPCR, compared to a naturally-occurring GPCR, may be modified in order to improve, for example, solubility, proteolytic stability (eg by truncation, deletion of loops, mutation of glycosylation sites or mutation of reactive amino acid side chains such as cysteine). In any event, the parent GPCR is a protein that is able to bind to the selected ligand which ligand is one which is known to bind the naturally occurring GPCR. Conveniently, the parent GPCR is one which, on addition of an appropriate ligand, can affect any one or more of the downstream activities which are commonly known to be affected by G-protein activation.

However, it will be appreciated that the stability of the mutant is to be compared to a parent in order to be able to assess an increase in stability.

A ligand is selected, the ligand being one which binds to the parent GPCR when residing in a particular conformation. Typically, the ligand will bind to one conformation of the parent GPCR (and may cause the GPCR to adopt this conformation), but does not bind as strongly to another conformation that the GPCR may be able to adopt. Thus, the presence of the ligand may be considered to encourage the GPCR to adopt the particular conformation. Thus, the method may be considered to be a way of selecting mutant GPCRs which are trapped in a conformation of biological relevance (eg ligand bound state), and which are more stable with respect to that conformation.

Preferably the particular conformation in which the GPCR resides in step (c) corresponds to the class of ligand selected in step (b).

Preferably the selected ligand is from the agonist class of ligands and the particular conformation is an agonist conformation, or the selected ligand is from the antagonist class of ligands and the particular conformation is an antagonist conformation.

Preferably the selected ligand is from the agonist class of ligands and the particular conformation in which the GPCR resides in step (c) is the agonist conformation. Preferably, the selected ligand binding affinity for the mutant receptor should be equal to or greater than that for the wild type receptor; mutants that exhibit significantly reduced binding to the selected ligand are typically rejected.

By “ligand” we include any molecule which binds to the GPCR and which causes the GPCR to reside in a particular conformation. The ligand preferably is one which causes more than half of the GPCR molecules overall to be in a particular conformation.

Many suitable ligands are known.

Typically, the ligand is a full agonist and is able to bind to the GPCR and is capable of eliciting a full (100%) biological response, measured for example by G-protein coupling, downstream signaling events or a physiological output such as vasodilation. Thus, typically, the biological response is GDP/GTP exchange in a G-protein, followed by stimulation of the linked effector pathway. The measurement, typically, is GDP/GTP exchange or a change in the level of the end product of the pathway (eg cAMP, cGMP or inositol phosphates). The ligand may also be a partial agonist and is able to bind to the GPCR and is capable of eliciting a partial (<100%) biological response.

The ligand may also be an inverse agonist, which is a molecule which binds to a receptor and reduces its basal (ie unstimulated by agonist) activity sometimes even to zero.

The ligand may also be an antagonist, which is a molecule which binds to a receptor and blocks binding of an agonist, so preventing a biological response. Inverse agonists and partial agonists may under certain assay conditions be antagonists.

The above ligands may be orthosteric, by which we include the meaning that they combine with the same site as the endogenous agonist; or they may be allosteric or allotopic, by which we include the meaning that they combine with a site distinct from the orthosteric site. The above ligands may be syntopic, by which we include the meaning that they interact with other ligand(s) at the same or an overlapping site. They may be reversible or irreversible.

In relation to antagonists, they may be surmountable, by which we include the meaning that the maximum effect of agonist is not reduced by either pre-treatment or simultaneous treatment with antagonist; or they may be insurmountable, by which we include the meaning that the maximum effect of agonist is reduced by either pre-treatment or simultaneous treatment with antagonist; or they may be neutral, by which we include the meaning the antagonist is one without inverse agonist or partial agonist activity. Antagonists typically are also inverse agonists.

Ligands for use in the invention may also be allosteric modulators such as positive allosteric modulators, potentiators, negative allosteric modulators and inhibitors. They may have activity as agonists or inverse agonists in their own right or they may only have activity in the presence of an agonist or inverse agonist in which case they are used in combination with such molecules in order to bind to the GPCR.

Neubig et al (2003) Pharmacol. Rev. 55, 597-606, incorporated herein by reference, describes various classes of ligands.

Preferably, the above-mentioned ligands are small organic or inorganic moieties, but they may be peptides or polypeptides. Typically, when the ligand is a small organic or organic moiety, it has a M_r of from 50 to 2000, such as from 100 to 1000, for example from 100 to 500.

Typically, the ligand binds to the GPCR with a K_d of from mM to pM, such as in the range of from μM (micromolar) to nM. Generally, the ligands with the lowest Kd are preferred.

Small organic molecule ligands are well known in the art, for example see the Examples below. Other small molecule ligands include 5HT which is a full agonist at the 5HT1A receptor; eltoprazine which is a partial agonist at the 5HT1A receptor (see Newman-Tancredi et al (1997) Neurophamacology 36, 451-459); (+)-butaclamol and spiperone are dopamine D2 receptor inverse agonists (see Roberts & Strange (2005) Br. J. Pharmacol. 145, 34-42); and WIN55212-3 is a neutral antagonist of CB2 (Savinainen et al (2005) Br. J. Pharmacol. 145, 636-645).

The ligand may be a peptidomimetic, a nucleic acid, a peptide nucleic acid (PNA) or an aptamer. It may be an ion such as Na⁺ or Zn²⁺, a lipid such as oleamide, or a carbohydrate such as heparin.

The ligand may be a polypeptide which binds to the GPCR. Such polypeptides (by which we include oligopeptides) are typically from M_r 500 to M_r 50,000, but may be larger. The polypeptide may be a naturally occurring GPCR-interacting protein or other protein which interacts with the GPCR, or a derivative or fragment thereof, provided that it binds selectively to the GPCR in a particular conformation. GPCR-interacting proteins include those associated with signaling and those associated with trafficking, which often act via PDZ domains in the C terminal portion of the GPCR.

Polypeptides which are known to bind certain GPCRs include any of a G protein, an arrestin, a RGS protein, G protein receptor kinase, a RAMP, a 14-3-3 protein, a NSF, a periplakin, a spinophilin, a GPCR kinase, a receptor tyrosine kinase, an ion channel or subunit thereof, an ankyrin and a Shanks or Homer protein. Other polypeptides include NMDA receptor subunits NR1 or NR2a, calcyon, or a fibronectin domain framework. The polypeptide may be one which binds to an extracellular domain of a GPCR, such as fibulin-1. The polypeptide may be another GPCR, which binds to the selected GPCR in a hetero-oligomer. A review of protein-protein interactions at GPCRs is found in Milligan & White (2001) Trends Pharmacol. Sci. 22, 513-518, or in Bockaert et al (2004) Curr. Opinion Drug Discov. Dev. 7, 649-657 incorporated herein by reference.

The polypeptide ligand may conveniently be an antibody which binds to the GPCR. By the term “antibody” we include naturally-occurring antibodies, monoclonal antibodies and fragments thereof. We also include engineered antibodies and molecules which are antibody-like in their binding characteristics, including single chain Fv (scFv) molecules and domain antibodies (dAbs). Mention is also made of camelid antibodies and engineered camelid antibodies. Such molecules which bind GPCRs are known in the art and in any event can be made using well known technology. Suitable antibodies include ones presently used in radioimmunoassay (RIAs) for GPCRs since they tend to recognise conformational epitopes.

The polypeptide may also be a binding protein based on a modular framework, such as ankyrin repeat proteins, armadillo repeat proteins, leucine rich proteins, tetratriopeptide repeat proteins or Designed Ankyrin Repeat Proteins (DARPins) or proteins based on lipocalin or fibronectin domains or Affilin scaffolds based on either human gamma crystalline or human ubiquitin.

In one embodiment of the invention, the ligand is covalently joined to the GPCR, such as a G-protein or arrestin fusion protein. Some GPCRs (for example thrombin receptor) are cleaved N-terminally by a protease and the new N-terminus binds to the agonist site. Thus, such GPCRs are natural GPCR-ligand fusions.

It will be appreciated that the use of antibodies, or other “universal” binding polypeptides (such as G-proteins which are known to couple with many different GPCRs) may be particularly advantageous in the use of the method on “orphan” GPCRs for which the natural ligand, and small molecule ligands, are not known.

Once the ligand has been selected, it is then determined whether the or each mutant GPCR has increased stability with respect to binding the selected ligand compared to the parent GPCR with respect to binding that ligand. It will be appreciated that this step (c) is one in which it is determined whether the or each mutant GPCR has an increased stability (compared to its parent) for the particular conformation which is determined by the selected ligand. Thus, the mutant GPCR has increased stability with respect to binding the selected ligand as measured by ligand binding or whilst binding the selected ligand. As is discussed below, it is particularly preferred if the increased stability is assessed whilst binding the selected ligand.

The increased stability is conveniently measured by an extended lifetime of the mutant under the imposed conditions which may lead to instability (such as heat, harsh detergent conditions, chaotropic agents and so on). Destabilisation under the imposed condition is typically determined by measuring denaturation or loss of structure. As is discussed below, this may manifest itself by loss of ligand binding ability or loss of secondary or tertiary structure indicators.

As is described with respect to FIG. 12 below (which depicts a particular, preferred embodiment), there are different assay formats which may be used to determine stability of the mutant GPCR.

In one embodiment the mutant GPCR may be brought into contact with a ligand before being subjected to a procedure in which the stability of the mutant is determined (the mutant GPCR and ligand remaining in contact during the test period). Thus, for example, when the method is being used to select for mutant GPCRs which in one conformation bind to a ligand and which have improved thermostablity, the receptor is contacted with the ligand before being heated, and then the amount of ligand bound to the receptor following heating may be used to express thermostability compared to the parent receptor. This provides a measure of the amount of the GPCR which retains ligand binding capacity following exposure to the denaturing conditions (eg heat), which in turn is an indicator of stability.

In an alternative (but less preferred) embodiment, the mutant GPCR is subjected to a procedure in which the stability of the mutant is determined before being contacted with the ligand. Thus, for example, when the method is being used to select for mutant membrane receptors which in one conformation bind to a ligand and which have improved thermostability, the receptor is heated first, before being contacted with the ligand, and then the amount of ligand bound to the receptor may be used to express thermostability. Again, this provides a measure of the amount of the GPCR which retains ligand binding capacity following exposure to the denaturing conditions.

In both embodiments, it will be appreciated that the comparison of stability of the mutant is made by reference to the parent molecule under the same conditions.

It will be appreciated that in both of these embodiments, the mutants that are selected are ones which have increased stability when residing in the particular conformation compared to the parent protein.

The preferred route may be dependent upon the specific GPCR, and will be dependent upon the number of conformations accessible to the protein in the absence of ligand. In the embodiment described in FIG. 12, it is preferred if the ligand is present during the heating step because this increases the probability that the desired conformation is selected.

From the above, it will be appreciated that the invention includes a method for selecting a mutant GPCR with increased thermostability, the method comprising (a) providing one or more mutants of a parent GPCR, (b) selecting an antagonist or an agonist which binds the parent GPCR, (c) determining whether the or each mutant has increased thermostability when in the presence of the said antagonist or agonist by measuring the ability of the mutant GPCR to bind the selected said antagonist or agonist at a particular temperature and after a particular time compared to the parent GPCR and (d) selecting those mutant GPCRs that bind more of the selected said antagonist or agonist at the particular temperature and after the particular time than the parent GPCR under the same conditions. In step (c), a fixed period of time at the particular temperature is typically used in measuring the ability of the mutant GPCR to bind the selected said antagonist or agonist. In step (c), typically a temperature and a time is chosen at which binding of the selected said antagonist or agonist by the parent GPCR is reduced by 50% during the fixed period of time at that temperature (which is indicative that 50% of the receptor is inactivated; “quasi” Tm).

Conveniently, when the ligand is used to assay the GPCR (ie used to determine if it is in a non-denatured state), the ligand is detectably labelled, eg radiolabeled or fluorescently labelled. In another embodiment, ligand binding can be assessed by measuring the amount of unbound ligand using a secondary detection system, for example an antibody or other high affinity binding partner covalently linked to a detectable moiety, for example an enzyme which may be used in a colorimetric assay (such as alkaline phosphatase or horseradish peroxidase). FRET methodology may also be used. It will be appreciated that the ligand used to assay the mutant GPCR in determining its stability need not be the same ligand as selected in step (b) of the method.

Although it is convenient to measure the stability of the parent and mutant GPCR by using the ability to bind a ligand as an indicator of the presence of a non-denatured protein, other methods are known in the art. For example, changes in fluorescence spectra can be a sensitive indicator of unfolding, either by use of intrinsic tryptophan fluorescence or the use of extrinsic fluorescent probes such as 1-anilino-8-napthaleneulfonate (ANS), for example as implemented in the Thermofluor™ method (Mezzasalma et al, J Biomol Screening, 2007, April; 12(3):418-428). Proteolytic stability, deuterium/hydrogen exchange measured by mass spectrometry, blue native gels, capillary zone electrophoresis, circular dichroism (CD) spectra and light scattering may also be used to measure unfolding by loss of signals associated with secondary or tertiary structure. However, all these methods require the protein to be purified in reasonable quantities before they can be used (eg high pmol/nmol quantities), whereas the method described in the Examples makes use of pmol amounts of essentially unpurified GPCR.

In a preferred embodiment, in step (b) two or more ligands of the same class are selected, the presence of each causing the GPCR to reside in the same particular conformation. Thus, in this embodiment, one or more ligands (whether natural or non-natural) of the same class (eg full agonist or partial agonist or antagonist or inverse agonist) may be used. Including multiple ligands of the same class in this process, whether in series or in parallel, minimises the theoretical risk of inadvertently engineering and selecting multiply mutated receptor conformations substantially different to the parent, for example in their binding site, but still able, due to compensatory changes, to bind ligand. The following steps may be used to mitigate this risk:

1. Select a chemically distinct set (eg n=2-5) of ligands, in a common pharmacological class as evidenced by for example a binding or functional or spectroscopic assay. These ligands should be thought to bind to a common spatial region of the receptor, as evidenced for example by competitive binding studies using wild type and/or mutated receptors, and/or by molecular modelling, although they will not necessarily express a common pharmacophore.
2. Make single or multiple receptor mutants intended to increase stability, and assay for tight binding using the full set of ligands. The assays can be parallelised, multiplexed or run in series.
3. Confirm authenticity of stabilised receptor mutant by measurement for example of the binding isotherm for each ligand, and by measurement of the stability shift with ligand (the window should typically be narrowed compared to wild type).

In order to guard against changes in apparent affinity caused by perturbations to the binding site upon mutation, preferably ligands of the same pharmacological class, but different chemical class, should be used to profile the receptor. These should typically show similar shifts in affinity (mutant versus parent, e.g. wild type) in spite of having different molecular recognition properties. Binding experiments should preferably be done using labelled ligand within the same pharmacological class.

Nonetheless it should be recognised that conformational substrates may exist that are specific to chemical classes of ligand within the same pharmacological class, and these may be specifically stabilised in the procedure depending on the chemical class of the selected ligand.

Typically the selected ligand binds to the mutant GPCR with a similar potency to its binding to the parent GPCR. Typically, the K_d values for the particular ligand binding the mutant GPCR and the parent GPCR are within 5-10 fold of each other, such as within 2-3 fold. Typically, the binding of the ligand to the mutant GPCR compared to the parent GPCR would be not more than 5 times weaker and not more than 10 times stronger.

