Mini-G proteins: Novel tools for studying GPCRs in their active conformation

1G protein-coupled receptors (GPCRs) are integral membrane proteins that play a ubiquitous role in intercellular communication throughout the human body. Binding of an agonist at the extracellular surface of the inactive receptor results ultimately in a conformational change at its intracellular surface that allows G protein coupling. The conformation of the bound G protein is altered around the nucleotide binding pocket, resulting in GDP release, binding of GTP and subsequent activation of the G protein. There has been great progress in the structure determination of GPCRs in the inactive state bound to antagonists which highlights common structural features of the GPCR superfamily, a common mode of activation and a conserved mechanism for the activation of G proteins. However, it is still a major challenge to determine the structures of receptors in their fully active state, which can be defined structurally as the conformation bound to a heterotrimeric G protein or pharmacologically as the high-affinity agonist binding state of the receptor. The range of GPCR structures now available allows a clear distinction between agonist-bound structures either in the inactive state, in an active-intermediate state or in an active state. It is, however, difficult to determine structures of receptors in an active state due to the instability of activated GPCRs and receptor-bound G proteins.

2Four solutions have been developed to determine the active structures of GPCRs. The first active structure determined was of opsin stabilized at low pH and bound to a C-terminal fragment of the G protein transducin. This methodology has not been successful for any other GPCR. The second active state structure of a GPCR was determined of the β2-adrenergic receptor (β2AR) bound to a single-chain camelid antibody (nanobody) that mimics the pharmacological effects of a heterotrimeric G protein. This has been applied successfully to other GPCRs, but currently the methodology still requires the immunization of llamas with purified receptors, which makes it difficult to apply to many unstable GPCRs. The third methodology was to crystallise β2AR with the entire heterotrimeric G protein. This was a great achievement and showed for the first time how a GPCR facilitated nucleotide exchange in the G protein. However, no other structures of a GPCR coupled to a heterotrimeric G protein have been published and the resolution of the GPCR portion of the structure was poor and required the previously determined active state structure of β2AR bound to a nanobody to allow the model to be built. The fourth methodology was based on the observation that only the GTPase domain of the Gα subunit made significant contacts with active β2AR and therefore this domain was engineered and thermostabilised to generate a mini-G protein, mini-Gs. Like the nanobodies, mini-Gs recapitulated all the receptor pharmacology upon coupling, but in addition, mini-Gs could couple to any Gs-coupled receptor whereas nanobodies bound to only a specific receptor. The structure of the adenosine A2A receptor (A2AR) coupled to mini-Gs was determined recently, which showed similar structural re-arrangements observed in the β2AR–Gs complex.

3The concept of mini-G proteins shows great promise for accelerating the rate of structure determination of GPCRs in their active states, from providing biophysical and pharmacological data on receptors in the active state, as well as providing a potential binding partner in co-crystallisation applications. However, there are four families of Gα subunits (Fig 1; Gαs, Gαi, Gαq, and Gα12) that show different specificities for various GPCRs. Thus to be truly useful as tools in structural biology, at least one member from each family needs to be converted into a mini-G protein. Here we report the development of mini-G proteins for all the major Gα families. We also describe five different assays that can be used to characterize the binding of the mini-G proteins to GPCRs and show in three cases that the complexes can be purified by size exclusion chromatography. The two different methodologies for generating the mini-G proteins can be applied easily to any other Gα subunit, opening the doorway to studies on potentially any GPCR from any species.

4The β1-adrenergic receptor (β1AR) agonist isoproterenol hydrochloride and inverse agonist ICI118551 hydrochloride were from Sigma-Aldrich. The adenosine A2A receptor (A2AR) agonist NECA and antagonist ZM241385 were also from Sigma Aldrich. Serotonin 5HT1B receptor (5HT1BR) agonist donitriptan hydrochloride and selective antagonist SB224289 hydrochloride were from Santa Cruz Biotechnology; the agonist sumatriptan succinate was from Cayman chemical. Angiotensin II receptor (AT1R) agonist angiotensin II was from Tocris. All radioactive ligands were from PerkinElmer.

5Two different A2AR constructs were used during this work. For SEC experiments using purified receptor, an A2AR construct was used that contained an N-terminal thioredoxin fusion protein to increase the molecular weight of the receptor. Without this fusion protein, A2AR and the mini-G protein had identical mobility on SDS-PAGE, thus making it difficult to visualise the separate components when analyzing a complex. The thioredoxin-A2AR fusion protein consisted of an N-terminal cleavable leader sequence (gp67), His10 tag and TEV protease cleavage site, followed by thioredoxin, which was connected to wild-type human A2AR (residues 6–316) through an EAAAKA linker. A2AR contained the N154A mutation to remove a potential N-linked glycosylation site. For all other experiments, a C-terminally truncated human A2AR construct was used (residues 1–317), which contained a C-terminal His10 tag and TEV protease cleavage site and the N154A mutation to remove the potential N-linked glycosylation site. Both constructs were expressed using the baculovirus expression system as described previously. Cells were harvested by centrifugation 72 hours post infection, resuspended in hypotonic buffer (20 mM HEPES pH7.5, 1 mM EDTA, 1 mM PMSF), flash-frozen in liquid nitrogen and stored at –80°C until use. Purification of the receptor was performed in DDM using Ni2+-affinity chromatography followed by SEC essentially as described previously.

6A truncated version of wild type turkey β1AR (construct βAR6;) contained truncations at the N-terminus and the C-terminus and a C-terminal His6 tag for purification, and was expressed using the baculovirus expression system at 27°C as described previously. Cells were harvested by centrifugation 48 hours post infection, resuspended in hypotonic buffer (20 mM Tris-HCl pH8, 1 mM EDTA, 1 mM PMSF), flash-frozen in liquid nitrogen and stored at -80°C until use.

7Wild type AT1R (residues 1–359) had a C-terminal factor X cleavage site followed by GFP and a His10 tag for purification, and was expressed using the tetracycline-inducible mammalian expression system as a stable cell line in HEK293 cells. Cells were grown in DMEM containing 5% tetracycline-free FBS until they were 80% confluent and then tetracycline was added to a final concentration of 1 μg/ml. Cells were grown for 24 hours and then harvested, and resuspended in PBS, flash frozen in liquid nitrogen and stored at -80°C until use.

8NTSR1 was expressed as described previously. The baculovirus construct NTSR1 consisted of the hemagglutinin signal peptide and the Flag tag, followed by the wild-type rat NTSR1 (residues 43–396) and a C-terminal His10 tag. Recombinant baculovirus was generated using a modified pFastBac1 transfer plasmid (Invitrogen). Trichoplusia ni cells were infected with recombinant virus, and the temperature was lowered from 27°C to 21°C. Cells were harvested by centrifugation 48 hours post infection, resuspended in hypotonic buffer (10 mM HEPES pH 7.5, 10 mM MgCl2, 20 mM KCl), flash-frozen in liquid nitrogen and stored at -80°C until use.

