Ligand recognition and allosteric regulation of DRD1-Gs signaling complexes.

RESOURCE AVAILABILITY

Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead contact, Professor Zhenhua Shao ( zhenhuashao@scu.edu.cn ).

Materials availability

The plasmids generated from this study may be obtained directly from the lead contact.

Data and code availability

All data produced or analyzed in this study are included in the main text or the supplementary materials . The cryo-EM density maps and atomic coordinates have been deposited in the Electron Microscopy Data Bank (EMDB) and Protein Data Bank (PDB) under accession numbers EMD-30392 and 7CKW for Fenoldopam-DRD1 complex; EMD-30393 and 7CKX for A77636-DRD1 complex; EMD-30452 and 7CRH for A77636-DRD1 complex; EMD-30394 and 7CKY for PW0464-DRD1 complex; and EMD-30395 and 7CKZ for dopamine/LY3154207-DRD1 complex.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Spodoptera frugiperda ( Sf9 ) cells were purchased from Expression system and were cultured in ESF921 medium (Expression system) at 27°C with 140 rpm. HEK293T cells were obtained from American Type Culture Collection (CRL-11268) and were grown in DEME medium (GIBCO, #11995) supplemented with 10% fetal bovine serum (FBS) (CELL-BOX, #SAG-01U-02), 1% penicillin/streptomycin (GIBCO, #15140122) at 37°C with 5% CO 2 .

METHOD DETAILS Constructs

The human wild-type DRD1 gene was cloned into pFastBac1 with an N-terminal FLAG tag (DYKDDDDA). To facilitate expression and purification, the DRD1 signal peptide was substituted with that of hemagglutinin (HA). Human DNGα s , human Gβ1 with an N-terminal 6 × Histidine tag and human Gγ2 were also sub-cloned into pFastBac1 vector. The same constructs were used for the expression and functional assays in this study. Expression of DRD1, Gα s /Gβγ heterotrimer Recombinant baculovirus was generated using the Bac-to-Bac Baculovirus Expression System (Invitrogen). Briefly, FuGENE HD transfection reagent (Promega) was used to prepare baculovirus. Sf9 suspension cells were seeded in ESF921 medium at a density of 3 × 10 6 cells/ml and infected with DRD1 baculovirus. After 48h incubation at 27°C, shaking at 110 rpm, cells were collected by centrifugation, flash-frozen in liquid nitrogen and stored at −80°C. Gs heterotrimer was co-expressed via infection with virus of Gs and Gβ1γ2. Infected cells were also cultured at 27°C, 110 rpm for 48 h. The harvested cells were collected by centrifugation and the cell pellets were stored at −80°C.

