Plasmid constructs
The human glycine receptor α1 (NCBI: NP_001139512.1) and β (NCBI: NP_000815.1) sequence were amplified from cDNA clones (McDermott Center, UT Southwestern Medical center). The α1em sequence was derived by substitution of M3/M4 loop (residues R316-P381) s by GSSG peptide. For the βem construct, we used the previously described βem construct30. The α1em and α1 wild type sequence was subcloned into a BacMam expression vector47. The β wild type sequence were introduced into pLVX-IRES-ZsGreen1 vector (Clonetech) for electrophysiology. All α1em and βem mutants was generated using site-directed mutagenesis.
Protein expression and purification
The α1em and βem constructs were transformed into DH10BacY competent cells (Geneva Biotech) to produce bacmids. The bacmids were transfected into Sf9 cells to generate baculovirus. Recombinant baculovirus titer was determined as described before30,47. Virus was added at a multiplicity of infection (MOI) of 2(at 3βem:1α1em ratio) to the cell cultures, at a density of 2.5×106 cells/ml. To increase the expression level, 10 mM sodium butyrate was added, and culture temperature was changed to 30˚C after transduction 12h. Cells were collected after induction 60h by centrifugation at 30,000 g for 20 minutes at 4°C and stored at −80 °C until further use.
Cell pellets were thawed and resuspended in a by lysis buffer (40 mM Tris pH 8.0, 50 mM NaCl, 2mM MgCl2, 1mM CaCl2, 20μg/ml Dnase, 2μg/ml leupeptin, 2μM pepstatin, 0.8μM aprotinin, 0.2 mM PMSF) rotated at 4°C for 30min under constant stirring, followed by centrifugation at 40,000g for 20min to collect cell debris. The cell debris was dounced and centrifugated at 40,000g at 4 °C for 20 min. The pellets were further homogenized and solubilized with buffer A (40 mM Tris pH 8.0, 200 mM NaCl, 2mM MgCl2, 1mM CaCl2, 20μg/ml Dnase, 2μg/ml leupeptin, 2μM pepstatin, 0.8μM aprotinin, 0.2 mM PMSF, 0.75%(w/v) DDM, 0.075%(w/v) CHS and 0.075% (w/v) Na Cholate) for 40min at 4°C. Solubilized membranes were cleared by centrifugation at 40,000g for 30 min. Afterwards, supernatant was added to PA-tag antibody (NZ-1) 48 resin at RT. The resin was collected by a gravity column and washed with 5CV buffer B (20 mM Tris pH 8.0, 200 mM NaCl, 2mM MgCl2, 1mM CaCl2, 0.2 mM PMSF, 0.05% (w/v) DDM (Anatrace), 0.005% (w/v) CHS (Anatrace), 0.001% (w/v) Na Cholate (Anatrace)) and 5CV buffer C (20 mM Tris pH 8.0, 200 mM NaCl, 2mM MgCl2, 1mM CaCl2, 0.06% (w/v) digitonin (Sigma-Aldrich)). Then, beads were mixed with PreScission protease (1:30 v/v) to cleave PA tag at RT for 1h. The flow through was collected, and resin were washed with 2CV buffer C. All proteins were pooled and concentrated to load onto Superose6 increase 10/300 GL column (GE Healthcare) in SEC buffer (20 mM Tris pH8.0, 200 mM NaCl, 0.06%(w/v) digitonin). Peak fractions were collected and concentrated to 6 mg/ml for grids freeze.
Cryo-EM sample preparation, data collection and image processing
For apo α1emβem GlyR, the sample was vitrified without any ligand. For glycine-bound α1β GlyR, the sample was incubated for 1h with 2 mM glycine on ice. 1 × CMC final concentration (~ 3 mM) of Fluorinated fos-choline 8 (Anatrace) was added into sample immediately before freezing. Grids (Quantifoil R1.2/1.3 400-mesh Au holey carbon grid) were glow-discharged. An FEI Vitrobot Mark IV Vitrobot (Thermo Fisher) was employed to plunge freeze the grids after application of 3 µl sample at 4℃ under 100% humidity.
Micrographs were collected using a Titan Krios microscope (Thermo Fisher) with a K3 Summit direct electron detector (Gatan) operating at 300 kV using the SerialEM data acquisition software. The GIF-Quantum energy filter was set to a slit width of 20 eV. Images were recorded in the super-resolution counting mode with the pixel size of 0.415 Å. Micrographs were dose-fractioned into 50 frames with a dose rate of 1.4 e-/Å/frame.