9Wild-type 5HT1BR (residues 34–390) was modified to contain a C-terminal TEV cleavage site and a His10 tag, cloned into plasmid pBacPAK8 and recombinant baculoviruses were prepared using the FlashBAC ULTRA system (Oxford Expression Technologies). Trichoplusia ni cells were grown in ESF921 media (Expression Systems) to a density of 3x106 cells/ml, infected with 5HT1BR baculovirus and incubated for 48 h at 27°C for expression. Purification of the receptor was performed in either DDM or LMNG using Ni2+-affinity chromatography followed by SEC.

10For constructs see Fig 2 and S1–S3 Figs. Expression, purification and stability measurements by differential scanning fluorimetry (DSF) of the mini-G proteins as well as the non-lipidated Gβ1γ2 dimer, were performed following the protocols described in. The stability of mini-G proteins was also determined in detergent using native DSF (NanoTemper Prometheus). Mini-G proteins (2 mg/ml) in 50 mM HEPES pH 7.5 (KOH), 20 mM MgCl2, 50 mM NaCl, 1 μM GDP were mixed with either no detergent (control), 0.1% LMNG or 0.1% DDM. Samples were incubated on ice (minimum 30 min) prior to heating on the Prometheus (20% excitation, 15°C-85°C, rate of 2.0°C/min) and the onset of scattering determined. The conditions used for the native DSF measurements mimic those used in the formation of a mini-G protein–GPCR complex (contains 1 μM GDP and detergent), which differ from the conditions used for good stability of the GDP-bound state of mini-G proteins (contains 1 mM GDP and no detergent).

11The thioredoxin fusion construct of A2AR was purified in DDM essentially as described previously. A2AR and mini-G protein were mixed in a 1:1.2 molar ratio, apyrase (0.1 U/ml final concentration) was added and the sample was incubated overnight on ice before loading onto a Superdex S200 10/300 size exclusion column (10 mM HEPES pH 7.5, 100 mM NaCl, 1 mM MgCl2, 100 μM NECA, 0.02% DDM; 4°C, 0.5 ml/min). Peak fractions were analysed by SDS-PAGE.

12Donitriptan-bound 5HT1BR purified in DDM was mixed with mini-Go1 in a 1:1.2 molar ratio. Apyrase (0.1 U/ml final concentration) was added and the sample was incubated 4 h on ice before loading onto a Superdex S200 10/300 size exclusion column (20 mM HEPES pH 7.5, 100 mM NaCl, 1 mM MgCl2, 1 μM donitriptan, 0.03% DDM; 4°C, 0.5 ml/min). Peak fractions were analysed by SDS-PAGE.

13Insect cell membranes containing a total of 20 μg (560 pmol) wild-type A2AR (20 x 106 cells) were solubilized for 30 min on ice in 40 mM HEPES pH7.5, 500 mM NaCl, 2 mM MgCl2, 2 U/mL apyrase (Sigma-Aldrich), and 0.5% (v/v) DDM (final volume of 2 ml, concentration of A2AR 280 nM). Insoluble material was removed by ultracentrifugation (30 min, 4°C, 135,000 xg). The supernatant was divided into aliquots for the subsequent assay. To 900 μl of the supernatant was added either the agonist NECA or the inverse agonist ZM241385 (negative control), both at a final concentration of 60 μM. GFP-mini-Gs (6 μg; 110 pmol, final concentration 122 nM) was then added and allowed to bind for 90 min on ice before loading 200 μl onto a Superdex S200 10/300 size exclusion column (buffer 20 mM HEPES pH 7.5, 100 mM NaCl, 10 mM MgCl2, 1 μM NECA or ZM241385, 0.03% DDM, 4°C, flow rate 0.45 ml/min). The control sample contained 6 μg GFP-mini-Gs only in 500 μl assay buffer. GFP fluorescence was detected by a Hitachi fluorometer (mV) set to an excitation of 488 nm and an emission of 525 nm.

14Insect cell membranes containing a total of 8 μg (178 pmol) wild-type β1AR (30 x 106 cells) were solubilized for 30 min on ice in 20 mM Tris-HCl pH8, 500 mM NaCl, 5 mM MgCl2, 2 U/mL apyrase and 0.5% (v/v) DDM (final volume 2 ml, β1AR concentration 90 nM). Insoluble material was removed by ultracentrifugation (30 min, 4°C, 135,000 xg). The supernatant was divided into aliquots for the subsequent assay. Isoprenaline (100 μM final concentration) or ICI118551 (10 μM final concentration) were added to 500 μl of the supernatant. GFP-mini-Gs (6 μg, 110 pmol) was then added to give a final concentration of 122 nM and allowed to bind for 90 min on ice before loading 200 μl onto a Superdex S200 10/300 size exclusion column (buffer 20 mM HEPES pH 7.5, 100 mM NaCl, 10 mM MgCl2, 1 μM isoprenaline or ICI118551, 0.03% DDM, 4°C, flow rate 0.45 ml/min). The control sample contained 6 μg GFP-mini-Gs only in 500 μl assay buffer.

15When detergent-solubilized unpurified receptor was used, insect cells expressing 610 pmol 5HT1BR (40 x 106 cells) were resuspended in 20 mM HEPES pH 7.5, 100 mM NaCl, 10 mM MgCl2, 2 U/ml apyrase to a final cell density of 20 x 106 cells/ml and solubilized with 0.5% DDM (45 min, 4°C, final volume 2 ml, 5HT1BR concentration 305 nM). Insoluble material was removed by ultracentrifugation (30 min, 4°C, 135,000 xg). The supernatant was divided into 900 μl aliquots for the subsequent assay. GFP-mini-Go1 (5 μg, 100 pmol) was added (final concentration 111 nM) with either donitriptan or SB224289, each to a final concentration of 100 μM, and allowed to bind for 90 min on ice before loading 500 μl onto a Superdex S200 10/300 size exclusion column. The control sample contained 5 μg GFP-mini-Go1 in 500 μl assay buffer.

16In some FSEC experiments, purified 5HT1BR was used. Donitriptan-bound, purified receptor (120 μg; 3 nmol) in either LMNG or DDM was incubated for 90 min on ice with 4 μg (60–80 pmol) either of GFP-mini-Gi1, GFP-mini-Go1 or GFP-mini-Gs (negative control) in a final volume of 450 μl. Samples (200 μl) were then loaded onto Superdex S200 10/300 size exclusion column (buffer 20 mM HEPES pH 7.5, 100 mM NaCl, 10 mM MgCl2, 1 μM donitriptan, 0.03% DDM or 0.001% LMNG buffer, 4°C, flow rate 0.45 ml/min). The final concentration of purified 5HT1BR was 6.7 μM and the final concentrations of the mini-G proteins were as follows: GFP-mini-Gs/i1, 133 nM; GFP-mini-Gi1, 167 nM; GFP-mini-Go1, 170 nM; GFP-mini-Gs, 167 nM; GFP-mini-Gi1β1γ2, 90 nM.