Typically, mutant receptors which have been stabilised in the selected conformation should bind the selected ligand with approximately equal affinity (that is to say typically within 2-3 fold) or greater affinity than does the parent receptor. For agonist-conformation mutants, the mutants typically bind the agonists with the same or higher affinity than the parent GPCR and typically bind antagonists with the same or lower affinity than the parent GPCR. Similarly for antagonist-conformation mutants, the mutants typically bind the antagonists with the same or higher affinity than the parent GPCR and typically bind agonists with the same or lower affinity than the parent GPCR.

Mutants that exhibit a significant reduction (typically greater than 2-3 fold) in affinity for the selecting ligand are typically rejected.

Typically, the rank order of binding of a set of ligands of the same class are comparable, although there may be one or two reversals in the order, or there may be an out-lier from the set.

In a further embodiment, two or more ligands that bind simultaneously to the receptor in the same conformation may be used, for example an allosteric modulator and orthosteric agonist.

For the avoidance of doubt, and as is evident from the Examples, it is not necessary to use multiple ligands for the method to be effective.

In a further embodiment, it may be advantageous to select those mutant GPCRs which, while still being able to bind the selected ligand, are not able to bind, or bind less strongly than the parent GPCR, a second selected ligand which is in a different class to the first ligand. Thus, for example, the mutant GPCR may be one that is selected on the basis that it has increased stability with respect to binding a selected antagonist, but the mutant GPCR so selected is further tested to determine whether it binds to a full agonist (or binds less strongly to a full agonist than its parent GPCR). Mutants are selected which do not bind (or have reduced binding of) the full agonist.

In this way, further selection is made of a GPCR which is locked into one particular conformation.

It will be appreciated that the selected ligand (with respect to part (b) of the method) and the further (second) ligand as discussed above, may be any pair of ligand classes, for example: antagonist and full agonist; full agonist and antagonist; antagonist and inverse agonist; inverse agonist and antagonist; inverse agonist and full agonist; full agonist and inverse agonist; and so on.

It is preferred that the mutant receptor binds the further (second) ligand with an affinity which is less than 50% of the affinity the parent receptor has for the same further (second) ligand, more preferably less than 10% and still more preferably less than 1% or 0.1% or 0.01% of affinity for the parent receptor. Thus, the K_d for the interaction of the second ligand with mutant receptor is higher than for the parent receptor. As is shown in Example 1, the mutant β-adrenergic receptor βAR-m23 (which was selected by the method of the invention using an antagonist) binds an agonist 3 orders of magnitude more weakly than its parent (ie K_d is 1000× higher). Similarly, in Example 2, the mutant adenosine A2a receptor Rant21 binds agonist 2-4 orders of magnitude more weakly than its parent.

This type of counter selection is useful because it can be used to direct the mutagenesis procedure more specifically (and therefore more rapidly and more efficiently) along a pathway towards a pure conformation as defined by the ligand.

Preferably, the mutant GPCR is provided in a suitable solubilised form in which it maintains structural integrity and is in a functional form (eg is able to bind ligand). An appropriate solubilising system, such as a suitable detergent (or other amphipathic agent) and buffer system is used, which may be chosen by the person skilled in the art to be effective for the particular protein. Typical detergents which may be used include, for example, dodecylmaltoside (DDM) or CHAPS or octylglucoside (OG) or many others. It may be convenient to include other compounds such as cholesterol hemisuccinate or cholesterol itself or heptane-1,2,3-triol. The presence of glycerol or proline or betaine may be useful. It is important that the GPCR, once solubilised from the membrane in which it resides, must be sufficiently stable to be assayed. For some GPCRs, DDM will be sufficient, but glycerol or other polyols may be added to increase stability for assay purposes, if desired. Further stability for assay purposes may be achieved, for example, by solubilising in a mixture of DDM, CHAPS and cholesterol hemisuccinate, optionally in the presence of glycerol. For particularly unstable GPCRs, it may be desirable to solubilise them using digitonin or amphipols or other polymers which can solubilise GPCRs directly from the membrane, in the absence of traditional detergents and maintain stability typically by allowing a significant number of lipids to remain associated with the GPCR. Nanodiscs may also be used for solubilising extremely unstable membrane proteins in a functional form.

Typically, the mutant GPCR is provided in a crude extract (eg of the membrane fraction from the host cell in which it has been expressed, such as E. coli). It may be provided in a form in which the mutant protein typically comprises at least 75%, more typically at least 80% or 85% or 90% or 95% or 98% or 99% of the protein present in the sample. Of course, it is typically solubilised as discussed above, and so the mutant GPCR is usually associated with detergent molecules and/or lipid molecules.

A mutant GPCR may be selected which has increased stability to any denaturant or denaturing condition such as to any one or more of heat, a detergent, a chaotropic agent or an extreme of pH.

In relation to an increased stability to heat (ie thermostability), this can readily be determined by measuring ligand binding or by using spectroscopic methods such as fluorescence, CD or light scattering at a particular temperature. Typically, when the GPCR binds to a ligand, the ability of the GPCR to bind that ligand at a particular temperature may be used to determine thermostability of the mutant. It may be convenient to determine a “quasi T_m” ie the temperature at which 50% of the receptor is inactivated under stated conditions after incubation for a given period of time (eg 30 minutes). Mutant GPCRs of higher thermostability have an increased quasi Tm compared to their parents.

In relation to an increased stability to a detergent or to a chaotrope, typically the GPCR is incubated for a defined time in the presence of a test detergent or a test chaotropic agent and the stability is determined using, for example, ligand binding or a spectroscopic method as discussed above.

In relation to an extreme of pH, a typical test pH would be chosen (eg in the range 4.5 to 5.5 (low pH) or in the range 8.5 to 9.5 (high pH).

Because relatively harsh detergents are used during crystallisation procedures, it is preferred that the mutant GPCR is stable in the presence of such detergents. The order of “harshness” of certain detergents is DDM, C₁₁→C₁₀→C₉→C₈ maltoside or glucoside, lauryldimethylamine oxide (LDAO) and SDS. It is particularly preferred if the mutant GPCR is more stable to any of C₉ maltoside or glucoside, C₈ maltoside or glucoside, LDAO and SDS, and so it is preferred that these detergents are used for stability testing.

Because of its ease of determination, it is preferred that thermostability is determined, and those mutants which have an increased thermostability compared to the parent protein with respect to the selected condition are chosen. It will be appreciated that heat is acting as the denaturant, and this can readily be removed by cooling the sample, for example by placing on ice. It is believed that thermostability may also be a guide to the stability to other denaturants or denaturing conditions. Thus, increased thermostability is likely to translate into stability in denaturing detergents, especially those that are more denaturing than DDM, eg those detergents with a smaller head group and a shorter alkyl chain and/or with a charged head group. We have found that a thermostable GPCR is also more stable towards harsh detergents.

When an extreme of pH is used as the denaturing condition, it will be appreciated that this can be removed quickly by adding a neutralising agent. Similarly, when a chaotrope is used as a denaturant, the denaturing effect can be removed by diluting the sample below the concentration in which the chaotrope exerts its chaotropic effect.

In a particular embodiment of the invention, the GPCR is β-adrenergic receptor (for example from turkey) and the ligand is dihydroalprenolol (DHA), an antagonist.

In a further preferred embodiment of the invention, the GPCR is the adenosine A_2a receptor (A_2aR) (for example, from man) and the ligand is ZM 241385 (4-[2-[[7-amino-2-(2-furyl) [1,2,4]-triazolo[2,3-α][1,3,5]triazin-5-yl]amino]ethyl]phenol), an antagonist or NECA (5′-N-ethylcarboxamido adenosine), an agonist.

In a still further preferred embodiment, the GPCR is the neurotensin receptor (NTR) (for example, from rat) and the ligand is neurotensin, an agonist.

A second aspect of the invention provides a method for preparing a mutant GPCR, the method comprising

• • (a) carrying out the method of the first aspect of the invention,
  • (b) identifying the position or positions of the mutated amino acid residue or residues in the mutant GPCR or GPCRs which has been selected for increased stability, and
  • (c) synthesising a mutant GPCR which contains a mutation at one or more of the positions identified.

As can be seen in the Examples, surprisingly, changes to a single amino acid within the GPCR may increase the stability of the protein compared to the parent protein with respect to a particular condition in which the protein resides in a particular conformation. Thus, in one embodiment of the method of the second aspect of the invention, a single amino acid residue of the parent protein is changed in the mutant protein. Typically, the amino acid residue is changed to the amino acid residue found in the mutant tested in the method of the first aspect of the invention. However, it may be replaced by any other amino acid residue, such as any naturally-occurring amino acid residue (in particular, a “codeable” amino acid residue) or a non-natural amino acid. Generally, for convenience, the amino acid residue is replaced with one of the 19 other codeable amino acids. Preferably, it is the replaced amino acid residue which is present in the mutant selected in the first aspect of the invention.

Also as can be seen in the Examples, a further increase in stability may be obtained by replacing more than one of the amino acids of the parent protein. Typically, each of the amino acids replaced is one which has been identified using the method of the first aspect of the invention. Typically, each amino acid identified is replaced by the amino acid present in the mutant protein although, as noted above, it may be replaced with any other amino acid.

Typically, the mutant GPCR contains, compared to the parent protein, from 1 to 10 replaced amino acids, preferably from 1 to 8, typically from 2 to 6 such as 2, 3, 4, 5 or 6 replaced amino acids.

It will be appreciated that the multiple mutants may be subject to the selection method of the first aspect of the invention. In other words, multiple mutants may be provided in step (a) of the method of the first aspect of the invention. It will be appreciated that by the first and/or second aspect of the invention multiply mutagenised GPCRs may be made, whose conformation has been selected to create a very stable multiple point mutant protein.

The mutant GPCRs may be prepared by any suitable method. Conveniently, the mutant protein is encoded by a suitable nucleic acid molecule and expressed in a suitable host cell. Suitable nucleic acid molecules encoding the mutant GPCR may be made using standard cloning techniques, site-directed mutagenesis and PCR as is well known in the art. Suitable expression systems include constitutive or inducible expression systems in bacteria or yeasts, virus expression systems such as baculovirus, semliki forest virus and lentiviruses, or transient transfection in insect or mammalian cells. Suitable host cells include E. coli, Lactococcus lactis, Saccharomyces cerevisiae, Schizosaccharomyces pombe, Pichia pastoris, Spodoptera frugiperda and Trichoplusiani cells. Suitable animal host cells include HEK 293, COS, S2, CHO, NSO, DT40 and so on. It is known that some GPCRs require specific lipids (eg cholesterol) to function. In that case, it is desirable to select a host cell which contains the lipid. Additionally or alternatively the lipid may be added during isolation and purification of the mutant protein. It will be appreciated that these expression systems and host cells may also be used in the provision of the mutant GPCR in part (a) of the method of the first aspect of the invention.

Molecular biological methods for cloning and engineering genes and cDNAs, for mutating DNA, and for expressing polypeptides from polynucleotides in host cells are well known in the art, as exemplified in “Molecular cloning, a laboratory manual”, third edition, Sambrook, J. & Russell, D. W. (eds), Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y., incorporated herein by reference.

In a further embodiment of the first or second aspect of the invention it is determined whether the selected or prepared mutant GPCR is able to couple to a G protein. It is also preferred if it is determined whether the selected or prepared mutant GPCR is able to bind a plurality of ligands of the same class as the selecting ligand with a comparable spread and/or rank order of affinity as the parent GPCR.

A third aspect of the invention provides a mutant GPCR prepared by the method of the second aspect of the invention.

The invention includes mutant GPCRs with increased stability compared to their parent GPCRs, particularly those with increased thermostability.

Mutant β-adrenergic Receptor

β-adrenergic receptors are well known in the art. They share sequence homology to each other and bind to adrenalin.

A fourth aspect of the invention provides a mutant β-adrenergic receptor which, when compared to the corresponding wild-type β-adrenergic receptor, has a different amino acid at a position which corresponds to any one or more of the following positions according to the numbering of the turkey β-adrenergic receptor as set out in FIG. 9: Ile 55, Gly 67, Arg 68, Val 89, Met 90, Gly 98, Ile 129, Ser 151, Val 160, Gln 194, Gly 197, Leu 221, Tyr 227, Arg 229, Val 230, Ala 234, Ala 282, Asp 322, Phe 327, Ala 334, Phe 338.

The mutant β-adrenergic receptor may be a mutant of any β-adrenergic receptor provided that it is mutated at one or more of the amino acid positions as stated by reference to the given turkey β-adrenergic receptor amino acid sequence.

It is particularly preferred if the mutant GPCR is one which has at least 20% amino acid sequence identity when compared to the given turkey β-adrenergic receptor sequence, as determined using MacVector and CLUSTALW (Thompson et al (1994) Nucl. Acids Res. 22, 4673-4680). More preferably, the mutant receptor has at least 30% or at least 40% or at least 50% amino acid sequence identity. There is generally a higher degree of amino acid sequence identity which is conserved around the orthosteric (“active”) site to which the natural ligand binds.

As is described in Example 1 and FIG. 1 below, individual replacement of the following amino acid residues in the parent turkey β-adrenergic sequence (as shown in FIG. 9) lead to an increase in thermostability: Ile 55, Gly 67, Arg 68, Val 89, Met 90, Gly 98, Ile 129, Ser 151, Val 160, Gln 194, Gly 197, Leu 221, Tyr 227, Arg 229, Val 230, Ala 234, Ala 282, Asp 322, Phe 327, Ala 334, Phe 338.

Thus, the invention includes mutant turkey β-adrenergic receptors in which, compared to its parent, one or more of these amino acid residues have been replaced by another amino acid residue. The invention also includes mutant β-adrenergic receptors from other sources in which one or more corresponding amino acids in the parent receptor are replaced by another amino acid residue. For the avoidance of doubt, the parent may be a β-adrenergic receptor which has a naturally-occurring sequence, or it may be a truncated form or it may be a fusion, either to the naturally occurring protein or to a fragment thereof, or it may contain mutations compared to the naturally-occurring sequenced provided that it retains ligand-binding ability.

By “corresponding amino acid residue” we include the meaning of the amino acid residue in another β-adrenergic receptor which aligns to the given amino acid residue in turkey β-adrenergic receptor when the turkey β-adrenergic receptor and the other β-adrenergic receptor are compared using MacVector and CLUSTALW.

FIG. 9 shows an alignment between turkey β-adrenergic receptor and human β1, β2 and β3 β-adrenergic receptors.

It can be seen that Ile 72 of human β1 corresponds to Ile 55 of turkey β-adrenergic receptor; Ile 47 of human β2 corresponds to Ile 55 of turkey β-adrenergic receptor; and Thr51 of human β3 corresponds to Ile 55 of turkey β-adrenergic receptor. Other corresponding amino acid residues in human β1, β2 and β3 can readily be identified by reference to FIG. 9.

It is preferred that the particular amino acid is replaced with an Ala. However, when the particular amino acid residue is an Ala, it is preferred that it is replaced with a Leu (for example, see turkey β-adrenergic Ala 234, Ala 282 and Ala 334 in FIG. 1).

It is preferred if the mutant β-adrenergic receptor has a different amino acid compared to its parent at more than one amino acid position since this is likely to give greater stability. Particularly preferred human β1 receptor mutants are those in which one or more of the following amino acid residues are replaced with another amino acid residue: K85, M107, Y244, A316, F361 and F372. Typically, the given amino acid residue is replaced with Ala or Val or Met or Leu or Ile (unless they are already that residue).

Mutant human β1 receptors which have combinations of 3 or 4 or 5 or 6 mutations as described above are prepared.

Particularly preferred human β2 receptor mutants are those in which one or more of the following amino acids are replaced with another amino acid residue: K60, M82, Y219, C265, L310 and F321. Typically, the given amino acid residue is replaced with Ala or Val or Met or Leu or Ile (unless they are already that residue).