2-fold binning (0.83 Å pixel size after binning), motion correction and dose weighting of the movie frames were performed using the Motioncorr2 program49. CTF correction was carried out using the CTFFIND 4 program50. The following image processing steps were carried out in RELION 351, as illustrated in Supplementary Fig. 2. Particles were initially picked using the Laplacian-of-Gaussian blobs and subjected to 2D classification to obtain good class-averages, which was then used as template for reference-based autopicking. Resulting particles were extracted with 4-fold binning for a further round of 2D classification (Supplementary Fig. 2a, i). Good 2D classes were selected and subjected to 3D classification using an initial model downloaded from EMDB database (EMD-23148)30. For both the apo- and glycine-bound samples, 1 out of 6 classes in 3D classification appeared with good density for the entire channel (Supplementary Fig. 2b, j). A single density blob for GFP was identified for both the apo and glycine-bound samples. A further 3D classification into 4 classes with non-binned particles (0.83 Å pixel size) without particle alignment was performed. For the apo- sample, partial signal subtraction52 was performed to focus on the TMD. 2 indistinguishable good classes were pooled, which resulted in a final of 29, 850 particles (Supplementary Fig. 2c). After reverting particles to un-subtracted version, CTF refinement, Bayesian polishing in RELION and non-uniform refinement53 in cryoSPARC54, an overall resolution of 3.6 Å was achieved, with local resolutions exceeding 3.0 Å in many regions (Supplementary Fig. 2d, f). For the glycine-bound sample, second 3D classification was performed using a mask excluding GFP and micelle, resulting in three good classes with distinct conformations (Supplementary Fig. 2k). After CTF refinement, Bayesian polishing in RELION and non-uniform refinement in cryoSPARC, overall resolutions of 3.6 Å (21, 676 particles), 3.9 Å (24, 487 particles) and 4.1 Å (30, 723 particles) were achieved for the open, expanded-open and desensitized states, with local resolutions exceeding 3.0 Å in many regions (Supplementary Fig. 2l, n). Resolutions were estimated by applying a soft mask around the protein densities with the Fourier Shell Correlation (FCS) 0.143 criterion. Local resolutions were calculated using Resmap55.
Model building and refinement
Models of GlyR α1β heteromer were bulit by fitting the structure of Rattus norvegicus α1β homomer glycine-bound state (PDB ID: 7mly) 56 into the Cryo-EM density maps of GlyR α1β heteromer using Chimera57 and Coot58. The atomic model was manually adjusted in Coot. The final models were refined with real-space refinement module and validated with comprehensive validation module in PHENIX package59,60. Fourier shell correlation (FSC) curves were calculated between refined atomic model and the work/free half maps as well as the full map to assess the correlation between the model and density map (Supplementary Fig. 2e and m). Statistics of cryo-EM data processing and model refinement are listed in Table S1. Pore radii were calculated using the HOLE program61. Figures were prepared in UCSF Chimera57, ChimeraX62, and PyMOL63.
The final model of apo α1β GlyR contained the α1 and β subunit amino acids except the following: α1 subunit of chain A and chain C (total 367aa, 345aa built, 22aa not built) A1-P7,L314-L315, GSSG linker, E382 and V421- Q429; α1 subunit of chain B (total 367aa, 342aa built, 25aa not built) A1-P7, H311-L315, GSSG linker, E382 and V421- Q429; α1 subunit of chain D (total 367aa, 342aa built, 25aa not built) A1-M8, K312-L315, GSSG linker, E382-E383 and V421- Q429; β subunit (total 444aa, 338aa built, 106aa not built) K1-N32, GSSAAA-EGFP-SGSGSG insertion and V378-P442.
The final model of expanded open α1β GlyR contained amino acids except the following: α1 subunit of chain A (total 367aa, 341aa built, 26aa not built) A1-P7,K312-L315, GSSG linker, E382-E383 and V421- Q429; α1 subunit of chain B (total 367aa, 339aa built, 28aa not built) A1-P7, R309-L315,GSSG linker, E382-E383 and R422- Q429; α1 subunit of chain C (total 367aa, 337aa built, 30aa not built) A1-M8, H311-L315, GSSG linker, E382-K385 and V421- Q429; α1 subunit of chain D (total 367aa, 340aa built , 27aa not built) A1-P7, H311-L315,GSSG linker,E382-E383 and V421-Q429; The model of β subunit for expanded open is the same as apo.
