No statistical methods were used to predetermine sample size. The experiments were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment.
To facilitate expression and purification, human GABAB with the haemagglutinin (HA) signal peptide—including GB1a (UniProt: Q9UBS5) and GB2 (UniProt: O75899)—were cloned into the pEG BacMam vector34. An 8× histidine tag and 3C protease cleavage site were inserted at the C terminus of the GB1a (residues 15–919) subunit, and a Flag epitope tag (DYKDDDD) and a 2× GSG linker were added to the N terminus of the GB2 (residues 42–819) subunit. GABAB and Gil mutants were generated using site-directed mutagenesis. All the constructs were confirmed by sequencing.
Expression and purification of scFv16
scFv16 was expressed and purified as previously described35. In brief, the 6×histidine-tagged scFv16 was expressed in secreted form in Trichoplusia ni Hi5 insect cells for 48 h using the Bac-to-Bac system. The expressed scFv16 was purified using a Ni-NTA resin. The C-terminal 6×His tag of the Ni-NTA eluent was cleaved by 3C protease and further purified by gel filtration chromatography using a Superdex 200 column. Finally, the purified scFv16 was concentrated and stored at −80 °C until further use.
Expression and purification of heterotrimeric Gi1
Heterotrimeric Gi1 was expressed and purified as previously described35. In brief, the dominant-negative Gαi1 (S47N, G203A, E245A and A326S) and human β1γ2 subunits (β1–8×His tag) were co-expressed in Hi5 insect cells for 48 h using the Bac-to-Bac system. The cells were collected and lysed with a buffer containing 10 mM HEPES (pH 7.5), 100 μM MgCl2 and 10 μM GDP. The cell membrane was collected by centrifugation and heterotrimeric Gi1 was extracted in a buffer containing 1% sodium cholate. The supernatant was purified by Ni-NTA column and the detergent was exchanged with n-dodecyl-β-d-maltoside (Anatrace) on a column. Afterward, Gi1 was mixed with a 1.2 molar excess of scFv16 and further purified by Superdex 200 column. Finally, the Gi1–scFv16 complex was concentrated and flash-frozen in liquid nitrogen until further use.
Formation of the GABAB–Gi1–scFv16 complex
The GB1 and GB2 plasmids mixed with PEI 25 K at a 3:0.5:0.5 ratio of PEI to GB1 and GB2 plasmid (w/w) were added to HEK293F cells when the density reached about 2.8 million per ml. Seventeen hours after infection, sodium butyrate was added at a final concentration of 10 mM and the cells were grown for another 3 days at 30 °C before being collected11. The collected cells were solubilized for 3 h at 4 °C in a buffer containing 0.5% (w/v) lauryl maltose neopentyl glycol (Anatrace) and 0.1% (w/v) cholesteryl hemisuccinate (Anatrace). After centrifugation at 30,000g for 30 min, the GABAB was purified by Ni-NTA column and M1 anti-Flag affinity resin. The GABAB was further concentrated and mixed with a 1.3 molar excess of Gi1–scFv16 complex in the presence of 100 μM baclofen and 50 μM BHFF. The sample was incubated at 25 °C for 1 h, followed by the addition of 0.2 U ml−1 apyrase for an additional 1.5-h incubation at 24 °C36. Finally, the sample was purified using a Superose 6 Increase column (GE Healthcare) to acquire a homogeneous GABAB receptor–Gi1 complex. The entire purification procedure was accomplished in 12 h, followed by immediate verification to acquire a stable and fresh sample for structural determination.
Cryo-EM grid preparation and data collection
To prepare cryo-EM grids, 3.0 μl of the purified baclofen- and BHFF-activated GABAB–Gi1 complex at 1.8 mg ml−1 was applied onto the glow-discharged holey carbon grids (Quantifoil, R1.2/1.3, 300 mesh). The grids were blotted for 3.0 s with a blot force of 3 at 4 °C, 100% humidity, and then plunge-frozen in liquid ethane using Vitrobot Mark IV (Thermo Fischer Scientific). Cryo-EM data collection was performed on a Titan Krios at 300 kV accelerating voltage in the Center of Cryo-Electron Microscopy (Zhejiang University). Micrographs were recorded using a Gatan K2 Summit Detector in counting mode with a pixel size of 1.014 Å using SerialEM software37. Image stacks were obtained at a dose rate of about 8.0 electrons per Å2 per second with a defocus ranging from −1.0 to −2.5 μm. The total exposure time was 8 s, and 40 frames were recorded per micrograph. A total of 13,843 movies were collected for the GABAB–Gi1 complex.