Mutant human β2 receptors which have combinations of 3 or 4 or 5 or 6 mutations as described above are preferred.

FIG. 26 shows the effect on thermostability when six thermostabilising mutations in β1-m23 (R68S, M90V, Y227A, A282L, F327A, F338M) were transferred directly to the human β2 receptor (equivalent mutations K60S, M82V, Y219A, C265L, L310A, F321M), making human β2-m23. The Tms for human β2 and β2-m23 were 29° C. and 41° C. respectively, thus exemplifying the transferability of thermostabilising mutations from one receptor to another receptor. Accordingly, a particularly preferred human β2 receptor mutant is one which comprises the mutations K60S, M82V, Y219A, C265L, L310A, F321M.

Particularly preferred human β3 receptor mutants are those in which one or more of the following amino acids are replaced with another amino acid residue: W64, M86, Y224, P284, A330 and F341. Typically, the given amino acid residue is replaced with Ala or Val or Met or Leu or Ile (unless they are already that residue).

Mutant human β3 receptors which have combinations of 3 or 4 or 5 or 6 mutations as described above are preferred.

Particularly preferred combinations of mutations are described in detail in Tables 1 and 2 in Example 1, and the invention includes the mutant turkey β-adrenergic receptors, and also includes mutant β-adrenergic receptors where amino acids in corresponding position have been replaced by another amino acid, typically the same amino acid as indicated in Tables 1 and 2 in Example 1.

Particularly preferred mutants are those which contain mutations in the amino acids which correspond to the given amino acid residue by reference to turkey β-adrenergic receptor: (R68S, Y227A, A282L, A334L) (see m6-10 in Table 2 below); (M90V, Y227A, F338M) (see m7-7 in Table 2 below); (R68S, M90V, V230A, F327A, A334L) (see m10-8 in Table 2 below); and (R68S, M90V, Y227A, A282L, F327A, F338M) (see m23 in Table 2 below).

Mutant Adenosine Receptor

Adenosine receptors are well known in the art. They share sequence homology to each other and bind to adenosine.

A fifth aspect of the invention provides a mutant adenosine receptor which, when compared to the corresponding wild-type adenosine, has a different amino acid at a position which corresponds to any one or more of the following positions according to the numbering of the human adenosine A_2a receptor as set out in FIG. 10: Gly 114, Gly 118, Leu 167, Ala 184, Arg 199, Ala 203, Leu 208, Gln 210, Ser 213, Glu 219, Arg 220, Ser 223, Thr 224, Gln 226, Lys 227, His 230, Leu 241, Pro 260, Ser 263, Leu 267, Leu 272, Thr 279, Asn 284, Gln 311, Pro 313, Lys 315, Ala 54, Val 57, His 75, Thr 88, Gly 114, Gly 118, Thr 119, Lys 122, Gly 123, Pro 149, Glu 151, Gly 152, Ala 203, Ala 204, Ala 231, Leu 235, Val 239.

The mutant adenosine receptor may be a mutant of any adenosine receptor provided that it is mutated at one or more of the amino acid positions as stated by reference to the given human adenosine A_2a receptor amino acid sequence.

It is particularly preferred if the mutant GPCR is one which has at least 20% amino acid sequence identity when compared to the given human adenosine A_2a receptor sequence, as determined using MacVector and CLUSTALW. Preferably, the mutant GPCR has at least 30% or at least 40% or at least 50% or at least 60% sequence identity. Typically, there is a higher degree of sequence conservation at the adenosine binding site.

As is described in Example 2 below, individual replacement of the following amino acid residues in the human adenosine A_2a receptor sequence (as shown in FIG. 10) lead to an increase in thermostability when measured with the agonist 5′-N-ethylcarboxamidoadenosine (NECA):

Gly 114, Gly 118, Leu 167, Ala 184, Arg 199, Ala 203, Leu 208, Gln 210, Ser 213, Glu 219, Arg 220, Ser 223, Thr 224, Gln 226, Lys 227, His 230, Leu 241, Pro 260, Ser 263, Leu 267, Leu 272, Thr 279, Asn 284, Gln 311, Pro 313, Lys 315.

Replacement of the following amino acid residues in the human A_2a receptor sequence (as shown in FIG. 10) lead to an increase in thermostability when measured with the antagonist ZM 241385 (4-[2-[[7-amino-2-(2-furyl) [1,2,4]-triazolo[2,3-α][1,3,5]triazin-5-yl]amino]ethyl]phenol):

Ala 54, Val 57, His 75, Thr 88, Gly 114, Gly 118, Thr 119, Lys 122, Gly 123, Pro 149, Glu 151, Gly 152, Ala 203, Ala 204, Ala 231, Leu 235, Val 239.

Thus, the invention includes mutant human adenosine A_2a receptors in which, compared to its parent, one or more of these amino acid residues have been replaced by another amino acid residue. The invention also includes mutant adenosine receptors from other sources in which one or more corresponding amino acids in the parent receptor are replaced by another amino acid residue. For the avoidance of doubt, the parent may be an adenosine receptor which has a naturally-occurring sequence, or it may be a truncated form or it may be a fusion, either to the naturally-occurring protein or to a fragment thereof, or it may contain mutations compared to the naturally-occurring sequence, provided that it retains ligand-binding ability.

By “corresponding amino acid residue” we include the meaning of the amino acid residue in another adenosine receptor which aligns to the given amino acid residue in human adenosine A_2a receptor when the human adenosine A_2a receptor and the other adenosine receptor are compared using MacVector and CLUSTALW.

FIG. 10 shows an alignment between human adenosine A_2a receptor and three other human adenosine receptors (A2b, A3 and A1).

It can be seen that, for example, Ser 115 in the A_2b receptor (indicated as AA2BR) corresponds to Gly 114 in the A_2a receptor. Similarly, it can be seen that Ala 60 in the A₃ receptor (indicated as AA3R) corresponds to Ala 54 in the A_2a receptor, and so on. Other corresponding amino acid residues in human adenosine receptors A_2b, A₃ and A₁ can readily be identified by reference to FIG. 10.

It is preferred that the particular amino acid in the parent is replaced with an Ala. However, when the particular amino acid residue in the parent is an Ala, it is preferred that it is replaced with a Leu.

It is preferred that the mutant adenosine receptor has a different amino acid compared to its parent at more than one amino acid position. Particularly preferred human adenosine A2b receptors are those in which one or more of the following amino acid residues are replaced with another amino acid residue: A55, T89, R123, L236 and V240. Typically, the given amino acid residue is replaced with Ala or Val or Met or Leu or Ile (unless they are already that residue).

Mutant human adenosine A2b receptors which have combinations of 3 or 4 or 5 mutations as described above are preferred.

Particularly preferred human adenosine A3 receptors are those in which one or more of the following amino acid residues are replaced with another amino acid residue: A60, T94, W128, L232 and L236. Typically, the given amino acid residue is replaced with Ala or Val or Met or Leu or Ile (unless they are already that residue).

Mutant human adenosine A3 receptors which have combinations of 3 or 4 or 5 mutations as described above are preferred.

Particular preferred human adenosine A1 receptors are those in which one or more of the following residues are replaced: A57, T91, A125, L236, and L240. Typically, the given amino acid residue is replaced with Ala or Val or Met or Leu or Ile (unless they are already that residue).

Particularly preferred combinations of mutations are described in detail in Example 2. The invention includes these mutant human adenosine A_2a receptors, and also includes other mutant adenosine receptors where amino acids in corresponding positions have been replaced by another amino acid, typically the same amino acid as indicated in Example 2.

Particularly preferred adenosine receptor mutants are those which contain mutations in the amino acids which correspond to the given amino residue by reference to human adenosine A2a receptor: (A54L, K122A, L235A) (Rant 17); (A54L, T88A, V239A, A204L) (Rant 19); and (A54L, T88A, V239A, K122A) (Rant 21).

Mutant Neurotensin Receptor

Neurotensin receptors are known in the art. They share sequence homology and bind neurotensin.

A sixth aspect of the invention provides a mutant neurotensin receptor which, when compared to the corresponding wild-type neurotensin receptor, has a different amino acid at a position which corresponds to any one or more of the following positions according to the numbering of the rat neurotensin receptor as set out in FIG. 11: Ala 69, Leu 72, Ala 73, Ala 86, Ala 90, Ser 100, His 103, Ser 108, Leu 109, Leu 111, Asp 113, Ile 116, Ala 120, Asp 139, Phe 147, Ala 155, Val 165, Glu 166, Lys 176, Ala 177, Thr 179, Met 181, Ser 182, Arg 183, Phe 189, Leu 205, Thr 207, Gly 209, Gly 215, Val 229, Met 250, Ile 253, Leu 256, Ile 260, Asn 262, Val 268, Asn 270, Thr 279, Met 293, Thr 294, Gly 306, Leu 308, Val 309, Leu 310, Val 313, Phe 342, Asp 345, Tyr 349, Tyr 351, Ala 356, Phe 358, Val 360, Ser 362, Asn 370, Ser 373, Phe 380, Ala 385, Cys 386, Pro 389, Gly 390, Trp 391, Arg 392, His 393, Arg 395, Lys 397, Pro 399.

It is particularly preferred if the mutant GPCR is one which has at least 20% amino acid sequence identity when compared to the given rat neurotensin receptor sequence, as determined using MacVector and CLUSTALW. Preferably, the mutant GPCR has at least 30% or at least 40% or at least 50% amino acid sequence identity.

The mutant neurotensin receptor may be a mutant of any neurotensin receptor provided that it is mutated at one or more of the amino acid positions as stated by reference to the given rat neurotensin receptor amino acid sequence.

As is described in Example 3 below, individual replacement of the following amino acid residues in the rat neurotensin receptor sequence (as shown in FIGS. 11 and 28) lead to an increase in thermostability when considered with respect to the absence of neurotensin. Leu 72, Ala 86, Ala 90, Ser 100, His 103, Ser 108, Leu 109, Leu 111, Asp 113, Ile 116, Ala 120, Asp 139, Phe 147, Ala 155, Lys 176, Thr 179, Met 181, Ser 182, Phe 189, Leu 205, Thr 207, Gly 209, Gly 215, Leu 256, Asn 262, Val 268, Met 293, Asp 345, Tyr 349, Tyr 351, Ala 356, Phe 358, Ser 362, Ala 385, Cys 386, Trp 391, Arg 392, His 393, Lys 397, Pro 399.

As is described in Example 3 below, individual replacement of the following amino acid residues in the rat neurotensin receptor sequence (as shown in FIGS. 11 and 28) lead to an increase in thermostability when considered with respect to the presence of neurotensin. Ala 69, Ala 73, Ala 86, Ala 90, His 103, Val 165, Glu 166, Ala 177, Arg 183, Gly 215, Val 229, Met 250, Ile 253, Ile 260, Thr 279, Thr 294, Gly 306, Leu 308, Val 309, Leu 310, Val 313, Phe 342, Phe 358, Val 360, Ser 362, Asn 370, Ser 373, Phe 380, Ala 385, Pro 389, Gly 390, Arg 395.

Thus, the invention includes mutant rat neurotensin receptor in which, compared to its parent, one or more of these amino acid residues have been replaced by another amino acid residue. The invention also includes mutant neurotensin receptors from other sources in which one or more corresponding amino acids in the parent receptor are replaced by another amino acid residue. For the avoidance of doubt the parent may be a neurotensin receptor which has a naturally-occurring sequence, or it may be a truncated form or it may be a fusion, either to the naturally-occurring protein or to a fragment thereof, or it may contain mutations compared to the naturally-occurring sequence, providing that it retains ligand-binding ability.

By “corresponding amino acid residue” we include the meaning of the amino acid residue in another neurotensin receptor which aligns to the given amino acid residue in rat neurotensin receptor when the rat neurotensin receptor and the other neurotensin receptor are compared using MacVector and CLUSTALW.

FIG. 11 shows an alignment between rat neurotensin receptor and two human neurotensin receptors 1 and 2. It can be seen, for example, that Ala 85 of the human neurotensin receptor 1 corresponds to Ala 86 of the rat neurotensin receptor, that Phe 353 of the human neurotensin receptor 1 corresponds to Phe 358 of the rat neurotensin receptor, and so on. Other corresponding amino acid residue in the human neurotensin receptors 1 and 2 can readily be identified by reference to FIG. 11.

It is preferred that the particular amino acid in the parent is replaced with an Ala. However, when the particular amino acid residue in the parent is an Ala, it is preferred that it is replaced with a Leu.

It is preferred that the mutant neurotensin receptor has a different amino acid compared to its parent at more than one amino acid position. Particularly preferred human neurotensin receptors (NTR1) are those in which one or more of the following amino acid residues are replaced with another amino acid residue: Ala 85, His 102, Ile 259, Phe 337 and Phe 353. Typically, the given amino acid residues is replaced with Ala or Val or Met or Leu or Ile (unless they are already that residue).

Mutant human neurotensin receptors (NTR1) which have combinations of 3 or 4 or 5 mutations as described above are preferred.

Particularly preferred human neurotensin receptors (NTR2) are those in which one or more of the following amino acid residues are replaced with another amino acid residue: V54, R69, T229, P331 and F347. Typically, the given amino acid residue is replaced with Ala or Val or Met or Leu or Ile (unless they are already that residue). Mutant human neurotensin receptors (NTR2) which have combinations of 3 or 4 or 5 mutations as described above are preferred.

Particularly preferred combinations of mutations are described in detail in Example 3. The invention includes these mutant rat neurotensin receptors, and also includes other mutant neurotensin receptors where amino acids in corresponding positions have been replaced by another amino acid, typically the same amino acid as indicated in Example 3.

Particularly preferred neurotensin receptor mutants are those which contain mutations in the amino acid residues which correspond to the given amino acid residue by reference to the rat neurotensin receptor: (F358A, A86L, I260A, F342A) (Nag7m); (F358A, H103A, I260A, F342A) (Nag7n).

Mutant Muscarinic Receptor

Muscarinic receptors are known in the art. They share sequence homology and bind muscarine.

A seventh aspect of the invention provides a mutant muscarinic receptor which, when compared to the corresponding wild-type muscarinic receptor, has a different amino acid at a position which corresponds to any one or more of the following positions according to the numbering of the human muscarinic receptor M1 as set out in FIG. 17: Leu 65, Met 145, Leu 399, Ile 383 and Met 384.

It is particularly preferred if the mutant GPCR is one which has at least 20% amino acid sequence identity when compared to the given human muscarinic receptor sequence, as determined using MacVector and CLUSTALW. Preferably, the mutant GPCR has at least 30% or at least 40% or at least 50% amino acid sequence identity.

The mutant muscarinic receptor may be a mutant of any muscarinic receptor provided that it is mutated at one or more of the amino acid positions as stated by reference to the given muscarinic receptor amino acid sequence.

Thus, the invention includes a mutant human muscarinic receptor in which, compared to its parent, one or more of these amino acid residues have been replaced by another amino acid residue. The invention also includes mutant muscarinic receptors from other sources in which one or more corresponding amino acids in the parent receptor are replaced by another amino acid residue. For the avoidance of doubt the parent may be a muscarinic receptor which has a naturally-occurring sequence, or it may be a truncated form or it may be a fusion, either to the naturally-occurring protein or to a fragment thereof, or it may contain mutations compared to the naturally-occurring sequence, providing that it retains ligand-binding ability.

By “corresponding amino acid residue” we include the meaning of the amino acid residue in another muscarinic receptor which aligns to the given amino acid residue in human muscarinic receptor when the human muscarinic receptor and the other muscarinic receptor are compared using MacVector and CLUSTALW.