17Membranes prepared from insect cells expressing β1AR (50 x 106 cells) were solubilized in 20 mM Tris-HCl pH8, 500 mM NaCl, 3 mM imidazole, 2 U/ml apyrase 0.5% DDM (1 hour, 4°C, final volume 8 ml, β1AR concentration 37.5 nM). Insoluble material was removed by ultracentrifugation (30 min, 4°C, 135,000 xg) and the supernatant was divided into two aliquots. The agonist isoprenaline was added to one sample (final concentration 10 μM) and the inverse agonist ICI118551 was added to the other (final concentration 1 μM). Samples were then aliquoted 200 μl per well into a black Ni2+-coated 96-well plate (Pierce; Thermo Fisher). The receptor was allowed to bind via its His tag for 1 h on ice. The supernatant was then aspirated and 200 μl GFP-mini-Gs at varying concentrations (0 to 2.8 μM) were added and incubated for a further 90 min on ice. The supernatant was then removed by aspiration and each well washed 4 times with buffer A (10 μM isoprenaline (agonist), 20 mM Tris-HCl pH8, 100 mM NaCl, 1 mM MgCl2, 1 mg/mL BSA, 30 mM imidazole, 0.03% DDM,) or buffer B (1 μM ICI118551 (inverse agonist), 20 mM Tris-HCl pH8, 100 mM NaCl, 1 mM MgCl2, 1 mg/mL BSA, 30 mM imidazole, 0.03% DDM). Elution of the receptor–GFP-mini-Gs complex from the sides of the well to make a homogeneous solution was performed with 200 μl of the respective wash buffers that contained 300 mM imidazole. Fluorescence was then measured using a Pherastar plate reader (BMG Labtech, Inc.) with excitation at 485 nm and emission at 520 nm. ΔF data (fluorescence agonist condition minus fluorescence antagonist condition) corresponding to specific binding were analysed by non-linear regression using GraphPad Prism version 5.0 (GraphPad Software, San Diego, CA) and apparent KD values derived from one site-specific binding analysis.

18The assay was performed essentially as described above for β1AR, but the buffer conditions were different. Solubilisation of insect cell membranes (40 x 106 cells) was performed in 10 ml of 20 mM Tris-HCl pH8, 500 mM NaCl, 10 mM imidazole, 2 U/ml apyrase and 0.5% DDM (1 h, 4°C, final volume 10 ml, A2AR concentration 112 nM). After ultracentrifugation, the agonist NECA (10 μM final concentration) was added to one supernatant sample and the inverse agonist ZM241385 (10 μM final concentration) to the other. Washing buffers for A2AR were buffer C (10 μM NECA, 20 mM Tris-HCl pH8, 100 mM NaCl, 1 mM MgCl2, 1 mg/mL BSA, 50 mM imidazole, 0.03% DDM) or buffer D (10 μM ZM241385, 20 mM Tris-HCl pH8, 100 mM NaCl, 1 mM MgCl2, 1 mg/mL BSA, 50 mM imidazole, 0.03% DDM).

19Insect cells expressing 5HT1BR (50 x 106 cells) were solubilized with buffer containing 10 μM Donitriptan, 20 mM Tris-HCl pH8; 500 mM NaCl; 10 mM imidazole, 2 U/ml apyrase and 0.5% DDM (1 h, 4°C, final volume 6 ml, 5HT1BR concentration 127 nM). Insoluble material was removed by ultracentrifugation (30 min, 4°C, 135,000 xg) and 200 μl of supernatant was then aliquoted per well into a black Ni2+-coated 96-well plate. The receptor was allowed to bind via its His tag for 1 h on ice. The supernatant was then aspirated and 200 μl either of GFP-mini-Go1, GFP-mini-Gs/i1 or GFP-mini-Gs (negative control) at varying concentrations (from 0 to 5 μM) were added and incubated for a further 90 min on ice. The supernatant was then removed by aspiration and each well washed 4 times with buffer E (1 μM Donitriptan, 20 mM Tris-HCl pH8, 100 mM NaCl, 1 mM MgCl2, 1 mg/mL BSA, 50 mM imidazole, 0.03% DDM). Elution was carried out with 200 μL of buffer E containing 300 mM imidazole. ΔF data (fluorescence Go1 condition minus fluorescence Gs condition) corresponding to specific binding were analysed by non-linear regression using GraphPad Prism version 5.0 (GraphPad Software, San Diego, CA) and apparent KD values derived from one site-specific binding analysis.

20Insect cells expressing 5HT1BR were resuspended in 1 ml of assay buffer (20 mM HEPES pH7.5, 100 mM NaCl, 1 mM MgCl2, 1 mM ascorbate, 20 μM pargyline) at a final concentration of 2 x 106 cells/ml. Cells were sheared by 10 passages through a bent 26G needle. The supernatant was diluted 10-fold in assay buffer and aliquots (900 μl) taken for each sample. Mini-G protein (100 μl, 25 μM final concentration) or buffer (negative control) was added. The mixture was aliquoted into a 0.2 ml PCR plate, 96 μl per well. Sumatriptan (12 μl), prepared in assay buffer also containing 2 U/ml apyrase, was added to each well (final concentrations in the range of 100 pM to 1 mM). Non-specific binding was determined in the presence of 100 μM donitriptan. Samples were mixed and incubated at 4°C for 2 h. [3H]-GR125743 (12 μl) was added at its apparent KD (10 nM) concentration. Samples were mixed and incubated at 4°C for 2 h before filtering through 96-well glass fibre GF/B filter plate (Merck Millipore) and washing with ice-cold assay buffer. Filters were dried, punched into scintillation vials and 4 ml Ultima Gold scintillant (Perkin Elmer) were added. Radioactivity was quantified by scintillation counting (1 min per sample) using a Tri-Carb counter (Perkin Elmer), and Ki values were determined using GraphPad Prism version 5.0 (GraphPad Software, San Diego, CA)

21Membranes from Trichoplusia ni cells expressing wild-type human A2AR were resuspended in Tm buffer (25 mM HEPES pH 7.5, 100 mM NaCl, 1 mM MgCl2) and homogenized by ten passages through a 26G needle. Mini-G protein was added at a final concentration of 25 μM. 3H-NECA and unlabeled NECA, prepared in assay buffer containing 2 U/ml apyrase, were mixed in a molar ratio of 1:5 and added to the membranes to give a final concentration of 1 μM (approximately ten-fold above the apparent KD). The samples were incubated at room temperature for 1 h, then chilled on ice for 30 min. Decylmaltoside (DM) was added to a final concentration of 0.13%, and samples were incubated on ice for 1 h. Cell debris and insoluble material were removed by centrifugation (5 min, 20,000 xg, 4°C) and the supernatant was aliquoted (120 μl) into PCR strips. Samples were heated to the desired temperature for exactly 30 min, then quenched on ice for 30 min. Samples (50 μl) were loaded onto gel-filtration resin (Toyopearl HW-40F) packed into a 96-well filter plate (Millipore), which was centrifuged to separate receptor-bound from free radioligand. Nonspecific binding was determined in the presence of 200 μM unlabelled NECA. Radioactivity was quantified by liquid scintillation counting using a MicroBeta TriLux scintillation counter (PerkinElmer). Data were analysed by nonlinear regression using GraphPad Prism software. Apparent Tm values were derived from sigmoidal dose-response analysis performed by non-liner regression. Results represent the mean ± SEM of two independent experiments, performed in duplicate.