DRD1-Gs complex formation and purification

Sf9 cell pellets infected with DRD1 and Gs heterotrimer were resuspended in lysis buffer (20 mM HEPES, pH 7.4, 100 mM NaCl, 3 mM MgCl 2 , 5 mM CaCl 2 , 2.5 mg/ml leupeptin and 0.2 mg/ml benzamidine). The DRD1-Gs complex was prepared in membranes, and the mixture was incubated for 2 h at room temperature by adding either 10 μM fenoldopam, A77636 and SKF83959 , 10 mg/ml Nb35, and 25 mU/ml Apyrase. The complex was then solubilized in a buffer containing 20 mM HEPES, pH 7.4, 100 mM NaCl, 3 mM MgCl 2 , 5 mM CaCl 2 , 2.5 mg/ml leupeptin, 0.2 mg/ml benzamidine, 0.5% (w/v) lauryl maltose neopentylglycol (LMNG, Anatrace), 0.1% (w/v) cholesteryl hemisuccinate TRIS salt (CHS) for 3 h at 4°C. The supernatant was collected by centrifugation at 25,000 rpm for 30 min, and the solubilized complex was incubated with M1 anti-FLAG resin for 2 h at 4°C. The complex was immobilized on Flag-M1 resin, loaded on a Flag-M1 column and washed with 20 column volumes of 20 mM HEPES, pH 7.4, 100 mM NaCl, 3 mM MgCl 2 , 5 mM CaCl 2 , 0.01% (w/v) LMNG, 0.002% (w/v) CHS, 10 μM ligand, 2.5 mg/ml leupeptin, 0.2 mg/ml benzamidine. The DRD1-Gs complex was eluted with 20 mM HEPES, pH 7.4, 100 mM NaCl, 0.01% (w/v) LMNG, 0.002% (w/v) CHS, 10 μM ligand, 5 mM EGTA and 0.2 mg/ml FLAG peptide. The complex was collected and concentrated, then loaded onto a Superdex 6 Increase 10/300 GL column (GE Healthcare) with buffer containing 20 mM HEPES, pH 7.4, 100 mM NaCl, 0.00075% (w/v) LMNG, 0.00025% GDN, 0.0002% (w/v) CHS, 10 μM ligand. The complex fractions were concentrated with a 100 kDa MWCO Millipore concentrator for electron microscopy. The DRD1-Gs-Nb35 in complex with PW0464, or dopamine and LY3154207 were prepared with similar procedure as fenoldopam bound DRD1 complex. All samples were concentrated over 5mg/ml for making cryo grid.

Cryo-grid preparation and EM data collection

Forcryo-EM grid preparation, 3.5 μL aliquots of concentrated ligands bound DRD1-Gs-Nb35 complexes were loaded onto glow-discharged holey carbon grids (Quantifoil Au R1.2/1.3, 300 mesh). Grids were blotted for 3.0 s and plunge-frozen in liquid ethane cooled by liquid nitrogen using a Vitrobot Mark IV (Thermo Fisher) at 4°C and with 100% humidity. Grids were then transferred to a Titan Krios electron microscope (Thermo Fisher) operating at 300 kV and equipped with spherical aberration (Cs) image corrector. Micrographs were recorded using a K2 Summit counting camera (Gatan Company) in super-resolution mode with a nominal magnification of 22,500 × , resulting in a physical pixel size of 1 Å. Movie stacks were obtained with a defocus range of −1.0 to −2.0 μm ( Table S1 ) using SerialEM with a set of customized scripts enabling automated low-dose image acquisition. Each movie stack was recorded for a total of 8 s with 0.2 s exposure per frame and exposure dose set to 8 electrons per pixel per second. For micrographs recording of SKF83959 bound DRD1 complex, the magnification is 105,000 × and pixel size is 0.85 Å, movie stacks were obtained with a defocus range of −1.0 to −2.5 μm ( Table S1 ).