It is preferred that the particular amino acid is replaced with an Ala. However, when the particular amino acid residue is an Ala, it is preferred that it is replaced with a Leu.

As shown in Examples 1-3 and described above, we have identified thermostabilising mutations scattered widely throughout the sequences of the turkey beta1 adrenergic receptor, human adenosine receptor, rat neurotensin receptor and human muscarinic receptor. FIG. 17 provides an alignment of these sequences with the sequence of the human beta-2AR such that when the thermostabilising mutations are positioned onto the sequences then, in 11 instances out of a total of 70, two sequences contain mutations at the same position (denoted in FIG. 17 with a star). Thus it will be appreciated that once one or more stabilising mutations have been identified in one GPCR, a further GPCR with increased stability can be generated by aligning the amino acid sequences of the GPCRs and making the same one or more mutations at the corresponding position or positions. This concept is clearly exemplified in FIG. 26 wherein the six thermostabilising mutations in turkey β1-m23 were transferred directly to the human β2 receptor. The resultant mutant, β2-m23, had a Tm 12° C. higher than that of the human β2 receptor.

Accordingly, an eighth aspect of the invention provides a method for producing a mutant GPCR with increased stability relative to its parent GPCR, the method comprising:

• • (i) identifying in the amino acid sequence of one or more mutants of a first parent GPCR with increased stability relative to the first parent GPCR, the position or positions at which the one or more mutants have at least one different amino acid residue compared to the first parent GPCR, and
  • (ii) making one or more mutations in the amino acid sequence that defines a second GPCR at the corresponding position or positions, to provide one or more mutants of a second parent GPCR with increased stability relative to the second parent GPCR.

The one or more mutants of a first parent GPCR may be selected or prepared according to the methods of the first or second aspects of the invention. Accordingly, it will be appreciated that the one or more mutants of a first parent GPCR may be any of the mutants of the third, fourth, fifth, sixth or seventh aspects of the invention. Hence, the method of the eighth aspect of the invention may be used to create stable, conformationally locked GPCRs by mutagenesis. For example, following the selection of mutant GPCRs which have increased stability in a particular conformation, the stabilising mutation can be identified and the amino acid at a corresponding position in a second GPCR replaced to produce a mutant GPCR with increased stability in a particular conformation relative to a second parent GPCR.

For the avoidance of doubt the first parent GPCR may be a GPCR which has a naturally-occurring sequence, or it may be a truncated form or it may be a fusion, either to the naturally-occurring protein or to a fragment thereof, or it may contain mutations compared to the naturally-occurring sequence, providing that it retains ligand-binding ability.

Typically, identifying the position or positions at which the one or more mutants have at least one different amino acid residue compared to the first parent GPCR involves aligning their amino acid sequences with that of the parent GPCR, for example using the ClustalW program (Thompson et al., 1994).

By “corresponding position or positions”, we include the meaning of the position in the amino acid sequence of a second GPCR which aligns to the position in the amino acid sequence of the first GPCR, when the first and second GPCRs are compared by alignment, for example by using MacVector and ClustalW. For example, as shown in the alignment in FIG. 17, the six stabilising mutations in turkey β1-m23, R68S, M90V, Y227A, A282L, F327A and F338M, are at positions which correspond to residues K60, M82, Y219, C265, L310 and F321 respectively in the human β2 receptor.

Having identified the corresponding position or positions in the amino acid sequence of a second GPCR, the amino acids at those positions are replaced with another amino acid. Typically, the amino acids are replaced with the same amino acids which replaced the amino acids at the corresponding positions in the mutant of the first parent GPCR (unless they are already that residue). For example, at position 68 in turkey β1-m23 (R68S), an arginine residue was replaced with a serine residue. Therefore, at the corresponding position in the human β2 receptor, position 60 (K60), the lysine residue is preferably replaced with a serine residue.

Mutations can be made in an amino acid sequence, for example, as described above and using techniques well-established in the art.

It will be appreciated that the second GPCR may be any other GPCR. For example, stabilising mutations in a GPCR from one species may be transferred to a second GPCR from another species. Similarly, stabilising mutations in one particular GPCR isoform may be transferred to a second GPCR which is a different isoform. Preferably, the second parent GPCR is of the same GPCR class or family as the first parent GPCR. Phylogenetic analyses have divided GPCRs into three main classes based on protein sequence similarity, i.e., classes 1, 2, and 3 whose prototypes are rhodopsin, the secretin receptor, and the metabotropic glutamate receptors, respectively (Foord et al (2005) Pharmacol. Rev. 57, 279-288). Thus, the second GPCR may be a GPCR which is of the same GPCR class as the first parent GPCR. Similarly, GPCRs have been divided into families by reference to natural ligands such as glutamate and GABA. Thus, the second GPCR may be of the same GPCR family as the first parent GPCR. A list of GPCR classes and families has been produced by the International Union of Pharmacology (Foord et al (2005) Pharmacol. Rev. 57, 279-288) and this list is periodically updated at www.iuphar-db.org/GPCR/ReceptorFamiliesForward.

It will be appreciated that the second parent GPCR must be able to be aligned with the first parent GPCR such that the corresponding positions of the mutations in the first GPCR can be determined in the second GPCR. Thus typically, the second parent GPCR has at least 20% sequence identity to the first parent GPCR and more preferably at least 30%, 40%, 50%, 60%, 70%, 80% or 90% sequence identity to the first parent GPCR. However, some GPCRs have low sequence identity (e.g. family B and C GPCRs) and at the same time are very similar in structure. Thus the 20% sequence identity threshold is not absolute.

The inventors have reasoned that the identification of structural motifs in which the one or more mutations in a mutant GPCR with increased stability reside, will be useful in producing further mutant GPCRs with increased stability.

Accordingly, a ninth aspect of the invention provides a method for producing a mutant G-protein coupled receptor (GPCR) with increased stability relative to its parent GPCR, the method comprising:

• • (i) providing one or more mutants of a first parent GPCR with increased stability relative to the first parent GPCR
  • (ii) identifying in a structural membrane protein model the structural motif or motifs in which the one or more mutants have at least one different amino acid residue compared to the first parent GPCR, and
  • (iii) making one or more mutations in the amino acid sequence that defines a corresponding structural motif or motifs in a second parent GPCR, to provide one or more mutants of a second parent GPCR with increased stability relative to the second parent GPCR.

Mapping stabilising mutations onto one or more known structural models can be used to identify particular structural motifs in which such stabilising mutations reside. We have mapped stabilising mutations of the β1-adrenergic receptor onto structural models of the β2-adrenergic receptor (Rasmussen et al (2007) Nature 450, 383-387; Cherezov et al (2007) Science 318:1258-65; Rosenbaum et al (2007) Science 318:1266-1273) in order to identify such motifs. For example, Table (vi) lists the turkey β1-adrenergic receptor mutations which we have mapped onto the human β2-adrenergic receptor and describes the corresponding structural motifs in which they reside. As discussed in Example 4, mapping of the Y227A mutation (equivalent to Y219 in the human β₂ receptor) onto the human β₂-adrenergic receptor reveals its position at the interface between helices such that the mutation may improve packing at the helical interface (see FIGS. 15, 16 and 23). Similarly, mapping of the M90V mutation (equivalent to M82 in the human β₂ receptor) onto the human β₂-adrenergic receptor reveals it to be in helix 2 at a point where the helix is kinked (see FIGS. 15, 16 and 20). Other mutations were found to reside in further structural motifs including transmembrane helix surfaces pointing into the lipid bilayer, hydrophobic-hydrophilic boundary regions, protein binding pockets and loop regions (see Table (vi) and FIGS. 18-19, 21-22 and 24-25).

Such structural motifs, by virtue of them containing stabilising mutations, are important in determining protein stability. Therefore, targeting mutations to these motifs will facilitate the generation of stabilised mutant GPCRs. Indeed, there were several instances where more than one mutation mapped to the same structural motif. For example, the Y227A, V230A and A234L mutations in the turkey β1 adrenergic receptor mapped to the same helical interface, the V89L and M90V mutations mapped to the same helical kink and the F327A and A334L mutations mapped to the same helical surface pointing towards the lipid bilayer (Table (vi)). Thus, when one stabilising mutation has been identified, the determination of the structural motif in which that mutation is located will enable the identification of further stabilising mutations.

In an embodiment of the ninth aspect of the invention, the one or more mutants of a first parent GPCR are selected or prepared according to the methods of the first, second or eighth aspects of the invention. Accordingly, it will be appreciated that the one or more mutants of a first parent GPCR may be any of the mutants of the third, fourth, fifth, sixth or seventh aspects of the invention. Hence, the method of the ninth aspect of the invention may also be used to create stable, conformationally locked GPCRs by mutagenesis. For example, following the selection of mutant GPCRs which have increased stability in a particular conformation, the structural motifs in which such stabilising mutations reside can be identified. Making one or more mutations in the amino acid sequence that defines the corresponding structural motif in another GPCR can then be used to produce a mutant GPCR with increased stability in a particular conformation relative to its parent GPCR.

We have performed a multiple sequence alignment of the human beta-2AR, rat NTR1, turkey beta-1 AR, human Adenosine A2aR and human muscarinic M1 receptor amino acid sequences (FIG. 17) which shows that, when the thermostabilising mutations identified (see Examples 1-3) are positioned on the sequences then, in 11 instances out of a total of 70, two sequences contain mutations at the same position (denoted in FIG. 17 with a star). Without wishing to be bound by any theory, the inventors believe that thermostabilising mutations at these positions should be of enhanced transferability for mapping onto a structural membrane protein model. Thus in one embodiment, the mutant of the first parent GPCR is a mutant human beta-2AR, rat NTR1, turkey beta-1 AR, human Adenosine A2aR or human muscarinic M1 receptor which, when compared to its corresponding parent receptor, has a different amino acid at a position which corresponds to any one or more of the following positions according to the numbering of the human beta2 AR as set out in FIG. 17: Ala 59, Val 81, Ser 143, Lys 147, Val 152, Glu 180, Val 222, Ala 226, Ala 271, Leu 275 and Val 317.

In order to identify the structural motif or motifs, the stabilising mutations are mapped onto a known structure of a membrane protein.

By “membrane protein” we mean a protein that is attached to or associated with a membrane of a cell or organelle. Preferably, the membrane protein is an integral membrane protein that is permanently integrated into the membrane and can only be removed using detergents, non-polar solvents or denaturing agents that physically disrupt the lipid bilayer.

The structural model of a membrane protein may be any suitable structural model. For example, the model may be a known crystal structure. Examples of GPCR crystal structures include bovine rhodopsin (Palczewski, K. et al., Science 289, 739-745. (2000)) and human β₂ adrenergic receptor (Rasmussen et al, Nature 450, 383-7 (2007); Cherezov et al (2007) Science 318:1258-65; Rosenbaum et al (2007) Science 318:1266-1273). The coordinates for the human β₂ adrenergic receptor structure can be found in the RCSB Protein Data Bank under accession codes: 2rh1, 2r4r and 2r4s. Alternatively, the structural model may be a computer generated model based upon homology or using de novo structure prediction methods (Qian et al Nature (2007) 450: 259-64).

It will be appreciated that stabilising mutations of a given mutant GPCR can be mapped onto a structural model of any membrane protein which has sufficient structural similarity to the GPCR. In particular, the domain of the membrane protein must have sufficient structural similarity to the GPCR domain in which the stabilising mutation resides, for a given mutation to be transferable.

A protein domain is typically defined as a discretely folded assembly of secondary structure elements which may stand alone as a single protein or be part of a larger protein in combination with other domains. It is commonly a functional evolutionary unit.

GPCRs are essentially single domain proteins excluding those with large N-terminal domains. Therefore, typically, the structural model is of a membrane protein which comprises at least one domain that has sufficient structural similarity to the GPCR.

Structural similarity can be determined indirectly by the analysis of sequence identity, or directly by comparison of structures.

With regard to sequence identity, the amino acid sequence encoding the GPCR domain in which the mutant has at least one different amino acid residue compared to the first parent GPCR, is aligned with an amino acid sequence encoding a domain of a membrane protein for which a structural model is available. It will be appreciated that one or more of these sequences may contain an inserted sequence or N-terminal or C-terminal extensions which are additional to the core conserved domain. For optimal alignment, such sequences are removed so as not to skew the analysis. Membrane proteins with sufficient sequence identity across the domain in question may then be used as the structural model for mapping mutations. It has been shown for soluble protein domains that their 3D structure is broadly conserved above 20% sequence identity and well conserved above 30% identity, with the level of structural conservation increasing as sequence identity increases up to 100% (Ginalski, K. Curr Op Struc Biol (2006) 16, 172-177). Thus, it is preferred if the structural membrane protein model is a model of a membrane protein which contains a domain that shares at least 20% sequence identity with the mutant GPCR domain containing the at least one different amino acid residue compared to the first parent GPCR, and more preferably at least 30%, 40%, 50%, 60%, 70%, 80% or 90% sequence identity, and yet more preferably at least 95% or 99% sequence identity.

Sequence identity may be measured by the use of algorithms such as BLAST or PSI-BLAST (Altschul et al, NAR (1997), 25, 3389-3402) or methods based on Hidden Markov Models (Eddy S et al, J Comput Biol (1995) Spring 2 (1) 9-23). Typically, the percent sequence identity between two polypeptides may be determined using any suitable computer program, for example the GAP program of the University of Wisconsin Genetic Computing Group and it will be appreciated that percent identity is calculated in relation to polypeptides whose sequence has been aligned optimally. The alignment may alternatively be carried out using the Clustal W program (Thompson et al., 1994). The parameters used may be as follows: Fast pairwise alignment parameters: K-tuple(word) size; 1, window size; 5, gap penalty; 3, number of top diagonals; 5. Scoring method: x percent. Multiple alignment parameters: gap open penalty; 10, gap extension penalty; 0.05. Scoring matrix: BLOSUM.

In addition to sequence identity, structural similarity can be determined directly by comparison of structural models. Structural models may be used to detect regions of structural similarity not evident from sequence analysis alone, and which may or may not be contiguous in the sequence. For example, family B and C GPCRs are thought to share similar structures; however, their sequence identity is very low. Similarly, the water transporting aquaporins spinach SoPip2, E. coli AqpZ, Methanococcus AqpM, rat Aqp4, human Aqp1 and sheep Aqp0 share low sequence identity but all have similar structures.

Structural models of high fidelity may be constructed for proteins of unknown structure using standard software packages such as MODELLER (Sali A and Blundell T, J Mol Biol (1993) 234(3) 779-815), wherein the structure is modelled on a known structure of a homologous protein. Such modelling improves with increasing sequence identity. Typically, the sequence identity between the sequence of unknown structure and a sequence of known 3D structure is more than 30% (Ginalski, K. Curr Op Struc Biol (2006) 16, 172-177). In addition, de novo structure prediction methods based on sequence alone may be used to model proteins of unknown structure (Qian et al, (2007) Nature 450:259-64). Once structures have been experimentally determined or derived by modelling, regions of structural similarity may be detected by direct comparison of two or more 3D structures. They may, for example, comprise secondary structure elements of a particular architecture and topology which can be detected by the use of software such as DALI (Holm, L and Sander, C (1996) Science 273, 595-603). They may comprise local arrangements of amino acid side chains and the polypeptide backbone, or specific sets of atoms or groups of atoms in a particular spatial arrangement, which may for example also be detected by the use of graph theoretical representations (Artymiuk, P et al, (2005) J Amer Soc Info Sci Tech 56 (5) 518-528). In this approach, the atoms or groups of atoms within the proteins or regions of proteins to be compared are typically represented as the nodes of a graph, with the edges of the graph describing the angles and distances between the nodes. Common patterns in these graphs indicate common structural motifs. This approach may be extended to include any descriptor of atoms or groups of atoms, such as hydrogen bond donor or acceptor, hydrophobicity, shape, charge or aromaticity; for example proteins may be spatially mapped according to such descriptors using GRID and this representation used as a basis for similarity searching (Baroni et al (2007) J Chem Inf Mod 47, 279-294). Descriptions of the methods, availability of software, and guidelines for user-defined selection of parameters, thresholds and tolerances are described in the references given above.