The final model of open α1β GlyR contained amino acids except the following: α1 subunit of chain A (total 367aa, 341aa built, 26aa not built) A1-P7,K312-L315, GSSG linker, E382-E383 and V421- Q429; α1 subunit of chain B (total 367aa, 340aa built, 27aa not built) A1-P7, R309-L315, GSSG linker, E382-R385 and V421- Q429; α1 subunit of chain C (total 367aa, 336aa built, 31aa not built) A1-M8, H311-L315, GSSG linker, E382-K386 and V421- Q429; α1 subunit of chain D (total 367aa, 340aa built, 27aa not built) A1-P7, H311-L315, GSSG linker, E382-E383 and V421- Q429; The model of β subunit for open is the same as apo.
The final model of desensitized state contained the α1 and β subunit amino acids except the following: α1 subunit of chain A (total 367aa, 335aa built, 32aa not built) A1-P7,Q310-L315, GSSG linker, E382-L387 and V421- Q429; α1 subunit of chain B (total 367aa, 336aa built, 29aa not built) A1-P7, Q310-L315,GSSG linker, E382-M384 and V421- Q429; α1 subunit of chain C (total 367aa, 337aa built, 30aa not built) A1-P7, H311-L315, GSSG linker, E382-K386 and V421-Q429; α1 subunit of chain D (total 367aa, 340aa built, 27aa not built) A1-P7, K312-L315, GSSG linker, E382-M384 and V421- Q429; The model of β subunit for desensitized is the same as apo.
Fluorescence-Detection Size-Exclusion Chromatography (FSEC) expression assay
In the FSEC assay, fluorescence was detected using the RF-20Axs fluorescence detector for HPLC (Shimadzu, Japan) (for EGFP, excitation: 480 nm, emission: 512 nm) as EGFP was fused into βem construct. Using 2μl of Lipofectamine 3000 (Thermo Fisher Scientific, US), 1ug of plasmid (at 1α1:3β ratio) was transfected into HEK293T cells for each well of 12 well plate. Cells were incubated in a CO2 incubator (37 °C, 8% CO2) for 48 h after transfection and solubilized with 50μl buffer B for 1 h. After centrifugation (40,000 × g, 30 min), 50 μl of the sample was applied to a Superose 6 Increase 10/300 GL column (GE Healthcare) equilibrated with buffer D (20 mM Tris pH 8.0, 200 mM NaCl,0.025%DDM) for the FSEC assay.
Whole Cell Patch Clamp
The glycine EC50 values were determined on α1β GlyRs expressed in HEK293T cells. Plasmids were transiently transfected using Lipofectamine 3000 reagent (Invitrogen). Total 0.8µg of DNA was transfected at 1α1:3β ratios for each 35 mm dish. Whole-cell recordings were made after 17-24h transfected at room temperature. GFP fluorescence was used to identify the cells expressing the heteromeric α1β GlyRs. The bath solution contained (in mM): 10 HEPES pH 7.4, 10 KCl, 125 NaCl, 2 MgCl2, 1 CaCl2 and 10 glucoses. The pipette solution contained (in mM): 10 HEPES pH 7.4, 150 KCl, 5 NaCl, 2 MgCl2, 1 CaCl2 and 5 EGTA. The resistance of borosilicate glass pipettes between 2∼7 MΩ. For data acquisition, voltage held at -50 mV and a Digidata 1550B digitizer (Molecular Devices) was connected to an Axopatch 200B amplifier (Molecular Devices). Analog signals were filtered at 1 kHz and subsequently sampled at 20 kHz and stored on a computer running pClamp 10.5 software. Data analysis was performed by Origin software (Origin Lab). Hill1 equation was used to fit the dose-response data and derive the EC50 (k) and Hill coefficient (n). For glycine activation, we used \(I={I}_{0}+({I}_{max}-{I}_{0})\frac{{x}^{n}}{{k}^{n}+{x}^{n}}\), where I is current, I0 is the basal current (accounting mostly for leak, very close to 0), Imax is the maximum current and x is glycine concentration. All start point is fixed at 0 during fit. Measurements were from 4–11 cells, average and S.E.M. values were calculated for each data point.