22Cell pellets from 10 ml of insect cell cultures were resuspended in 1.8 ml buffer containing DDM to give a final buffer composition of 50 mM Tris-HCl pH 7.4, 100 mM NaCl, 1 mM MgCl2, 1% (w/v) DDM. The samples were placed on a rotating mixer at 4°C for 1 hour. Cell debris and non-solubilized material were removed by ultracentrifugation (152,800 xg, 4°C, 30 min), and the supernatant containing detergent-solubilized NTSR1 was used to test for thermal stability in the presence of NTS and mini-G proteins. For thermal denaturation curves, the supernatants were diluted 6.67-fold into assay buffer (50 mM Tris-HCl pH 7.4, 100 mM NaCl, 1 mM MgCl2) containing 22.5 μM mini-G protein and 10 nM 3H-NTS and incubated for 1 hour on ice. After addition of apyrase (0.25 units/ml, NEB), the sample was placed on ice for an additional 30 min. Samples (120 μl aliquots) were exposed to different temperatures between 0°C and 60°C for 30 min and placed on ice. Separation of receptor–ligand–mini-G protein complex from free 3H-NTS (100 μl) was achieved by centrifugation-assisted gel filtration (spin assay) using Bio-Spin 30 Tris columns (BioRad), equilibrated with RDB buffer [50 mM Tris-HCl pH7.4, 1 mM EDTA, 0.1% (w/v) DDM, 0.2% (w/v) CHAPS, 0.04% (w/v) CHS], essentially as described previously. Control reactions on ice were recorded at the start and at the end of each denaturation experiment. The percentage of activity remaining after heat exposure was determined with respect to the unheated control. Data were analyzed by nonlinear regression using a Boltzmann sigmoidal equation in the Prism software (GraphPad).

23HEK 293 cells expressing wild type AT1R were resuspended in a radioligand binding assay buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.1% BSA, 40 μg/ml bacitracin) and homogenized by sonication (4 sec pulse). Mini-G protein and apyrase were added at a final concentration of 25 μM and 0.1 U/ml, respectively. 125I-Ang II and unlabeled Ang II were added at a concentration of 0.5 nM and 25 nM respectively (approximately 50 times the apparent KD value). The sample was incubated at room temperature (20°C) for an hour, chilled on ice for 10 minutes and then digitonin was added to a final concentration of 1% and incubated on ice for an hour. Insoluble material was removed by centrifugation (2 min, 20,000 xg, 4°C). The reaction mix was split into a number of 115 μl aliquots and each was incubated at various temperatures for exactly 30 minutes. The reactions were then quenched on ice for 5 minutes. 125I-Ang II bound to AT1R was separated from unbound 125I-Ang II using centrifugation-assisted gel filtration column, essentially as described previously. Non-specific binding was determined using a 500-fold excess of cold ligand. Radioactivity was measured using liquid scintillation counting. Data was analysed by non-linear regression using GraphPad prism software and apparent Tm values were derived by non-linear regression of the sigmoidal dose-response curve.

24The recently designed minimal G protein, mini-Gs, comprises only the GαGTPase domain from Gs and 3 deletions and 7 mutations to thermostabilise it (Fig 2). Mini-Gs coupled to both the β1-adrenergic receptor (β1AR) and the adenosine A2A receptor (A2AR), and resulted in the same increase in agonist affinity as observed for heterotrimeric Gs coupling. However, there are 4 families of Gα subunits (Fig 1) and GPCRs couple to distinct G proteins depending upon their physiological function. Therefore, to provide tools for the structure determination of GPCRs in their fully active state, it was necessary to develop versions of mini-G proteins for at least one member from each of the other families. All of the mutations and deletions used to create mini-Gs are located within conserved regions of the Gα subunit (Fig 2). Therefore, in theory, these mutations were potentially transferable to the other Gα families, allowing the production of a panel of mini-G proteins capable of coupling to any GPCR.

25Archetypical members from each Gα family were selected and include the following: Golf from the Gs family, Gi1, Go1, Gz and Gt from the Gi family, Gq and G16 from the Gq/11 family, and G12 from the G12/13 family. The mutations required to convert Gαs into mini-Gs were transferred en bloc to the selected Gα proteins to produce a mini-G protein version of each (Fig 2). These mutations were the following: (i) deletion of all amino acid residues N-terminal of Ile/LeuHN43; (ii) deletion of the α-helical domain between residues HH1S2.12 and the Thr, three residues N-terminal to IleS2.1, and replacement with an 8 amino acid residue linker; (iii) deletion of 10 amino acid residues of switch III between TyrS4H3.4 and Asn/SerS4H3.15; (iv) mutating 7 residues to D49S1H1.3, N50S1H1.4, D249S4.7, D252S4H3.3, D272H3.8, A372H5.4, I375H5.7. Residue numbers are for Gαs and superscripts refer to the CGN system for comparing residues in G proteins. Initial characterization of each mini-G protein was performed by assessing expression in Escherichia coli and purification by Ni2+-affinity chromatography and size exclusion chromatography (SEC). Four out of the eight engineered mini-G proteins (mini-Golf, mini-Gi1, mini-Go1 and mini-G12) fulfilled these initial criteria i.e. they were all stable enough in their basal conformation to allow high-yield expression and purification. The yield of purified mini-G protein per litre of culture and their stability as measured by differential scanning fluorimetry (in parentheses) are as follows: mini-Gs, 100 mg/L (65°C); mini-Golf, 80 mg/L (65°C); mini-Go1 100 mg/L (64°C); mini-G12 25 mg/L (73°C). The worst expressed of the four new mini-G proteins was mini-Gi1, so an additional mutation G217D was incorporated and the truncation at the N-terminus shortened, which increased the yield of pure protein to 12 mg/L, although the stability was only 48°C. Thus, mini-Golf, mini Gi1, mini-Go1 and mini-G12 were all of sufficient stability to be used to test their ability to couple to relevant GPCRs. The amino acid sequences of the mini-G proteins are given in S1 Fig.

26Four mini-G proteins were not expressed in E. coli, namely mini-Gz, mini-Gt, mini-G16 and mini-Gq (amino acid sequences are given in S2 Fig). The failure of the en bloc transfer of the deletions and mutations from mini-Gs, despite the high conservation of G protein structures, highlights our lack of understanding of the folding of these proteins. Indeed, it is well known that an accessory factor, Ric8, is required for the efficient folding of Gq in mammalian cells, and other unknown factors may also be required. For the purpose of this study, we therefore did not perform any further development of mini-Gt1 and mini-Gz, given that two other members of the Gi family, mini-Gi1 and mini-Go1, already gave stable mini-G proteins. In contrast, as neither member of the Gq family tested produced a stable mini-G protein, we decided to develop alternative strategies to make a usable version of mini-Gq, whilst further work on mini-G16 was terminated. The successful engineering of a version of mini-Gq chimera will be discussed later.