In a preferred embodiment, the structural membrane protein model is a structural GPCR model. It will be appreciated that the structural model of a GPCR may be a model of the first parent GPCR. For example, stabilising mutations within a mutant GPCR having increased stability can be directly mapped onto the first parent GPCR structure and the structural motifs in which such mutations are located, identified.

Where the structure of the first parent GPCR is unknown, structural models of other GPCRs may be used. For example, stabilising mutations in a GPCR from one species may be mapped onto a known structural model of the same GPCR from another species. Similarly, stabilising mutations in one particular GPCR isoform may be mapped onto a known structural model of another GPCR isoform. Moreover, stabilising mutations from one GPCR may be mapped onto a GPCR of the same class or family. A list of GPCR classes and families has been produced by the International Union of Pharmacology (Foord et al (2005) Pharmacol. Rev. 57, 279-288) and this list is periodically updated at www.iuphar-db.org/GPCR/ReceptorFamiliesForward.

As described above, it will be appreciated that the structural model may be of any GPCR provided it has sufficient structural similarity across the domain in which the mutant GPCR has at least one different amino acid compared to the first parent GPCR. Thus, it is preferred if the GPCR shares at least 20% sequence identity with the mutant of the first parent GPCR across the protein domain containing the at least one different amino acid residue compared to the first parent GPCR, and more preferably at least 30%, 40%, 50%, 60%, 70%, 80% or 90% sequence identity, and yet more preferably at least 95% or 99% sequence identity. However, the inventors recognise that the 20% sequence identity threshold is not absolute. GPCRs with less than 20% sequence identity to the first parent GPCR may also serve as a structural model to which stabilising mutations are transferred, wherein the low sequence identity is counterbalanced by other similarities, including, for example, the presence of the same sequence motifs, binding to the same G-protein or having the same function, or having substantially the same hydropathy plots compared to the first parent GPCR.

Mapping of stabilising mutations onto the structural model can be done using any suitable method known in the art. For example, typically, the amino acid sequence of the GPCR for which the structural model is available is aligned with the amino acid sequence of the mutant of the first parent GPCR. The position or positions of the at least one different amino acid residue in the mutant GPCR relative to the first parent GPCR can then be located in the amino acid sequence of the GPCR for which a structural model is available.

By ‘structural motif’ we include the meaning of a three dimensional description of the location in a GPCR structural model of a thermostabilising mutation. For example, the structural motif may be any secondary or tertiary structural motif within the GPCR. By ‘tertiary structural motif’ we include any descriptor of atoms or groups of atoms, such as hydrogen bond donor or acceptor, hydrophobicity, shape, charge or aromaticity. For example, proteins may be spatially mapped according to such descriptors using GRID and this representation used as a basis for defining a structural motif (Baroni et al (2007) J Chem Inf Mod 47, 279-294).

Table (vi) lists the structural motifs in which the turkey β1 adrenergic receptor stabilising mutations were found to reside. As seen from the table, the mutations are positioned in a number of distinct localities. Three mutations are in loop regions that are predicted to be accessible to aqueous solvent. Eight mutations are in the transmembrane α-helices and point into the lipid bilayer; three of these mutations are near the end of the helices and may be considered to be at the hydrophobic-hydrophilic boundary layer. Eight mutations are found at the interfaces between transmembrane α-helices, three of which are either within a kinked or distorted region of the helix and another two mutations occur in one helix but are adjacent to one or more other helices which contain a kink adjacent in space to the mutated residue. These latter mutations could affect the packing of the amino acids within the kinked region, which could result in thermostabilisation. Another mutation is in a substrate binding pocket.

Accordingly, in one embodiment, the structural motif is any of a helical interface, a helix kink, a helix opposite a helix kink, a helix surface pointing into the lipid bilayer, a helix surface pointing into the lipid bilayer at the hydrophobic-hydrophilic boundary layer, a loop region or a protein binding pocket.

Identifying a structural motif in which a stabilising mutation resides suggests the importance of that motif in protein stability. Therefore, making one or more mutations in the amino acid sequence that defines a corresponding structural motif or motifs in a second parent GPCR, should provide one or more mutants of a second parent GPCR with increased stability relative to the second parent GPCR.

The amino acid sequence which defines a structural motif is the primary amino acid sequence of the amino acid residues which combine in the secondary or tertiary structure of the protein to form the structural motif. It will be appreciated that such a primary amino acid sequence may comprise contiguous or non-contiguous amino acid residues. Thus, identifying the amino acid sequence which defines the structural motif will involve determining the residues involved and subsequently defining the sequence. Mutations can be made in an amino acid sequence, for example as described above and using techniques well-established in the art.

By “corresponding structural motif or motifs”, we mean the analogous structural motif or motifs identified in the structural model which are present in the second parent GPCR. For example, if a helical interface was identified, the corresponding helical interface in the second parent GPCR would be the interface between the helices which are analogous to the helices present in the structural model. If a helical kink was identified, the corresponding helical kink would be the kink in the helix which is analogous to the kinked helix present in the structural model. An analogous structural motif or motifs in the second parent GPCR can be identified by searching for similar amino acid sequences in the sequence of the second parent GPCR which define the motif or motifs in the structural model, for example, by sequence alignment. Moreover, computer based algorithms are widely available in the art that can be used to predict the presence of protein motifs based on an amino acid sequence. Thus, based upon the relative position of a particular motif within the amino acid sequence and its position relative to other motifs, an analogous structural motif can readily be identified. It will be appreciated that if a structural model of the second parent GPCR is available, the analogous structural motif or motifs can be directly mapped onto the structure of the protein. Typically, the amino acid sequence defining the analogous structural motif has at least 20% sequence identity with the sequence defining the motif in the structural model, more preferably at least 30%, 40%, 50%, 60%, 70%, 80% and 90% sequence identity and yet more preferably 95% and 99% sequence identity.

In one embodiment, the second parent GPCR is the first parent GPCR. For the avoidance of doubt, the second parent GPCR may have the naturally-occurring sequence of the first parent GPCR, or it may be a truncated form or it may be a fusion, either to the naturally occurring protein or to a fragment thereof, or it may contain mutations compared to the naturally-occurring sequence, providing that it retains ligand-binding.

In an alternative embodiment, the second parent GPCR is not the first parent GPCR. For example, a mutant of a first parent GPCR may have been identified that has increased stability but it is desired to generate a mutant of a different GPCR with increased stability. Preferably, the second parent GPCR is of the same GPCR class or family as the first parent GPCR as described above. However, it will be appreciated that the second parent GPCR may be any known GPCR provided that it shares sufficient structural similarity with the first parent GPCR, such that it contains a corresponding structural motif in which the stabilising mutation of the mutant of the first parent GPCR resides. Thus typically, the second parent GPCR has at least 20% sequence identity to the first parent GPCR and more preferably at least 30%, 40%, 50%, 60%, 70%, 80% or 90% sequence identity. However, as mentioned above, some GPCRs have low sequence identity (e.g. family B and C GPCRs) but are similar in structure. Thus the 20% sequence identity threshold is not absolute.

Since there are potentially thousands of mutations that can be screened in a GPCR for increased stability, it is advantageous to target particular mutations which are known to be important in conferring stability. Therefore, it will be appreciated that the methods of the eighth and ninth aspects of the invention may be used in a method of selecting mutant GPCRs with increased stability. In particular, carrying out the methods of the eighth or ninth aspects of the invention can be used to target mutations to particular amino acid residues or to amino acid sequences which define structural motifs important in determining stability.

Accordingly, in one embodiment the methods of the eighth or ninth aspects further comprise:

• • (I) selecting a ligand, the ligand being one which binds to the second parent GPCR when the GPCR is residing in a particular conformation
  • (II) determining whether the or each mutant of the second parent GPCR when residing in a particular conformation has increased stability with respect to binding the selected ligand compared to the stability of the second parent GPCR when residing in the same particular conformation with respect to binding that ligand, and
  • (III) selecting those mutants that have an increased stability compared to the second parent GPCR with respect to binding the selected ligand.

It will be noted that steps (I), (II) and (III) correspond to steps (b), (c) and (d) of the method of the first aspect of the invention described above. Accordingly, preferences for the ligand and methods of assessing stability are as defined above with respect to the method of the first aspect of the invention.

A tenth aspect of the invention provides a mutant GPCR with increased stability relative to its parent GPCR produced by the method of the tenth aspect of the invention.

In one embodiment, the mutant GPCR of the tenth aspect of the invention is a mutant GPCR which has, compared to its parent receptor, at least one different amino acid at a position which corresponds to any one or more of the following positions: (i) according to the numbering of the turkey β-adrenergic receptor as set out in FIG. 9: Ile 55, Gly 67, Arg 68, Val 89, Met 90, Gly 67, Ala 184, Arg 199, Ala 203, Leu 208, Gln 210, Ser 213, Glu 219, Arg 220, Ser 223, Thr 224, Gln 226, Lys 227, His 230, Leu 241, Pro 260, Ser 263, Leu 267, Leu 272, Thr 279, Asn 284, Gln 311, Pro 313, Lys 315, (iii) according to the numbering of the rat neurotensin receptor as set out in FIG. 11: Ala 69, Leu 72, Ala 73, Ala 86, Ala 90, Ser 100, His 103, Ser 108, Leu 109, Leu 111, Asp 113, Ile 116, Ala 120, Asp 139, Phe 147, Ala 155, Val 165, Glu 166, Lys 176, Ala 177, Thr 179, Met 181, Ser 182, Arg 183, Phe 189, Leu 205, Thr 207, Gly 209, Gly 215, Val 229, Met 250, Ile 253, Leu 256, Ile 260, Asn 262, Val 268, Asn 270, Thr 279, Met 293, Thr 294, Gly 306, Leu 308, Val 309, Leu 310, Val 313, Phe 342, Asp 345, Tyr 349, Tyr 351, Ala 356, Phe 358, Val 360, Ser 362, Asn 370, Ser 373, Phe 380, Ala 385, Cys 386, Pro 389, Gly 390, Trp 391, Arg 392, His 393, Arg 395, Lys 397, Pro 399, and (iv) according to the numbering of the muscarinic receptor as set out in FIG. 17: Leu 65, Met 145, Leu 399, Ile 383 and Met 384.

Alignment of the turkey β1 AR, human adenosine receptor, rat neurotensin receptor and human muscarinic receptor amino acid sequences in FIG. 17, shows that in 11 instances out of 70, two sequences contain mutations art the same position, namely at the following positions according to the numbering of the human beta2 AR as set out in FIG. 17: Ala 59, Val 81, Ser 143, Lys 147, Val 152, Glu 180, Val 222, Ala 226, Ala 271, Leu 275 and Val 317. Therefore, in a preferred embodiment, the mutant GPCR of the tenth aspect of the invention is one which has, compared to its parent receptor, a different amino acid at any one of these positions.

In one embodiment the mutant GPCR of the tenth aspect of the invention is a mutant β-adrenergic receptor. For example, the mutant β-adrenergic receptor may have at least one different amino acid residue in a structural motif in which the mutant receptor compared to its parent receptor has a different amino acid at a position which corresponds to any of the following positions according to the numbering of the turkey β-adrenergic receptor as set out in FIG. 9: Ile 55, Gly 67, Arg 68, Val 89, Met 90, Gly 98, Ile 129, Ser 151, Val 160, Gln 194, Gly 197, Leu 221, Tyr 227, Arg 229, Val 230, Ala 234, Ala 282, Asp 322, Phe 327, Ala 334, Phe 338.

In one embodiment the mutant GPCR of the tenth aspect of the invention is a mutant adenosine receptor. For example, the mutant adenosine receptor may have at least one different amino acid residue in a structural motif in which the mutant receptor compared to its parent receptor has a different amino acid at a position which corresponds to any of the following positions according to the numbering of the human adenosine _Ata receptor as set out in FIG. 10: Gly 114, Gly 118, Leu 167, Ala 184, Arg 199, Ala 203, Leu 208, Gln 210, Ser 213, Glu 219, Arg 220, Ser 223, Thr 224, Gln 226, Lys 227, His 230, Leu 241, Pro 260, Ser 263, Leu 267, Leu 272, Thr 279, Asn 284, Gln 311, Pro 313, Lys 315.

In one embodiment the mutant GPCR of the tenth aspect of the invention is a mutant neurotensin receptor. For example, the mutant neurotensin receptor may have at least one different amino acid residue in a structural motif in which the mutant receptor compared to its parent receptor has a different amino acid at a position which corresponds to any of the following positions according to the numbering of the rat neurotensin receptor as set out in FIG. 11: Ala 69, Leu 72, Ala 73, Ala 86, Ala 90, Ser 100, His 103, Ser 108, Leu 109, Leu 111, Asp 113, Ile 116, Ala 120, Asp 139, Phe 147, Ala 155, Val 165, Glu 166, Lys 176, Ala 177, Thr 179, Met 181, Ser 182, Arg 183, Phe 189, Leu 205, Thr 207, Gly 209, Gly 215, Val 229, Met 250, Ile 253, Leu 256, Ile 260, Asn 262, Val 268, Asn 270, Thr 279, Met 293, Thr 294, Gly 306, Leu 308, Val 309, Leu 310, Val 313, Phe 342, Asp 345, Tyr 349, Tyr 351, Ala 356, Phe 358, Val 360, Ser 362, Asn 370, Ser 373, Phe 380, Ala 385, Cys 386, Pro 389, Gly 390, Trp 391, Arg 392, His 393, Arg 395, Lys 397, Pro 399.

In one embodiment the mutant GPCR of the tenth aspect of the invention is a mutant muscarinic receptor. For example, the mutant muscarinic receptor may have at least one different amino acid residue in a structural motif in which the mutant receptor compared to its parent receptor has a different amino acid at a position which corresponds to any of the following positions according to the numbering of the human muscarinic receptor as set out in FIG. 17: Leu 65, Met 145, Leu 399, Ile 383 and Met 384.

It is preferred that the mutant GPCRs of the invention have increased stability to any one of heat, a detergent, a chaotropic agent and an extreme of pH.

It is preferred if the mutant GPCRs of the invention have increased thermostability.

It is preferred that the mutant GPCRs of the invention, including the mutant β-adrenergic, adenosine and neurotensin receptors, have an increased thermostability compared to its parent when in the presence or absence of a ligand thereto. Typically, the ligand is an antagonist, a full agonist, a partial agonist or an inverse agonist, whether orthosteric or allosteric. As discussed above, the ligand may be a polypeptide, such as an antibody.

It is preferred that the mutant GPCRs of the invention, for example a mutant β-adrenergic receptor or a mutant adenosine receptor or a mutant neurotensin receptor is at least 2° C. more stable than its parent preferably at least 5° C. more stable, more preferably at least 8° C. more stable and even more preferably at least 10° C. or 15° C. or 20° C. more stable than its parent. Typically, thermostability of the parent and mutant receptors are measured under the same conditions. Typically, thermostability is assayed under a condition in which the GPCR resides in a particular conformation. Typically, this selected condition is the presence of a ligand which binds the GPCR.