Cryo-EM data processing
Image stacks for the GABAB–Gi1 complex were subjected to beam-induced motion correction using MotionCor238. Contrast transfer function parameters for non-dose-weighted micrographs were determined by Gctf39. Cryo-EM data processing was performed using Relion 3.140 and CryoSPARC 2.1541. Template-based particle selection yielded 5,889,932 particle projections using Relion. The projections were imported to CryoSPARC for 2D classification to discard poorly defined particles. The selected particle projections were further subjected to ab initio reconstruction and heterogeneous refinement in CryoSPARC. The well-defined subsets accounting for 1,366,533 particles were re-extracted for further processing in Relion. Three-dimensional classification showed that Gi1 predominantly bound to GB2, however, a small subset (112,338 particles) was also found to interact with GB1. To sort out conformational uniform particles for 3D reconstruction, these projections were subjected to 3D classification with a mask on the TMD–Gil, producing one good subset that accounted for 362,826 particles. Further 3D classifications focusing the alignment on the Gil produced two good subsets, which accounted for 275,089 particles that were subsequently subjected to 3D refinement, contrast transfer function refinement and Bayesian polishing. The overall refinement of GABAB–Gi1 generated a map with an indicated global resolution of 3.5 Å at a Fourier shell correlation of 0.143. To further improve the map quality of the complex (especially for Gi1), local 3D reconstruction focusing on the GABAB receptor and Gi1 was performed using the partial signal subtracted particles in Relion. The local refinement maps for the GABAB and Gi1 showed a global resolution of 3.3 Å and 3.4 Å, respectively, which were combined on the basis of the global refinement map using ‘vop maximum’ command in UCSF Chimera42. This composite map of the GABAB–Gi1 complex was used for subsequent model building and analysis. Global and local resolution was determined using the Bsoft 2.0.7 package43 with half maps as input maps.
Model building and refinement
The model of the active GABAB (PDB 7C7Q)11 was used to generate the initial template of the GABAB. The atomic coordinates of Gil and scFv16 from the structure of the human cannabinoid receptor 2–Gi1 complex (PDB 6PT0)36 were used to generate the initial template of the Gi1–scFv16 complex. Models of GABAB and Gil–scFv16 were docked into the electron microscopy density map using UCSF Chimera42. Agonist and PAM coordinates and geometry restraints were generated using a phenix.elbow44. The docked model was subjected to flexible fitting using Rosetta45 and was further rebuilt in Coot45 and real-space-refined in Rosetta45 and Phenix44. The final refinement statistics were validated using the module ‘comprehensive validation (cryo-EM)’ in Phenix. The goodness-of-fit of the model to the map was determined using a global model-versus-map Fourier shell correlation. The refinement statistics are provided in Supplementary Information and Extended Data Table 1. Structural figures were created using UCSF Chimera42 and the UCSF Chimera X package46.
Enzyme-linked immunosorbent assay
The cell-surface expression of the receptor subunits was detected using an enzyme-linked immunosorbent assay (ELISA). In brief, HEK293T cells were plated in each well of a 6-well plate at a concentration of 0.3 million per ml (2 ml per well). Plasmid transfection was performed with a mixture of 200 ng Gαi1–lgbit, 500 ng Gγ–smbit, 500 ng Gβ, 200 ng GABAB wild type (HA–GB1 and Flag–GB2) or mutants using Lipofectamine 2000 (Thermo Fisher Scientific) in 200 μl of Opti-MEM (Thermo Fisher Scientific). The Flag- and HA-tagged subunits were cotransfected into HEK293T cells and plated in a 96-well plate with a white transparent bottom. HEK293T cells were fixed with 4% paraformaldehyde and blocked with 10% fetal bovine serum (FBS). Bound antibodies coupled to horseradish peroxidase were detected by luminescence using SuperSignal ELISA Femto Maximum Sensitivity substrate (ThermoFisher Scientific), and luminescence was measured using a luminescence microplate reader (Tecan).
IP1 accumulation assay
IP1 accumulation was measured using the IP-One HTRF kit (PerkinElmer, CisBio Bioassays). Transfected HEK293 cells were seeded in a 96-well plate, and 24 h after transfection, cells were treated with baclofen diluted in stimulation buffer in a Cisbio kit for 30 min at 37 °C. Then, cryptate-labelled anti-IP1 monoclonal antibody and d2-labelled IP1 in lysis buffer were added to the wells. After 1 h of incubation at room temperature, the plates were read in PHERAstar FS with excitation at 337 nm and emission at 620 and 665 nm. The accumulation of IP1 was calculated according to a standard dose–response curve.
NanoBiT-G-protein dissociation assay
G-protein activation was detected using a Nanobit-G protein dissociation assay47. The transfection system was the same as that used in the ELISA. After 1 day of transfection, cells in the 6-well plate were digested and resuspended in complete medium DMEM (5% FBS, 1% antibiotic) and plated in 96-well flat-bottomed white microplates. After 24 h, the cells were washed twice with D-PBS and incubated in 40 μl of 5 μM coelenterazine H (Promega) solution diluted with 0.01% BSA- and 5 mM HEPES (pH 7.4)-containing HBSS (assay buffer) for 2 h at room temperature. Baseline luminescence was measured using a luminescent microplate reader (Tecan). The test compound (5×, diluted in the assay buffer) was added to the cells (10 μl) and incubated for 3–5 min at room temperature before the second measurement. The ligand-induced signal ratio was normalized to the baseline luminescence, and fold-change signals over vehicle treatment were used to show the G-protein dissociation response.
Statistical analyses were performed on at least three individual datasets and analysed using GraphPad Prism software. Bars represent differences in the calculated agonist potency (pEC50), maximum agonist response (Emax) and basal activity for each mutant relative to the wild-type receptor. Data are mean ± s.e.m. from at least three independent experiments, performed in triplicates. ND, not determined. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 (one-way analysis of variance (ANOVA) followed by Dunnett’s test, compared with the response of the wild type). For dose–response experiments, data were normalized and analysed using nonlinear curve fitting for the log (agonist) versus response (three parameters) curves.
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.