27The ultimate goal of developing mini-G proteins is the structure determination of GPCRs in the fully active state bound to an agonist and a mini-G protein. In the simplest format, this necessitates the purification of the GPCR in detergents and forming the G protein–GPCR complex from the purified components in vitro. It was therefore essential to devise some simple assays that could assess whether a mini-G protein had coupled to a GPCR in detergent solution. This turned out to be not as straightforward as originally anticipated due to the potential instability of either the GPCR and/or mini-G protein in either their inactive and/or active conformations. These issues were not obvious when the original work on the development of mini-Gs was performed, because mini-Gs is one of the most stable mini-G proteins developed and also the thermostabilised β1-adrenergic receptor (β1AR) and the wild type adenosine A2A receptor (A2AR) were both much more stable than other GPCRs. We therefore developed five separate assays for assessing whether a mini-G protein coupled to a GPCR and/or formed a stable complex in detergent. These were all first tested using mini-Gs coupling to β1AR and A2AR. Each assay has its own limitations, which are often apparent in the subsequent sections where they were used on less stable receptors and the newly developed mini-G proteins, and these are discussed below. The five different assays that were used are the following: (i) agonist affinity shift assay; (ii) thermostability assay (TSA); (iii) fluorescence-based saturation binding analysis (FSBA) of GFP-mini-G protein binding; (iv) fluorescence-detection size exclusion chromatography (FSEC); (v) size exclusion chromatography (SEC) of purified complex. A brief rationale for the use of each assay with their advantages and disadvantages are given below.

28The development of mini-Gs relied on the agonist affinity shift assay to identify those mutants that coupled to β1AR. It is generally considered that the defining feature of G protein coupling is an increase in the affinity of an agonist for the G protein–GPCR complex compared to the GPCR alone. For example, wild type β2AR binds an agonist 100-fold more tightly when coupled to a G protein than the receptor alone. However, the shift in agonist affinity in other receptors is often considerably smaller than that observed for β2AR, such as the 10-fold shift in agonist affinity observed in β1AR. However, the advantage of this assay is that it can be performed using standard pharmacological procedures in high-throughput, using receptors in either membrane preparations or solubilized in detergent. Assays may use either a radiolabelled agonist in saturation binding experiments or, more usually, a radiolabelled antagonist in competition binding experiments (and see experiments below on the serotonin 5HT1B receptor). The advantage of this assay is that it is very sensitive and can be performed on membrane-bound receptors i.e. in a format where the receptor is most stable in all conformations. The disadvantage of this assay is that it requires suitable radioligands for the receptor under study and these are not readily available for all GPCRs.

29The thermostability of a detergent-solubilised GPCR depends upon the type of detergent used and whether the receptor is either ligand-free, agonist-bound or antagonist-bound. In addition, the receptor stability tends to be increased by an increase in affinity and/or decrease in the off-rate of the ligand. Often, the agonist bound state is one of the least stable conformations of a receptor, presumably because agonists increase the probability of transitions to a fully active state. In the inactive state there is close packing of the intracellular surface of the transmembrane α-helices. Upon activation, the outward movement of helices 5 and 6 disrupts this close packed structure and creates a crevice where the C-terminus of the G protein binds, thus allowing G protein coupling. The structures of non-rhodopsin GPCRs in the fully active state have been determined only when they have been stabilized through binding of a heterotrimeric G protein, a conformation-specific nanobody or a mini-G protein. The interface between a GPCR and a G protein is over 1000 Å2, and is therefore predicted to increase the thermostability of the agonist-bound GPCR–G protein complex compared to the agonist-bound GPCR. This was observed for both β1AR and A2AR, which were consistently more stable in the agonist-bound state when coupled to mini-Gs in a variety of different detergents compared to when mini-Gs was absent.

30A typical thermostability assay measures how much of a radiolabelled agonist remains bound to a detergent-solubilised receptor after heating at different temperatures for 30 minutes. The advantage of this assay is that it is fast and high-throughput and can be performed in any detergent of choice. Another advantage is that the agonist–GPCR–mini-G protein complex can be pre-formed in membranes, which may stabilise the receptor upon detergent solubilisation. If there is a shift in thermostability in the presence of a mini-G protein, then this is strongly suggestive of binding or coupling.

31FSEC is a rapid methodology for assessing whether a membrane protein fused to GFP is stable in detergent by performing SEC on an unpurified detergent solubilisate and monitoring GFP fluorescence in the eluate. A membrane protein stable in detergent gives a symmetrical peak at a size consistent with the molecular weight of the membrane protein plus the mass of specifically bound detergent and lipid. By fusing GFP to the N-terminus of mini-G proteins (S3 Fig), it was possible to use FSEC to monitor whether a stable complex was formed between the mini-G protein and a GPCR. The GFP-mini-Gs fusion protein has a molecular weight of 54 kDa and migrated with a retention volume of 15.1 ml on FSEC. When this was mixed with either unpurified DDM-solubilised β1AR or A2AR in the presence of an agonist, then an additional peak was observed at 12.1–12.5 ml (Fig 3A and 3C), which was consistent with the molecular weight of the detergent-solubilised receptor bound to GFP-mini-Gs (~180 kDa). This additional peak was not observed if the receptors were bound to an inverse agonist. An additional peak was sometimes observed at a retention volume of 8 ml, which corresponds to the void volume of the SEC column and was due presumably to aggregates of GFP-mini-Gs.

32The advantage of this assay is that it is a quick assessment of whether a GPCR forms a complex with a mini-G protein, because the receptor does not need to be purified and the SEC experiment takes under an hour. However, a major limitation is that only small amounts of GFP-mini-Gs can be used per experiment to avoid saturation of the detector and producing a very broad peak that would obscure the presence of the complex between the GPCR and GFP-mini-G protein. An additional limitation is the amount of membranes that can be solubilized efficiently using 0.5% DDM (to ensure maximal stability of mini-Gs; see below) and maintaining the small volume (500 μl) required for loading on the FSEC column (24 ml). Thus in the data shown (Fig 3A and Fig 3C) the amount of receptor is limiting and only a proportion of the GFP-mini-Gs is bound to receptor and is shifted in apparent molecular weight. In addition, the receptor–mini-G protein complex must be detergent-stable for a peak to be observed. Many GPCR–G protein complexes are too unstable to be observed in DDM and therefore it is essential to assess milder detergents such as LMNG (see section on the serotonin 5HT1B receptor).

33To determine the affinity of mini-G protein binding to a receptor, the fluorescence-based saturation binding assay (FSBA) was developed. In this assay, the amount of the GFP-mini-G protein specifically bound to an immobilized receptor was determined using a fluorescent plate reader. As proof of principle, DDM-solubilized β1AR or A2AR were immobilized onto Ni2+-coated wells of a 96-well plate via their C-terminal poly-histidine tag, in the presence of either an agonist or inverse agonist. GFP-mini-Gs was then added at increasing concentrations. After washing to remove any non-specifically bound GFP-mini-Gs, the amount of GFP-mini-Gs fluorescence was measured (Fig 3B and 3D). GFP-mini-Gs showed a specific saturated binding to the receptor with apparent KD values of 200 ± 1 nM (n = 2) and 430 ± 24 nM (n = 2) for GFP-mini-Gs binding to β1AR and A2AR, respectively.