It is preferred that the mutant GPCRs of the invention, when solubilised and purified in a suitable detergent has a similar thermostability to bovine rhodopsin purified in dodecyl maltoside. It is particularly preferred that the mutant GPCR retains at least 50% of its ligand binding activity after heating at 40° C. for 30 minutes. It is further preferred that the mutant GPCR retains at least 50% of its ligand binding activity after heating at 55° C. for 30 minutes.

The mutant GPCRs disclosed herein are useful for crystallisation studies and are useful in drug discovery programmes. They may be used in biophysical measurements of receptor/ligand kinetic and thermodynamic parameters eg by surface plasmon resonance or fluorescence based techniques. They may be used in ligand binding screens, and may be coupled to solid surfaces for use in high throughput screens or as biosensor chips. Biosensor chips containing the mutant GPCRs may be used to detect molecules, especially biomolecules.

The invention also includes a polynucleotide which encodes a mutant GPCR of the invention. In particular, polynucleotides are included which encode the mutant β-adrenergic receptor or the mutant adenosine receptors or the mutant neurotensin receptors of the invention. The polynucleotide may be DNA or it may be RNA. Typically, it is comprised in a vector, such as a vector which can be used to express the said mutant GPCR. Suitable vectors are ones which propagate in and/or allow the expression in bacterial or mammalian or insect cells.

The invention also includes host cells, such as bacterial or eukaryotic cells, which contain a polynucleotide which encodes the mutant GPCR. Suitable cells include E. coli cells, yeast cells, mammalian cells and insect cells.

The invention will now be described in more detail with respect to the following Figures and Examples wherein:

FIG. 1 Amino acid changes in βAR that lead to thermostability. Stability quotient indicates the % remaining binding activity of the mutants after heating the sample for 30 min at 32° C. All values are normalized to βAR_34-424 (50%, showed 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. The original amino acid and its position in βAR_34-424 is indicated below. Bars corresponding to the same amino acid in βAR_34-424 are in the same colour with arrows indicating the best mutations. Errors were calculated from duplicate measurements; the best mutants were subsequently re-assayed to determine the Tm for each individual mutation and to give an accurate rank order of stability for each mutant (see Example 1).

FIG. 2 Side chains in rhodopsin that are at equivalent positions to the thermostable mutations in βAR_34-424. The equivalent amino acid residues in rhodopsin to the amino acid residues mutated in βAR_34-424 were located in the rhodopsin structure, based upon an alignment among rhodopsin, β1 adrenergic receptor, neurotensin receptor, and adenosine A_2a receptor (data not shown). Side chains in the same transmembrane helix are shown as space filling models in the same colour. The name and position of the amino acid residues are those in rhodopsin.

FIG. 3 Evolution of thermostability in βAR. Starting from βR-m10-8, combinations of mutations were rearranged systematically to find the optimum combination of mutations (see also Table 2).

FIG. 4 Stability of βAR-m23 and βAR_34-424 in the apo-state or containing the bound antagonist [³H]-DHA. To determine Tm in the absence of ligand (apo-state, discontinuous lines), detergent-solubilised receptors were incubated for 30 minutes at the temperatures indicated before carrying out the binding assay. For the Tm determination of the antagonist-bound form (continuous lines), detergent-solubilised receptors were pre-incubated with [³H]-DHA, followed by incubation at the temperatures indicated. βAR-m23 (circles), and βAR_34-424 (squares). Data points are from duplicates measurements in a representative experiment.

FIGS. 5a-c Competition binding of agonists to βAR-m23 and βAR_34-424. Binding assays were performed on receptors partially purified in DDM; βAR-m23 (triangles) and βAR_34-424 (squares). [³H]-DHA was used at a concentration three times greater than the K_D of partially purified receptor (see Methods). [³H]-DHA binding was competed with increasing concentrations of the agonists, norepinephrine (FIG. 5a) and isoprenaline (FIG. 5b), or with an antagonist, alprenolol (FIG. 5c). LogEC₅₀ and corresponding EC₅₀ values for the different ligands were calculated by nonlinear regression using GraphPad Prism software and the error for logEC₅₀s were lower than 10%. The EC₅₀s for ligand binding to βAR_34-424 and βAR-m23 are: norepinephrine, βAR_34-424 1.5 μM, βAR-m23 3.7 mM; isoprenaline, βAR_34-424 330 nM, βAR-m23 20 μM; alprenolol, βAR 78 nM, βAR-m23 112 nM.

FIGS. 6a-c Stability of βR-m23 and βAR_34-424 in five different detergents. Samples of βAR_34-424 (FIG. 6a), and βAR-m23 (FIG. 6b) solubilized in DDM were partially purified on Ni-NTA agarose columns allowing the exchange into various different detergents: DDM (squares), DM (triangles), OG (inverted triangles), LDAO (diamonds) and NG (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 the Methods and the Tm is shown at the intersection between the curves and the discontinuous line. Results are from duplicate measurements in a representative experiment performed in parallel. (FIG. 6c) Photomicrograph of a crystal of βAR-m23 mutant, which showed good order by X-ray diffraction.

FIG. 7 Curve of thermostability of βAR_34-424 (Tm). Binding assays were performed using [³H]-dihydroalprenolol (DHA) as radioligand as described under “Methods”. Samples were heated for 30 minutes at different temperatures before the assay. Tm represents the temperature at which the binding decreased to the 50%, value showed as a discontinuous line. Data points are from duplicates of one single experiment. This experiment has been repeated several times with similar results.

FIGS. 8a-b Saturation binding assays of membranes of βAR_34-424 and βAR-m23. Binding assays were performed as described in “Methods” using [³H]-dihydroalprenolol (DHA) as radioligand; βAR_34-424 (FIG. 8a) and βAR-m23 (FIG. 8b). Scatchard plots are shown as insets along with the corresponding values for B_max and K_D. Data points are from duplicates of two independent experiments for each protein. Data were analyzed by nonlinear regression using Prism software (GraphPad).

FIGS. 9A-B Alignment of the turkey β-adrenergic receptor with human β31, β2 and β3 receptors.

FIGS. 10A-B Alignment of human adenosine receptors.

FIGS. 11A-B Alignment of neurotensin receptors.

FIG. 12 Flow chart showing the two different assay formats of ligand (+) and ligand (−) used to determine receptor thermostablity.

FIGS. 13A-F Pharmacological profile of thermostable mutant adenosine A2a receptor, Rant21. Saturation binding of (FIG. 13A) antagonist and (FIG. 13B) agonist to solubilised receptors. (FIGS. 13C-F) Inhibition of [³H]ZM241385 binding by increasing concentrations of antagonists (FIG. 13C) XAC and (FIG. 13D) Theophylline, and agonists (FIG. 13E) NECA and (FIG. 13F) R-PIA; binding of [³H]ZM241385 (10 nM) in the absence of unlabeled ligand was set to 100%. Each solubilised receptor was incubated with ligands for one hour on ice in binding buffer (50 mM Tris pH 7.5 and 0.025% DDM) containing 400 mM NaCl (FIG. 13A, FIG. 13C-F). Data shown are from two independent experiments with each data point measured in triplicate. K_D and K_i values are given in Table (iii).

FIGS. 14A-B Thermostable mutants show a decreased dependence on lipids (FIG. 14A) and an increased survival at higher concentration of DDM (FIG. 14B) upon heating compared to the wild-type receptor. Receptors were solubilised in 1% DDM (diluted in 50 mM Tris pH 7.5 and 400 mM NaCl) and immobilised on Ni-NTA agarose for the IMAC step. Exchange of buffer containing the appropriate concentration of DDM and/or lipids was performed during washes and elution from the Ni-NTA beads.

FIG. 15 Mapping of the M90V, Y227A, A282L and F338M m23 mutations in turkey beta1 adrenergic receptor onto homologous residues (M82, Y219, C265 and A321 respectively) in the human beta2 adrenergic receptor structure (Rasmussen et al (2007) Nature 15; 383-387; pdb accession codes 2R4R and 2R4S) reveals their position at a helical interface and helical kink respectively. Amino acid residues in equivalent positions to the thermostabilising mutations in the turkey β1 adrenergic receptor are shown as labelled space filling models.

FIG. 16 Mapping of m23 mutations in turkey beta1 adrenergic receptor onto homologous residues in the human beta2 adrenergic receptor structure (Cherezov et al (2007) Science, 318:1258-65; pdb accession code 2RH1). The Cα trace of the β2AR is shown with the fusion moiety (T4 lysozyme) removed. The six mutations in βAR-m23 (R68S, M90V, Y227A, A282L, F327A, F338M) are equivalent to amino acid residues K60, M82, Y219, C265, L310, F321 in the human β2AR. Lys60 is on the intracellular end of Helix 1 and points into the lipid-water interface. Met82 is near the middle of Helix 2 and points into the ligand binding pocket; the nearest distance between the substrate carazolol and the Met side chain is 5.7 Å. Tyr219 is towards the intracellular end of helix 5 and is at the helix5-helix 6 interface. Cys265 is at the end of the loop region between helices 5 and 6 and points away from the transmembrane regions. Leu310 and Phe321 are both in helix 7 and both point out into the lipid bilayer.

FIGS. 17A-C Multiple sequence alignment of human beta-2AR, rat NTR1, turkey beta-1 AR, human Adenosine A2aR and human muscarinic M1 receptors. In each sequence, thermostabilising mutations are marked with a box. Mutations occurring in two or more sequences are denoted with a star.

FIG. 18 Mapping of turkey beta1AR mutation I55A (human beta2AR 147) onto human beta2AR structure (pdb accession code 2RH1). Mutation is at the interface between 3 helices (H1, H2 kink, H7 kink) Left: side view; right: top view.

FIG. 19 Mapping of turkey beta1AR V89L mutation (human beta2AR V81) onto human beta2AR structure (pdb accession code 2RH1). Mutation is in the kink in helix 2. The helices are numbered and the bound antagonist is shown as a space filling model. Amino acid residues in equivalent positions to the thermostabilising mutations in the turkey β1 adrenergic receptor are shown as space filling models and are arrowed for clarity. Left: side view; right: top view.

FIG. 20 Mapping of turkey beta1AR M90V mutation (human beta2AR M82) onto human beta2AR structure (pdb accession code 2RH1). Mutation is in kink in helix 2 oriented towards the binding pocket. The helices are numbered and the bound antagonist is shown as a space filling model. Amino acid residues in equivalent positions to the thermostabilising mutations in the turkey β1 adrenergic receptor are shown as space filling models and are arrowed for clarity. Left: side view; right: top view.

FIG. 21 Mapping of turkey beta1AR I129V mutation (human beta2AR I121) onto human beta2AR structure (pdb accession code 2RH1). Mutation is opposite a kink in helix 5. The helices are numbered and the bound antagonist is shown as a space filling model. Amino acid residues in equivalent positions to the thermostabilising mutations in the turkey β1 adrenergic receptor are shown as space filling models and are arrowed for clarity. Left: side view; right: bottom view.

FIG. 22 Mapping of turkey beta1AR F338M mutation (human beta2AR F321) onto human beta2AR structure (pdb accession code 2RH1). Mutation is in kink in helix 7. The helices are numbered and the bound antagonist is shown as a space filling model. Amino acid residues in equivalent positions to the thermostabilising mutations in the turkey β1 adrenergic receptor are shown as space filling models and are arrowed for clarity. Left: side view; right: top view.

FIG. 23 Mapping of turkey beta1AR Y227A mutation (human beta2AR Y219) onto human beta2AR structure (pdb accession code 2RH1). Mutation is at helix-helix interface. The helices are numbered and the bound antagonist is shown as a space filling model. Amino acid residues in equivalent positions to the thermostabilising mutations in the turkey β1 adrenergic receptor are shown as space filling models and are arrowed for clarity. Left: side view; right: bottom view.

FIG. 24 Mapping of turkey beta1AR A282L mutation (human beta2AR C265) onto human beta2AR structure (pdb accession code 2RH1). Mutation is in loop region. The helices are numbered and the bound antagonist is shown as a space filling model. Amino acid residues in equivalent positions to the thermostabilising mutations in the turkey β1 adrenergic receptor are shown as space filling models and are arrowed for clarity. Left: side view; right: top view.

FIG. 25 Mapping of turkey beta1AR R68S mutation (human beta2AR K60) onto human beta2AR structure (pdb accession code 2RH1). Mutation is at the lipid-water boundary, pointing into the solvent. The helices are numbered and the bound antagonist is shown as a space filling model. Amino acid residues in equivalent positions to the thermostabilising mutations in the turkey β1 adrenergic receptor are shown as space filling models and are arrowed for clarity. Left: side view; right: angled top view.

FIG. 26 Comparison of the thermostabilities of three β adrenergic receptors (turkey β1 (▪), human β1 (▾) and human β2 (●)) and two thermostabilised receptors (turkey β1-m23 (▴) and human β2-m23 (♦)). The six thermostabilising mutations in β1-m23 (R68S, M90V, Y227A, A282L, F327A, F338M) were all transferred directly to the human (32 receptor (K60S, M82V, Y219A, C265L, L310A, F321M) making β2-m23, based upon the alignment in FIGS. 9A-B. The resulting mutants were transiently expressed in mammalian cells, solubilised in 0.1% dodecylmaltoside and assayed for thermostability in the minus-ligand format (heating the apo-state, quenching on ice, adding 3H-DHA). The apparent Tms for turkey β1 and β2-m23 were 23° C. and 45° C. respectively, giving a ΔTm of 22° C. as seen previously in E. coli expressed receptor. The Tms for human β2 and β2-m23 were 29° C. and 41° C. respectively, showing that the apo receptor was stabilised by 12° C. This exemplifies the principle of the transferability of thermostabilising mutations from one receptor to another receptor, which in this case are 59% identical. The human β1 receptor (Tm˜12° C.) is much less stable than the turkey β1 receptor.

FIG. 27 Percentage identity of the turkey β1 adrenergic receptor, human adenosine receptor and rat neurotensin receptor to human β adrenergic receptors, human adenosine receptors and human neurotensin receptors, respectively.

FIGS. 28A-B Alignment of neurotensin receptors.

EXAMPLE 1: CONFORMATIONAL STABILISATION OF THE β3-ADRENERGIC RECEPTOR IN DETERGENT-RESISTANT FORM

Summary

There are over 500 non-odorant G protein-coupled receptors (GPCRs) encoded by the human genome, many of which are predicted to be potential therapeutic targets, but there is only one structure available, that of bovine rhodopsin, to represent the whole of the family. There are many reasons for the lack of progress in GPCR structure determination, but we hypothesise that improving the detergent-stability of these receptors and simultaneously locking them into one preferred conformation will greatly improve the chances of crystallisation. A generic strategy for the isolation of detergent-solubilised thermostable mutants of a GPCR, the β-adrenergic receptor, was developed based upon alanine scanning mutagenesis followed by an assay for receptor stability. Out of 318 mutants tested, 15 showed a measurable increase in stability. After optimisation of the amino acid residue at the site of each initial mutation, an optimally stable receptor was constructed by combining specific mutations. The most stable mutant receptor, βAR-m23, contained 6 point mutations that led to a Tm 21° C. higher than the native protein and, in the presence of bound antagonist, βARm23 was as stable as bovine rhodopsin. In addition, βAR-m23 was significantly more stable in a wide range of detergents ideal for crystallisation and was preferentially in an antagonist conformation in the absence of ligand.

Results

Selection of Single Mutations That Increase the Thermostability of the β1 Adrenergic Receptor

βAR from turkey erythrocytes is an ideal target for structural studies because it is well characterised and is expressed at high-levels in insect cells using the baculovirus expression system[10,11]. The best overexpression of βAR is obtained using a truncated version of the receptor containing residues 34-424 (βAR_34-424) [9] and this was used as the starting point for this work. Alanine scanning mutagenesis was used to define amino residues in βAR_34-424 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 residues were not obtained due to 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-solubilised receptors by heating the receptors at 32° C. for 30 minutes, quenching the reaction on ice and then performing a radioligand binding assay, using the antagonist [³H]-dihydroalprenolol, to determine the number of remaining functional βAR_34-424 molecules compared to the unheated control. Heating the unmutated βAR_34-424 at 32° C. for 30 min before the assay reduced binding to approximately 50% of the unheated control (FIG. 7); all the data for the mutants were normalised by including the unmutated βAR_34-424 as a control in every assay performed. In the first round of screening, eighteen mutants showed an apparent increase in stability, maintaining more than 75% of antagonist binding after heating and being expressed in E. coli to at least 50% of the native βAR_34-424 levels. In view of the possibility of increasing further the stability of these mutants, each of the 18 residues was mutated to 2-5 alternative amino acid residues of varying size or charge (FIG. 1). Out of these 18 mutants, 12 were not improved by further changes, 5 had better thermostability if another amino acid was present and one mutation from the first screen turned out to be a false positive. In addition, three residues that were not stabilised upon mutation to alanine (V89, S151, 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.

The position and environment predicted for each of the 16 amino residues that gave the best increases in thermostability when mutated were determined by aligning the βAR sequence with that of rhodopsin whose structure is known (FIG. 2). Fourteen of these residues were predicted to be present in transmembrane α-helices, with five of the residues predicted to be lipid-facing, 4 being deeply buried and 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 βAR_34-424 was determined by measuring the 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 minutes. Each mutation increased the Tm of βAR_34-424 by 1-3° C., with the exception of M90A and Y227A that 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. Constructions containing the mutations G67A and R68S, or different combinations of the mutations at the end of helix 5 (Y227A, R229Q, V230A and A234L) were expressed and assayed; the Tm values (results not shown) were only 1-3° C. higher than the Tm for βAR_34-424 and one mutant was actually slightly less stable, suggesting that combining mutations that are 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. PCR reactions were performed using various mixes of primers to combine up to 5 different mutations in a random manner and then tested for thermostability (Table 1). The best of these combinations increased the Tm more than 10° C. compared to the Tm of βAR_34-424. In some cases, there was a clear additive effect upon the Tm with the sequential incorporation of individual mutations. This is seen in a series of 3 mutants, m4-1, m4-7 and m4-2, with the addition of V230A to m4-1 increasing the Tm by 2° C. and the additional mutation D332A in m4-7 increasing the Tm a further 3° C. Mutants that contained Y227A and M90A all showed an increase in Tm of 10° C. or more. Just these two mutations together increased the Tm of βAR_34-424 by 13° C. (m7-5), however, the total antagonist binding was less than 50% of βAR_34-424 suggesting impaired expression of this mutant. The addition of F338M to m7-5 did not increase the thermostability, but it increased levels of functional expression in E. coli.

TABLE 1
Combinations of mutations by PCR. 10 PCR reactions were
performed combining different pairs of primers that contained
the selected mutations. Successful PCR reactions are shown
in the table. The stability of these new mutants was assayed
as described in FIG. 7 and the Tm calculated. The results
are shown as the mean ± S.E. from duplicates.
PCR	Receptor	Mutations	Tm (° C.)
	βAR34-424		31.7 ± 0.1
4	m4-1	G67A, G98A	35.5 ± 0.9
	m4-2	G67A, G98A, V230A, D322A	40.9 ± 0.9
	m4-6	G98A, D322A	35.0 ± 0.2
	m4-7	G67A, G98A, V230A	38.0 ± 1.2
6	m6-1	Y227A, A234L, A282L, A334L	41.6 ± 0.9
	m6-4	R68S, Y227A, A234L, A282L	41.6 ± 0.1
	m6-5	R68S, A234L, A282L, A334L	41.9 ± 0.5
	m6-9	R68S, Y227A, A234L, A282L, A334L	43.7 ± 0.4
	m6-10	R68S, Y227A, A282L, A334L	47.4 ± 1.1
	m6-11	R68S, A282L, A334L	39.1 ± 0.5
7	m7-1	M90V, A282L, F338M	43.0 ± 0.8
	m7-2	M90V, A282L	38.9 ± 0.6
	m7-5	M90V, Y227A	45.2 ± 1.0
	m7-6	M90V, I129V	40.0 ± 0.6
	m7-7	M90V, Y227A, F338M	45.2 ± 2.0
10	m10-4	R68S, M90V, V230A, A334L	46.9 ± 1.0
	m10-8	R68S, M90V, V230A, F327A, A334L	47.3 ± 1.4

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 8 mutations occurring in at least two of the mutants. A second round of mutagenesis was performed using m10-8 as the template and adding or replacing mutations present in m6-10 and m7-7 (FIG. 3); some of these mutations were very close in the primary amino acid sequence of βAR and therefore were not additive as noted above, but many mutations improved the Tm further (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 βAR_34-424 mutant so far and will be referred to as βAR-m23.

TABLE 2
Improvement of best combination of mutations. These new mutants were
obtained mixing the changes present in mutants m6-10, m7-7 and m10-8 by PCR.
The stability of these new mutants was assayed as described in FIG. 7 and the Tm
calculated. The results are shown as the mean ± S.E. from duplicates.
	Mutations	Tm (° C.)
m17	R68S	M90V	Y227A	V230A	—	F327A	A334L	—	48.2 ± 1.4
m18	R68S	M90V	Y227A	V230A	A282L	F327A	—	F338M	49.6 ± 0/9
m19	R68S	M90V	Y227A	—	A282L	F327A	—	F338M	49.0 ± 0.8
m20	R68S	M90V	—	—	—	F327A	A334L	—	48.4 ± 0.7
m21	R68S	M90V	Y227A	—	—	F327A	A334L	—	47.0 ± 1.3
m22	R68S	M90V	Y227A			F327A	A334L	—	47.4 ± 0.5
m23	R68S	M90V	Y227A	—	A282L	F327A	—	F338M	52.8 ± 1.4

The thermostability assays used to develop βAR_34-424 mutants were performed by heating the receptor in the absence of the antagonist, but it is well known that bound ligand stabilises receptors. Therefore, stability assays for βAR_34-424 and βAR-m23 were repeated with antagonist bound to the receptors during the heating step (FIG. 4). As expected, the Tm of the receptor that contained bound antagonist during the incubation was higher than that for the receptor without antagonist. For βAR_34-424 the Tm was 6° C. higher with bound antagonist and for βAR-m23 the Tm increased 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 βAR_34-424 (see also below). The Tm of βAR-m23 with antagonist bound is very similar to the Tm of dark-state rhodopsin in dodecylmaltoside (DDM)[12], whose structure has been solved by two independent laboratories[13,14]. This suggested that βAR-m23 is sufficiently stable for crystallisation.

Characterization of βAR-m23

The three characteristic activities measured for βAR-m23 and βAR_34-424 to identify the effect of the six mutations were the affinity of antagonist binding, the relative efficacies of agonist binding and the ability of βAR-m23 to couple to G proteins. Saturation binding experiments to membranes using the antagonist [³H]-dihydroalprenolol (FIG. 8) showed that the affinity of binding to βAR-m23 (K_D 6.5±0.2 nM, n=2) was slightly lower than for βAR_34-424 (K_D 2.8±0.1 nM, n=2), suggesting that there are no large perturbations in the structure of βARm23 in the antagonist-bound conformation. This is consistent with the observation that none of the mutations in βAR-m23 correspond with amino acids believed to be implicated in ligand binding. In contrast to antagonist binding, the efficacy of agonist binding by βAR-m23 is 3 orders of magnitude weaker than for βAR_34-424 (FIG. 5). The potency of the agonist isoprenaline is consistently lower in βAR-m23 and βAR_34-424 than for the native agonist norepinephrine, indicating that the agonist-bound conformation for the two receptors is likely to be similar. However, the large decrease in agonist efficacy in βAR-m23 compared to βAR_34-424 indicates that the 6 mutations in βAR-m23 have locked the receptor preferentially in an antagonist-bound conformation. From a crystallisation perspective, this is an added bonus to thermostabilisation, because it is essential to have a conformationally homogeneous protein population for the production of diffraction-quality crystals.

All of the thermostability assays used to derive βAR-m23 were performed on receptors solubilised in DDM. The aim of the thermostabilisation 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 crystallising integral membrane proteins. Membranes prepared from E. coli expressing βAR-m23 or βAR_34-424 were solubilised in DDM, bound to Ni-NTA agarose then washed with either DDM, decylmaltoside (DM), octylglucoside (OG), lauryldimethylamine oxide (LDAO) or nonylglucoside (NG). Stability assays were performed on the receptors in each of the different detergents (FIG. 6). βAR_34-424 was only stable 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 T_ms of 25° C. (NG), 23° C. (LDAO) and 17° C. (OG). The difference in Tm between βAR-m23 and βAR_34-424 is about 20° C., irrespective of whether the receptors were solubilised in either DDM or DM; it is therefore not surprising that no active βAR_34-424 could be found in even NG, because the predicted Tm would be about 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 upon 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 βAR_34-424 also resulted in increasing tolerance to small detergents ideal for crystallising integral membrane proteins.

Crystallisation of Mutant GPCR

Earlier attempts to crystallise several different constructs of turkey beta-adrenegic receptor failed. Despite experimenting with a variety of conditions, using both the native sequence and several truncated and loop-deleted constructs, over many years, no crystals were obtained.

However, once the stabilising mutations from βAR-m23 were transferred into the constructs, several different crystals were obtained in different detergents and different conditions.

The crystals that have been most studied so far were obtained using the purified beta-36 construct (amino acid residues 34-367 of the turkey beta receptor containing the following changes: point mutations C116L and C358A; the 6 thermostabilising point mutations in m23; replacement of amino acid residues 244-278 with the sequence ASKRK; a C terminal His6 tag) expressed in insect cells using the baculovirus expression system, after transferring the receptor into the detergent octyl-thioglucoside. The precipitant used was PEG600 or PEG1000 and the crystals obtained are elongated plates.

Experiments have also been carried out to see whether, once the crystallisation conditions had been defined using the stabilised receptor, it was possible to get crystals using the original non-stablised construct. It was possible that similar or perhaps very small crystals could have been obtained, but, in fact, the “wild type” (i.e. the starting structure from which the mutagenesis began) never gave any crystals.

The crystals are plate-shaped with space group C2 and diffract well, though the cell dimensions do vary depending on the freezing conditions used.

In general, once a GPCR has been stabilised it may be subjected to a variety of well-known techniques for structure determination. The most common technique for crystallising membrane proteins is by vapour diffusion (20, 21), usually using initially a few thousand crystallisation conditions set up using commercial robotic devices (22). However, sometimes the crystals formed by vapour diffusion are small and disordered, so additional techniques may then be employed. One technique involves the co-crystallisation (by vapour diffusion) of the membrane protein with antibodies that bind specifically to conformational epitopes on the proteins' surface (23, 24); this increases the hydrophilic surface of the protein and can form strong crystal contacts. A second alternative is to use a different crystallisation matrix that is commonly called either lipidic cubic phase or lipidic mesophase (25, 26), which has also been developed into a robotic platform (27). This has proven very successful for producing high quality crystals of proteins with only small hydrophilic surfaces e.g. bacteriorhodopsin (28). Membrane protein structures can also be determined to high-resolution by electron crystallography (29).

The evolution of βAR-m23 from βAR_34-424 by a combination of alanine scanning mutagenesis and the selection of thermostable mutants has resulted in a GPCR that is ideal for crystallography. The Tm for βAR-m23 is 21° C. higher than for βAR_34-424 and, in the presence of antagonist, βAR-m23 has a similar stability to rhodopsin. The increased Tm of βAR-m23 has resulted in an increased stability in a variety of small detergents that inactivate βAR_34-424. In addition, the selection strategy employed 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 βAR_34-424. Thus we have achieved a process of conformational stabilisation in a single selection procedure.

It is not at all clear why the particular mutations we have introduced lead to the thermostabilisation of the receptor. Equivalent positions in rhodopsin suggest that the amino acid residues mutated could be pointing into the lipid bilayer, into the centre of the receptor or at the interfaces between these two environments. Given the difficulties in trying to understand the complexities of the thermostabilisation of soluble proteins[15], it seems unlikely that membrane proteins will be any easier to comprehend; we found that there was no particular pattern in the amino acid residues in βAR that, when mutated, led to thermostability. However, since 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. While 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, an extreme of pH or in the presence of chaotropic salts. Conformational stabilisation 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 crystallised. It is to be hoped that conformational stabilisation will allow membrane protein crystallisation to become a far more tractable problem with a greater probability of rapid success than is currently the case. This should allow routine crystallisation of human membrane proteins in the pharmaceutical industry, resulting in valuable structural insights into drug development.

Methods

Materials. The truncated β1 adrenergic receptor from turkey (βAR_34-424)[9] was kindly provided by Dr Tony Warne (MRC Laboratory of Molecular Biology, Cambridge, UK). This βAR construct encoding residues 34-424 contains the mutation C116L to improve expression[11], 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 Ma1E fusion protein[16]. Mutants were generated by PCR using the expression plasmid as template using the QuikChange II methodology (Stratagene). PCR reactions 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 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 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), resuspended in 500 μl of buffer [20 mM Tris pH 8, 0.4 M NaCl, 1 mM EDTA and 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 minutes. Insoluble material was removed by centrifugation (15,000×g, 2 min, 4° C.) and the supernatant was used directly in radioligand binding assays.

For large-scale membrane preparations, 2L and 6L 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 lx protease inhibitor cocktail (Complete™ EDTA-free, Roche); 1 mg DNase I (Sigma) was added and the final volume was made to 100 ml. Cells were broken by a French press (2 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[17]. 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). 200 μl of Ni-NTA agarose was added to 2 ml of solubilized samples (10 mg/ml of 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 sec and washed twice with 250 μl of buffer (20 mM Tris pH 8, 0.4 M NaCl, 20 mM imidazole) containing detergent (either 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 K_D for [³H]-DHA binding to semipurified βAR_34-424 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 3 times the K_D ie 12 nM for βAR_34-424 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, 0.1% DDM (or corresponding detergent) with 50 nM [³H]-DHA and 20-100 μg 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 minutes; 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[18]. Non-specific binding was determined in the presence of 1 μM of s-propranolol. Saturation curves were obtained using a range of [³H]-DHA concentration from 0.4 nM to 100 nM. Competition assays were performed using a concentration of [³H]-DHA of 12 nM for βAR_34-424 and 40 nM for βAR-m23 (ie three times the K_D) 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 pdb file for the rhodopsin structure, accession code 1GZM[14], was downloaded from the Protein Data Bank website (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 upon an alignment among the four GPCRs with which we are most familiar, namely rhodopsin, β1 adrenergic receptor, neurotensin receptor and adenosine A_2a receptor[19].

EXAMPLE 2: MUTANTS OF THE ADENOSINE A RECEPTOR (A_2aR) WITH INCREASED THERMOSTABILITY

1. 315 site-directed mutants made between residues 2-316 of A_2aR.

2. All of these mutants have been assayed for thermostability using an assay measuring agonist and antagonist binding after the heating step (Ligand(−) format as described in FIG. 12).