34The FSBA is a simple assay for determining the affinity of mini-G protein binding to a receptor in vitro. However, it must be appreciated that the apparent affinity determined may be specific only for the conditions in the assay. In particular, the type of detergent used may have a profound effect on the affinity, especially if it slightly destabilizes the active state of the receptor. The agonist may also affect the apparent affinity of the mini-G protein, depending on how effective the agonist is in stabilizing the active state of the receptor. However, the FSBA remains a useful tool for biophysical analyses of mini-G protein binding to a receptor.

35The ultimate biochemical assay for observing coupling of mini-G proteins to a receptor is combining the purified components in vitro and then observing the co-elution of the relevant proteins on SEC. Purified A2AR and purified mini-Gs were mixed at a molar ratio of 1:1.2 in the presence of the agonist NECA, the complex allowed to form and then separation was performed by SEC. The A2AR–mini-Gs complex resolved as a predominant peak with an apparent molecular weight of 153 kDa compared with 133 kDa for the receptor alone and 22 kDa for mini-Gs alone. SDS-PAGE analysis confirmed the presence of both A2AR and mini-Gs in fractions from the 153 kDa complex (Fig 3E).

36The advantage of using purified components and SEC for analyzing complex formation is that complex formation is observed unambiguously. The conditions for complex formation can be refined and the stability of the complex can be assessed readily after a period of days by repeating the SEC. These data are essential for successful determination of the structure of a GPCR–mini-G protein complex. The disadvantage of this assay is that sufficient quantities of purified receptor are required and this may be limiting in the initial stages of a project. The limitation of the amount of purified receptor required for SEC may be partially overcome by analyzing complex formation using FSEC (see section on 5HT1BR).

37The GTPase domains of Golf and Gs share 87% sequence identity (80% for the full length α subunits) and both G proteins couple to A2AR. Of the 17 amino acid residues in mini-Gs that make direct contact to residues in A2AR in the crystal structure of the A2AR–mini-Gs complex, all of these residues are identical except that two Arg residues in Gs are replaced with two Lys residues in Golf. Despite the high degree of sequence homology between these two isoforms, Gαolf is far more difficult to overexpress than Gαs, in fact, the only method reported to produce functional Gαolf is co-expression with the molecular chaperone RIC8B in insect cells. Therefore, we constructed mini-Golf to investigate whether the mini-G protein version would be better expressed that native α subunit. Mini-Golf was constructed by transferring the 7 point mutations and 3 deletions from mini-Gs (Fig 2) and mini-Golf was highly expressed in E. coli and as stable as mini-Gs. The coupling of mini-Golf to A2AR was assessed by SEC of the complex assembled in vitro from purified proteins and a thermostability assay. Purified NECA-bound A2AR was mixed with mini-Golf and analysed by SEC and SDS-PAGE (Fig 4A). The apparent molecular weight of mini-Golf was 23 kDa (17.1 ml; theoretical molecular weight 26 kDa) and the apparent molecular weight of purified A2AR in DM was 133 kDa (13.3 ml). The complex A2AR–mini-Golf resolved as a predominant peak with an apparent molecular weight of 153 kDa (13 ml) and contained both A2AR and mini-Golf. Mini-Golf also stabilized agonist-bound DM-solubilised A2AR, with mini-Golf-coupled A2AR showing an apparent Tm of 32 ± 1°C in comparison with 27 ± 0.3°C for the receptor alone (Fig B). This stability was similar to that obtained with mini-Gs (33°C) under the same conditions.

38The results with mini-Golf were very encouraging in terms of both the transferability of the mutations, the expression and stability of the mini-Golf and the stability of the A2AR–mini-Golf complex. Thus where there is a high degree of homology between G proteins, then there is good transferability of the mutations, as was previously observed for the transfer of thermostabilising mutations between GPCRs. These data also suggested that even if the native α subunit is poorly expressed the mini-G protein version may be highly expressed and very stable.

39The expression of mini-Gq in E. coli was unsuccessful. One possibility to explain this is that efficient folding of Gq in vivo is dependent on the molecular chaperone Ric8 and that mini-Gq had a similar requirement. Indeed, co-expression of Ric8 with mini-Gq in the baculovirus expression system led to the overproduction of mini-Gq. However, upon purification of mini-Gq it was not possible to dissociate Ric8 (results not shown), suggesting that the mini-Gq was perhaps not correctly folded and/or was very unstable. Given the lack of success in transferring the mini-G protein mutations from Gs to Gq, another strategy was developed.

40The second strategy used to try and develop mini-Gq was to transfer the specificity determinants of Gq onto mini-Gs. It is well established that the C-terminal region of a Gα subunit forms the main receptor binding site and is one of the main determinants of coupling specificity. Mutating as few as 3–5 amino acids at the C-terminus of the G alpha subunit has been shown to switch the specificity of coupling to some GPCRs. However, the two GPCR–G protein structures published to date revealed an extensive interface between the receptors and Gα, suggesting that other regions of the G protein may also play a role in specificity. Recent in vivo FRET studies suggest that residues within the α5 helix, but distal to the five C-terminal residues, strongly influence specificity.

41Mini-Gs did not couple to any of the Gq-coupled receptors tested (results not shown). We then evaluated a number of mini-Gs/q chimeras (S4 Fig) for both gain of binding to Gq-coupled receptors (Fig 5A and 5B) and loss of binding to the cognate Gs-coupled receptor A2AR, predominantly using thermostability assays (Fig 5C) and SEC (S5 Fig). First, the chimera mini-Gs/q57 was constructed in which the five C-terminal amino acids of mini-Gs (QH5.22YELLH5.26) were changed to those found in Gαq, which required three mutations (Q390EH5.22, E392NH5.24 and L394VH5.26). There was no increase in thermostability of agonist-bound AT1R or NTSR1 in the presence of mini-Gs/q57 that might suggest coupling (Fig 5A and 5B). In fact, there was a decrease in thermostability of AT1R in the presence of mini-Gs/q57, suggesting that binding had occurred, but that residues of Gs were clashing with the receptor and causing its destabilisation. Furthermore, a complex between mini-Gs/q57 and A2AR was still observed (Fig 5C and S5 Fig), suggesting that the mutations were insufficient to change the specificity of Gs to Gq. Therefore, the chimera mini-Gs/q58 was constructed in which the final 19 amino acid residues in the α5 helix of mini-Gs (Phe376H5.8—Leu394H5.26) were changed to those in Gαq; this required 13 mutations (N377AH5.9, D378AH5.10, C379VH5.11, R380KH5.12, I382TH5.14, Q384LH5.16, R385QH5.17, M386LH5.18, H387NH5.19, R389KH5.21, Q390EH5.22, E392NH5.24 and L394VH5.26). Mini-Gs/q58 did not couple to A2AR (Fig 5C and S5 Fig), demonstrating that residues in the α5 helix beyond the C-terminal 5 amino acids are important in G protein specificity. However, there was no significant shift in the thermostability of the Gq-coupled receptor NTSR1 in the presence of mini-Gs/q58 (Fig 5B). We reasoned that this may be because the stability of mini-Gs/q58 was impaired, because mutating the last 19 amino acid residues in mini-Gs would have also changed residues buried in the core of the G protein, thus affecting the stability of the mini-Gs backbone. Therefore, a refined version of this chimera, mini-Gs/q70, was constructed in which residues in the α5 helix whose side chains formed direct contacts (3.9 Å cut-off) with either β2AR or A2AR in the G protein-bound structures were mutated to match those in Gαq (R380KH5.12, Q384LH5.16, R385QH5.17, H387NH5.19, E392NH5.24 and L394VH5.26; S4 Fig). In addition, the mutation Q390EH5.22 was included, despite only making contact to A2AR via its backbone, as it is buried in the receptor–G protein interface and may be important for binding to Gq-coupled receptors. Mini-Gs/q70 gave better binding to both Gq-coupled receptors tested, NTSR1 and AT1R, and showed no binding to A2AR (Fig 5 and S5 Fig).

42Two other chimeras were also constructed to try and improve on mini-Gs/q70. Mini-Gs/q72 contained the additional mutation C379VH5.11 compared to mini-Gs/q70 and, although the C379H5.11 side chain does not form direct contacts with either A2AR or β2AR, its mutation to Val is predicted to introduce a direct interaction between the Val γ2 carbon and Leu110 from A2AR. However, the AT1R–mini-Gs/q72 complex did not have a higher thermostability than AT1R–mini-Gs/q70 (results not shown). Finally, the chimera mini-Gs/q71 was constructed in which residues from other regions of Gα that form direct contacts with either β2AR or A2AR were mutated to match those in Gαq. This included the seven mutations in mini-Gs/q70 (R380KH5.12, Q384LH5.16, R385QH5.17, H387NH5.19, Q390EH5.22, E392NH5.24 and L394VH5.26) and six additional mutations (A39RHNS1.3, H41LS1.2, D343KH4.23, L346VH4.26, R347DH4.27 and Y358I H4S6.11). D343H4.23 was the only amino acid residue whose side chain did not interact with either A2AR or β2AR, but the mutation to Lys was included because the longer side chain could potentially interact with a receptor and the charge reversal may be important for specificity. Conversely, Thr350H4S6.3 was not mutated to Pro in mini-Gs/q71 even though its side chain forms direct contacts with β2AR. Alignment of Gαs with two independently solved structures of Gαq showed that this region of the G proteins differ significantly and thus, in Gαq, this residue is unlikely to interact with the receptor. However, after all these considerations to make an improved version of mini-Gs/q70, mini-Gs/q71 did not improve the thermostability of agonist-bound Gq-coupled receptors compared to mini-Gs/q70 (Fig 5A and 5B).

43Transfer of the 7 point mutations and 3 deletions from mini-Gs into Gαi1 to make mini-Gi1 was not successful, as the resultant protein was very poorly expressed and had low stability (results not shown). Whilst the work on developing chimeras of mini-Gs/q was underway, we decided to first study the reasons why mini-Gi1 appeared to be so unstable. Therefore, to improve expression, stability and to allow binding of the mini-Gi1 to the βγ subunits, the N-terminus (residues 4–18) was re-inserted, Asp249H3.8 was mutated back to Leu, and the G217DH2S4.3 mutation introduced based on a sequence comparison between Gi1 (poorly expressed) and Gs/Go (highly expressed) (Fig 2 and S1 Fig). The resultant mini-Gi1 (construct 46) yielded only 12 mg of purified protein per litre of culture and was 17°C less stable than mini-Gs, but was suitable for initial studies in GPCR coupling.

44The serotonin 5HT1B receptor (5HT1BR) was used as a model Gi-coupled receptor for developing mini-Gi1 because it could be expressed and purified in DDM using the baculovirus expression system and its structure determined in the inactive state. Initially, GFP-mini-Gi1 was tested using FSEC for binding to purified 5HT1BR (in DDM) and bound to the agonist donitriptan. However, the GFP-mini-Gi1 (S3 Fig) migrated at 13.5 ml in the absence of receptor or in the presence of donitriptan-bound 5HT1BR, indicating that no coupling occurred (Fig 6C). However, when the LMNG-purified 5HT1BR was used, the FSEC showed two peaks, one corresponding to free GFP-mini-Gi1 with a retention volume of 14.3 ml and the other corresponding to GFP-mini-Gi1 bound to donitriptan-activated 5HT1BR, with a retention volume of 12.2 ml (Fig 6D). The presence of two peaks, suggested that mini-Gi1 was not stable in detergent when in a complex with 5HT1BR. The detergent instability of the complex could also have been due to the receptor, but because agonist-bound 5HT1BR had been crystallised previously, we felt this was unlikely. In order to improve the stability of mini-Gi1, the heterotrimer was formed between GFP-mini-Gi146 (S3 Fig) and β1γ2 and tested by FSEC. The GFP-mini-trimer in complex with the LMNG-purified 5HT1BR resolved as a single peak with a retention volume of 11.8 ml compared to 14.3 ml for the free GFP-mini-Gi1β1γ2 trimer (Fig 6F). Thus the β1γ2 subunits restored the stability of mini-Gi1.

45Although the mini-Gi1β1γ2 trimer coupled successfully to LMNG-solubilised 5HT1BR, this is not as desirable for crystallography as a mini-G protein coupled receptor due to the large size of the heterotrimeric G protein. Therefore, following the successful strategy of changing the coupling of mini-Gs to that of Gq by making a mini-Gs/q chimera, the same strategy was applied to engineer a mini-Gs/i1 chimera (S1 Fig and S6 Fig). Therefore 9 mutations (C379VH5.11, R380TH5.12, Q384IH5.16, R385KH5.17, H387NH5.19, Q390DH5.22, Y391CH5.23, E392GH5.24 and L394FH5.26) were introduced into the α5 helix of mini-Gs to change its coupling specificity to that of Gi1. A complex between GFP-mini-Gs/i143 (S3 Fig) with donitriptan-bound DDM-purified 5HT1BR resolved as a single peak with a retention volume of 13.2 ml compared to 15.1 ml for the free GFP-mini-Gs/i1 (Fig 6E). Thus mini-Gs/i1 was indeed more stable than mini-Gi1. The specificity of mini-Gs compared to mini-Gs/i1 for donitriptan-bound, DDM-solubilised 5HT1BR was confirmed using FSBA (Fig 6B). No specific coupling of GFP-mini-Gs to 5HT1BR was observed, although specific coupling to GFP-mini-Gs/i1 (apparent KD 390 nM; Fig 6B) was confirmed.