• • a. 26 mutants showed improved thermostability when measured with ³H-NECA (agonist): G114 A, G118A, L167A, A184L, R199A, A203L, L208A, Q210A, S213A, E219A, R220A, S223A, T224A, Q226A, K227A, H230A, L241A, P260A, S263A, L267A, L272A, T279A, N284A, Q311A, P313A, K315A.
  • b. 18 mutants showed improved thermostability when assayed with ³H-ZM241385 (antagonist): A54L, V57A, H75A, T88A, G114A, G118A, T119A, K122A, G123A, P149A, E151A, G152A, A203L, A204L, A231L, L235A, V239A.

3. Mutations have been combined to generate mutants in a putative antagonist conformation. Wildtype A_2aR has a Tm of 31° C. with ZM241385 bound.

a. Rant17 A54L + K122A + L235A Tm 48° C. (ZM241385 bound)
b. Rant19 A54L, T88A, V239A +  Tm 47° C. (ZM241385 bound)
A204L
c. Rant21 A54L, T88A, V239A +  Tm 49° C. (ZM241385 bound)
K122A

4. Mutations from the agonist screen have been combined, but have led to only a very low level of improvement in Tm of +2° C.

TABLE (ii)
Stability of best combinations. Receptors were solubilised in 1% DDM
(no glycerol). A melting profile was obtained by heating the solubilised
receptor at different temperatures in absence (apo-state) or presence
of ligand (ligand-occupied state). Data shown are representative of
at least three independent experiments. S.D. is <1° C.
	Tm (° C.)		Tm (° C.)
	−	+		−	+
	ago-	ago-		antago-	antago-
	nist	nist		nist	nist
Wt	21	29	wt	31	32
Rag 1	26	34	Rant 5	42	46
(A184L/R199A/			(A54L/T88A/
L272A)			V239A)
Rag 23	22	38	Rant 21	41	49
(Rag 1 + F79A/			(Rant 5 +
L208A)			K122A)

TABLE (iii)
Summary of results for competition assays of detergent-
solubilised wild-type A2aR and thermo-stable mutant
Rant 21. Values are representative of two independent
experiments. Each data point was assayed in triplicate
and plotted as mean ± SD. Each solubilised receptor
was incubated with ligands for one hour on ice in
binding buffer (50 mM Tris pH 7.5 and 0.025% DDM)
containing 400 mM NaCl. Binding of [3H]ZM241385
(10 nM) in the absence of unlabeled ligand was set
to 100%. Data shown are from two independent
experiments with each data point measured in
triplicate. Incubation of samples with ligands was for
1 hour on ice with [3H]ZM241385 at a concentration
of 10 nM. Ki values were calculated according to the
Cheng and Prusoff equation using the non-linear
regression equation of the software Prism, applying
a KD for [3H]ZM241385 of 12 nM for the wild-type
and 15 nM for Rant 21. Rant 21 did not bind NECA
sufficiently for an accurate Ki determination (hence
indicated as >1 × 10−1). The affinity of Rant21
for agonist binding is weakened 232 fold for R-PIA
and at least by 1900 fold for NECA.
	Ki (M)
	Competitor	wt	Rant 21
	XAC	2.3 × 10−6	2.3 × 10−6
	Theophylline	1.5 × 10−3	0.9 × 10−3
	NECA	7.0 × 10−6	>1 × 10−1
	R-PIA	1.6 × 10−5	3.6 × 10−3

TABLE (iv)
Summary of results for saturation assays of detergent-solubilised
wild-type A2aR and thermo-stable mutants. Values are representative
of three independent experiments. Each data point was assayed
in triplicate and plotted as mean ± SD. Data were fitted
to the Michaelis-Menten equation using the non-linear regression
equation of the software Prism.
	KD (nM)
		[3H]NECA	[3H]ZM241385
	Receptor	(agonist)	(antagonist)
	wt	32 ± 1	12 ± 3
	Rag 1	26 ± 0.4	26 ± 0.5
	Rag 23	21 ± 1	62 ± 1
	Rant 21	>450	15 ± 3

TABLE (v)
Summary of stability of wild-type and mutant receptors in different
detergents. Solubilisation of receptors and detergent exchange
was performed during the IMAC step. S.D. is <1° C. It
was not possible to determine the Tm for some receptor-detergent
combinations, because the receptor was too unstable (†).
	Tm (° C.)
	Agonist-binding		Antagonist-binding
	wt	Rag 23	wt	Rant 21
	0.01% DDM	27	34	25	39
	0.1% DM	23	29	10	28
	0.3% NM	22	28	<4	25
	0.3% NG	†	†	†	22
	0.6% OG	<9	16	†	23
	0.003% LDAO	28	38	32	42
	0.006% FC12	37	39	43	49

EXAMPLE 3: MUTANTS OF THE NEUROTENSIN RECEPTOR (NTR) WITH INCREASED THERMOSTABILITY

1. 340 site-directed mutants have been made between residues 61-400 of NTR.

2. Initially, all of these mutants were assayed for thermostability using an assay measuring ³H-neurotensin (agonist) binding after the heating step. 24 mutations led to a small but significant increase in thermostability: A356L, H103A, D345A, A86L, A385L, Y349A, C386A, K397A, H393A, I116A, F358A, S108A, M181A, R392A, D113A, G209A, L205A, L72A, A120L, P399A, Y351A, V268A, T207A, A155L, S362A, F189A, N262A, L109A, W391A, T179A, S182A, M293A, L256A, F147A, D139A, S100A, K176A, L111A, A90L, N270A.

3. Mutants tested for thermostability by heating in the absence of the agonist were re-tested using a slightly different assay where the mutants were heated in the presence of ³H-neurotensin (Ligand(+) format in FIG. 12). Mutants with improved thermostability are: A69L, A73L, A86L, A90L, H103A, V165A, E166A, G215A, V229A, M250A, I253A, A177L, R183A, I260A, T279A, T294A, G306A, L308A, V309A, L310A, V313A, F342A, F358A, V360A, S362A, N370A, S373A, F380A, A385L, P389A, G390A, R395A.

4. There are also mutants that have a significantly enhanced expression level compared to the wildtype receptor and could be used to boost preceptor production levels for crystallisation: A86L, H103A, F358A, S362A, N370A, A385L, G390A. All of these also have increased thermostability.

5. Preferred combinations are

• • a. Nag7m F358A+A86L+I260A+F342A Tm 51° C. (neurotensin bound)
  • b. Nag7n F358A+H103A+I260A+F342A Tm 51° C. (neurotensin bound)
    Wildtype NTR has a Tm of 35° C. with neurotensin bound.

EXAMPLE 4: IDENTIFICATION OF STRUCTURAL MOTIFS IN WHICH STABILISING GPCR MUTATIONS RESIDE

The structure of the β2 adrenergic receptor has been determined (20, 21), which is 59% identical to the turkey β1 receptor, but with a distinctly different pharmacological profile (22, 23). In order to determine the structural motifs in which the stabilising mutations of the turkey β1 receptor reside, we mapped the mutations onto the human β2 structure (21).

The beta adrenergic receptors were first aligned using ClustalW in the MacVector package; thermostabilising mutations in turkey β1 were highlighted along with the corresponding residue in the human β2 sequence. The human β2 model (pdb accession code 2RH1) was visualised in Pymol and the desired amino acids were shown as space filling models by standard procedures known in the art. The structural motifs in which the stabilising mutations were located, were determined by visual inspection.

Table (vi) lists the equivalent positions in the β2 receptor corresponding to the thermostabilising mutations in βAR-m23 and the structural motifs in which they reside.

As seen from Table (vi), the mutations are positioned in a number of distinct localities. Three mutations are in loop regions that are predicted to be accessible to aqueous solvent (loop). Eight mutations are in the transmembrane α-helices and point into the lipid bilayer (lipid); three of these mutations are near the end of the helices and may be considered to be at the hydrophilic boundary layer (lipid boundary). Eight mutations are found at the interfaces between transmembrane α-helices (helix-helix interface), three of which are either within a kinked or distorted region of the helix (kink) and another two mutations occur in one helix but are adjacent to one or more other helices which contain a kink adjacent in space to the mutated residue (opposite kink) These latter mutations could affect the packing of the amino acids within the kinked region, which could result in thermostabilisation. Another mutation is in a substrate binding pocket.(pocket).

TABLE (vi)
Position in the human β2 structure of the amino acid residues
equivalent to the thermostabilising mutations found in the turkey
β1 receptor and the structural motifs in which they reside.
	Turkey β1	Human β2	Description
Helix 1	I55A	I47	3-helix kink interface	FIG. 18
Helix 1	G67A	A59	lipid boundary
Helix 1	R68S	K60	lipid boundary	FIG. 25
Helix 2	V89L	V81	kink	FIG. 19
Helix 2	M90V	M82	kink	FIG. 20
Helix 2	G98A	G90	pocket
Helix 3	I129V	I121	opposite kink	FIG. 21
	S151E	S143	loop
Helix 4	V160A	V152	lipid
	Q194A	A186	loop
Helix 5	L221V	V213	lipid
Helix 5	Y227A	Y219	helix-helix interface	FIG. 23
Helix 5	R229Q	R221	lipid
Helix 5	V230A	V222	helix-helix interface
Helix 5	A234L	A226	helix-helix interface
Helix 6	A282L	C265	loop	FIG. 24
	D322A	K305	lipid boundary
Helix 7	F327A	L310	lipid
Helix 7	A334L	V317	lipid
Helix 7	F338M	F321	kink	FIG. 22

Such structural motifs, by virtue of them containing stabilising mutations, are important in determining protein stability. Therefore, targeting mutations to these motifs will facilitate the generation of stabilised mutant GPCRs. Indeed, there were several instances where more than one mutation mapped to the same structural motif. For example, the Y227A, V230A and A234L mutations in the turkey β1 adrenergic receptor all mapped to the same helical interface, the V89L and M90V mutations mapped to the same helical kink and the F327A and A334L mutations mapped to the same helical surface pointing towards the lipid bilayer (Table (vi)). Thus, when one stabilising mutation has been identified, the determination of the structural motif in which that mutation is located will enable the identification of further stabilising mutations.

REFERENCES

• 1. S. H. White (2004) Protein Sci 13, 1948-1949.
• 2. C. G. Tate (2001) FEBS Lett 504, 94-98.
• 3. R. Grisshammer, C. G. Tate (1995) Q Rev Biophys 28, 315-422.
• 4. J. U. Bowie (2001) Curr Opin Struct Biol 11, 397-402.
• 5. F. W. Lau, S. Nauli, Y. Zhou, J. U. Bowie (1999) J Mol Biol 290, 559-564.
• 6. Y. Zhou, J. U. Bowie (2000) J Biol Chem 275, 6975-6979.
• 7. S. Faham, D. Yang, E. Bare, S. Yohannan, J. P. Whitelegge, J. U. Bowie (2004) J Mol Biol 335, 297-305.
• 8. Y. Yarden, H. Rodriguez, S. K. Wong, D. R. Brandt, D. C. May, J. Burnier, R. N. Harkins, E. Y. Chen, J. Ramachandran, A. Ullrich, et al (1986) Proc. Natl. Acad. Sci. USA 83, 6795-6799.
• 9. T. Warne, J. Chirnside, G. F. Schertler (2003) Biochim Biophys Acta 1610, 133-140.
• 10. E. M. Parker, E. M. Ross (1991) J Biol Chem 266, 9987-9996.
• 11. E. M. Parker, K. Kameyama, T. Higashijima, E. M. Ross (1991) J Biol Chem 266, 519-527.
• 12. W. J. Degrip (1982) Methods in Enzymology 81, 256-265.
• 13. K. Palczewski, T. Kumasaka, T. Hori, C. A. Behnke, H. Motoshima, B. A. Fox, I. Le Trong, D. C. Teller, T. Okada, R. E. Stenkamp, et al (2000) Science 289, 739-745.
• 14. J. Li, P. C. Edwards, M. Burghammer, C. Villa, G. F. Schertler (2004) J Mol Biol 343, 1409-1438.
• 15. R. Jaenicke, G. Bohm (1998) Current Opinion in Structural Biology 8, 738-748.
• 16. J. Tucker, R. Grisshammer (1996) Biochem J 317 (Pt 3), 891-899.
• 17. W. Schaffner, C. Weissmann (1973) Anal. Biochem. 56, 502-514.
• 18. C. G. Tate (1998) Methods Enzymol 296, 443-455.
• 19. H. M. Weiss, R. Grisshammer (2002) Eur J Biochem 269, 82-92.
• 20. Rasmussen, S. G., Choi, H. J., Rosenbaum, D. M., Kobilka, T. S., Thian, F. S., Edwards, P. C., Burghammer, M., Ratnala, V. R., Sanishvili, R., Fischetti, R. F., Schertler, G. F., Weis, W. I. and Kobilka, B. K. (2007) Nature 15, 383-387.
• 21. Cherezov, V., Rosenbaum, D. M., Hanson, M. A., Rasmussen, S. G., Thian, F. S., Kobilka, T. S., Choi, H. J., Kuhn, P., Weis, W. I., Kobilka, B. K. and Stevens, R. C. (2007) Science 318:1258-1265.
• 22. Minneman, K. P., Weiland, G. A. and Molinoff, P. B. (1980) Mol Pharmacol 17:1-7.
• 23. Parker, E. M., Swigart, P., Nunnally, M. H., Perkins, J. P. and Ross, E. M. (1995) J Biol Chem 270:6482-6487.

Claims

1. A mutant G-protein coupled receptor (GPCR) with increased conformational stability compared to its parent GPCR, which mutant GPCR has an extended lifetime of a particular conformation relative to the same conformation of its parent GPCR under denaturing conditions, the extended lifetime being manifest by retention of ligand binding ability.

2. A mutant GPCR according to claim 1, wherein the mutant GPCR has an extended lifetime under conditions selected from any one or more of heat, a detergent, a chaotropic agent and an extreme of pH.

3. A mutant GPCR according to claim 2, wherein the mutant GPCR has increased conformational thermostability, optionally which is at least 1° C. more stable than its parent.

4. A mutant GPCR according to claim 1, with increased conformational stability of an agonist or antagonist conformation.

5. A mutant GPCR according to claim 1, wherein the mutant GPCR is a Class 1 GPCR.

6. A mutant GPCR according to claim 1, wherein the mutant GPCR is a Class 2 GPCR.

7. A mutant GPCR according to claim 1, wherein the mutant GPCR is a Class 3 GPCR.

8. A polynucleotide encoding a mutant GPCR according to claim 1.

9. A host cell comprising a polynucleotide according to claim 8.

10. A mutant GPCR according to claim 1, which is in a solubilized form.

11. A mutant GPCR according to claim 1, which is substantially free of other proteins.

12. A mutant GPCR according to claim 1, which is immobilized to a solid support.

13. A mutant GPCR according to claim 1, wherein the mutant GPCR is not a human cannabinoid receptor.

14. A polynucleotide according to claim 8, wherein the mutant GPCR is not a human cannabinoid receptor.

15. A host cell according to claim 9, wherein the mutant GPCR is not a human cannabinoid receptor.

16. A crystallized form of a mutant GPCR according to claim 1.

17. A method comprising crystallizing a mutant GPCR of claim 1.

18. A method comprising

obtaining a mutant GPCR of claim 1, and

screening for a ligand that binds to the mutant GPCR.

19. The method of claim 18, wherein the mutant GPCR is coupled to a solid surface of a biosensor chip.

20. A method comprising

obtaining a mutant GPCR of claim 1, and

developing an assay involving binding of a ligand to the mutant GPCR.