46In order to compare all of the mini-Gi1 constructs and the role of β1γ2, agonist affinity shift assays were performed on 5HT1BR. The uncoupled receptor showed a Ki for the agonist sumatriptan in this assay of 280 ± 10 nM, which was shifted by mini-Gi146 and mini-Gs/i143 to 80 ± 13 nM and 36 ± 2 nM, respectively (Fig 6A). However, addition of β1γ2 to the mini-G proteins resulted in a further increase in agonist affinity to 15 ± 1 nM and 7.2 ± 0.8 nM for mini-Gi146-β1γ2 and mini-Gs/i143-β1γ2, respectively. Thus despite the successful generation of both mini-Gi1 and mini-Gs/i1, their stability is still not perfect as binding of β1γ2 stabilises the mini-G proteins and elicits a greater increase in agonist affinity upon coupling of the mini-trimers.

47The GTPase domain of Go1 and Gi1 are highly conserved (80% identity), but the mini-G proteins derived from them behaved very differently. Unlike the unstable mini-Gi1, mini-Go1 expressed well (100 mg/L), had high stability comparable to mini-Gs and it was largely insensitive to the presence of mild detergents. Since 5HT1BR couples to both Go and Gi family members, we tested mini-Go1 coupling to 5HT1BR and compared the results to coupling with mini-Gi1 (see above). On FSEC, GFP-mini-Go112 (S3 Fig) partially coupled to donitriptan-bound, DDM-solubilised 5HT1BR (unpurified), with the higher molecular weight species (retention volume 13 ml) reduced when the receptor was bound to an antagonist (Fig 7C). This was in contrast to the results with mini-Gi1 under the same conditions where no binding was observed (Fig 6C). The partial coupling probably resulted from the low amount of 5HT1BR used in the assay, because when the experiment was repeated using purified 5HT1BR and GFP-mini-Go1, all of the GFP-mini-Go1 bound to the receptor (Fig 7E). This is consistent with what was observed for β1AR and A2AR (Fig 3). The 5HT1BR–mini-Go1 complex was purified by SEC and SDS-PAGE indicated co-elution of 5HT1BR and mini-Go1 in a 1:1 molar ratio (Fig 7D). GFP-mini-Go1 bound to DDM-solubilised 5HT1BR in the presence of donitriptan with an apparent KD of 180 ± 24 nM (Fig 7B). In membranes, mini-Go112 shifted the agonist affinity for 5HT1BR from 280 ± 10 nM to 32 ± 3 nM (Fig 7A).

48The properties of mini-Go1 make this an ideal choice for structural studies of Go/Gi coupled receptors, rather than using mini-Gs/i, as it is more highly expressed and more tolerant of detergents. This is exemplified by the stability of the donitriptan-bound 5HT1BR–mini-Go1 complex made from purified components (S7 Fig).

49The aim of the work presented here was to generate a range of mini-G proteins that could be used as a basis for the structure determination of GPCRs in their fully active state. The original work in developing mini-G proteins was performed on Gs, which turned out to be one of the best expressed and most stable of the mini-G proteins. The structures of GPCRs and G proteins are highly conserved, and there are thought to be highly conserved networks of side chain interactions within the GPCR and G protein that are essential for receptor activation and G protein activation. It therefore seemed reasonable to use the two Gs-coupled GPCR structures to guide the engineering of Gi, Go or Gq, given that there are no structures of receptors coupled to these G proteins. Transfer of the relevant mutations from mini-Gs to other G proteins was successful in deriving mini-Golf, mini-Go1 and mini-G12. Both mini-Golf and mini-Go1 coupled to relevant receptors only in the presence of an agonist and formed stable complexes that could be purified by SEC. Currently, we have not been able to demonstrate binding of mini-G12 to any receptor (results not shown), even though it is highly expressed in E. coli and has high thermal stability, suggesting that the protein is in a folded state. In contrast, initial trials to generate mini-Gt1, mini-Gz mini-Gq and mini-G16 were unsuccessful due to no expression in E. coli; we have not yet tested the chimera strategy on these targets. Mini-Gi1 expressed very poorly, but was improved upon further mutagenesis, but was still not as stable as mini-Gs and required binding of βγ subunits to attain a full agonist affinity shift in the 5HT1BR.

50The second approach to generate mini-G proteins for those that did not work initially was to make chimeras by converting the specificity of mini-Gs to the specificity of the desired G protein. This was developed initially for Gq by mutating in mini-Gs only those residues in the α5 helix whose side chains make contact to either β2AR or A2AR in the crystal structures of the relevant complexes, to match the equivalent residues in Gq. The final mini-Gs/q chimera was stable, overexpressed in E. coli and coupled to Gq-coupled receptors, but not to Gs-coupled receptors. The process was also successful in generating a mini-Gs/i1 and mini-Gs/o chimera. The α5 helix provides ~70% of the buried surface area between the GTPase domain and the receptor in the two G protein–GPCR complexes crystallised to date. The work here shows that changing these contacts is sufficient to alter the specificity of coupling. However, this is not to say that the remaining 30% of the interface is not important, merely that a range of amino acid residues can be accommodated in this interface and therefore it plays a less important role in defining both specificity and the affinity of G protein binding.

51The mini-G proteins and their properties are shown in Table 1 and Table 2. On the whole, the expression levels are satisfactory in E. coli and the stability of the mini-G proteins in the absence of detergent is also good. However, their stability decreases in detergent, particularly in high concentrations, with the greatest decrease in stability observed at high detergent concentrations (greater than 0.5% w/v) and with detergents that are regarded as harsh for membrane protein purification. Thus care must be exercised in the initial choice of detergent for forming receptor–mini-G protein complexes. However, GPCRs in the active state are also unstable in detergent and this also needs to be carefully addressed before structural studies are a possibility. This may require the use of very high-affinity agonists that stabilize the receptor and/or the stabilisation of the GPCR–mini-G protein complex by conformational thermostabilisation. It is not trivial to define which component, receptor or mini-G protein, is the least stable within a complex under a given set of conditions. This is why we recommend that initial tests are performed on the receptor solubilized in very mild detergents such as digitonin, GDN or LMNG.

52Despite the successes described here, mini-G proteins are not a panacea for all GPCRs. The specificity determinants for G protein coupling are not determined solely by the α5 helix and the subtypes of βγ can also affect the efficiency and specificity of coupling within a cell. As we have not taken any subtype-specific G protein-GPCR interactions into account, further engineering of the mini-G proteins in the light of new structural data of GPCRs coupled to Gi, Go or Gq may improve their performance further. Additional complications to making stable receptor–G protein complexes are the kinetics of dissociation of the G protein from the activated receptor and the stability of the activated receptor itself. There has probably been strong evolutionary pressure for some GPCRs to evolve transient signaling complexes that are inherently unstable to provide tight regulation of the signaling process. Thus it may be necessary to thermostabilise the receptor in the active conformation to generate a complex sufficiently stable for structural studies, although for other GPCR–G protein complexes this may not be necessary.