Image processing and 3D reconstructions

For the fenoldopam-DRD1-Gs-Nb35 complex, a total 2316 movie stacks were collected and all 40 frames in each stack were summed and motion-corrected using MotionCorr2 ( Roh et al., 2017 ) ( Figure S1 ), with twofold binned to a pixel size of 1Å per pixel and dose weighting was performed. Each micrograph was manually inspected to remove bad pictures that were contaminated by crystalline ice or other forms of visible contamination, and CTF parameters were estimated by GCTF ( Zhang, 2016 ). After sorting, micrographs with maximum estimated resolution beyond 4.0 Å were discarded. A total of 2,386,124 particles were auto-picked by Gautomatch ( Zhang, 2016 ) for the fenoldopam-DRD1-Gs-Nb35 sample with a box size of 216 pixels. Particle picking was performed using low-pass filtered templates to 40 Å to limit reference bias. Particles extracted from the dataset were downscaled 4 times and subjected to reference-free 2D classification to remove false picks and obvious junk classes leaving behind approximately 1,581,300 particles. The map generated in cryoSPARC ( Punjani et al., 2017 ) low passed filtered to 60 Å was used as an initial reference model for maximum-likelihood-based three-dimensional classifications. The total good 1,581,300 particles were 3D classified with 6 classes in RELION3.0 ( Zivanov et al., 2018 ). Two good classes (887,222 particles) were selected for another round of 3D classification. A dataset of 539,327 particles were subjected to 3D auto-refinement, resulting in an initial 3.3 Å density map. Further Bayesian polishing of these particle projections was performed in RELION3.0, followed by another round of auto-refinement, which generated a final 3.2 Å map determined by gold standard Fourier shell correlation using the 0.143 criterion. The density maps were corrected for the modulation transfer function (MTF) of the K2 summit direct detector and then sharpened by applying a temperature factor that was estimated using post-processing in RELION3.0. Local resolution estimation was performed in RELION3.0 using the unfiltered half map. For A77636-DRD1-Gs-Nb35 complex, a total 3509 movie stacks were collected ( Data S1A ). 720,719 particles were extracted for further 3D classification and we obtained four different conformational subclasses. 391,771 particle projections of the best class were further applied for final homogeneous refinement in RELION3.0 and obtained the best density map with global l resolution of 3.5 Å. For SKF83959 -DRD1-Gs-Nb35 complex, a total 3610 movie stacks were collected ( Data S1D ). 140,000 particle projections (38.9%) of the best class were further applied for final homogeneous refinement in RELION3.0 and obtained the best density map with global resolution of 3.3 Å. For the PW0464-DRD1-Gs-Nb35 complex, a total of 4,690 movie stacks were collected ( Data S1B ). 720,786 particles were extracted for further 3D classification and we obtained four different conformational subclasses. 672,890 particle projections of the best class were further applied for final homogeneous refinement in RELION3.0 and obtained the best density map with global resolution of 3.1 Å. For dopamine-LY3154207-DRD1-Gs-Nb35 complex ( Data S1C ), a total of 4,475 movie stacks were collected. 1,206,885 particles were extracted for further 3D classification and obtained four different conformational subclasses. 584,166 particle projections of the best class were further applied for final homogeneous refinement in RELION3.0 and obtained the best density map with global resolution of 3.1 Å.

Model building and structure refinement

The crystal structure of β2 receptor with Gs protein and NB35 complex (PDB code: 3SN6) ( Rasmussen et al., 2011 ) was used as initial model for model rebuilding and refinement against EM density map. The model was docked into the EM density map using UCSF Chimera ( Pettersen et al., 2004 ), followed by manual adjustment and rebuilding in COOT ( Emsley and Cowtan, 2004 ). Real space refinement was carried out in Phenix programs ( Adams et al., 2010 ). The model statistics was validated using MolProbity ( Williams et al., 2018 ). The final structures showed a good model geometry, and the detailed refinement statistics are provided in Table S1 . The extent of any model overfitting during refinement was measured by refining the final model against one of the half-maps and by comparing the resulting map versus model FSC curves with the two half-maps and the full model. Structural figures were prepared in PyMoL ( https://pymol.org/2/ ) and UCSF Chimera.

Radioligand binding assay

Binding assays were performed using wild-type or mutant human DRD1 transfected into HEK293 cell. Briefly, wild-type or mutant DRD1 was transfected into HEK293 cells using Lipofectamine 2000 (Thermo Fisher) according to the manufacturer’s instructions, and cultured at 37°C with 5% CO 2 for 48 h. Cells were then harvested with a cell scraper and dounced to homogenize in ice-cold lysis buffer containing 50 mM HEPES, pH 7.4, 2 mM EDTA. Cell membranes were collected by centrifugation at 25,000 g for 45 min and homogenized in ice-cold buffer containing 75 mM Tris pH 7.4, 1 mM EDTA, and 12.5 mM MgCl 2 . The total membrane surface protein expression was determined by Bradford protein assay. For saturation binding assay, membranes (8~10 μg) were incubated for 45 min at 30°C while shaking with 0.075-5 nM [ 3 H]-SCH23390 (Perkin Elmer). Non-specific binding was determined in the presence of 2 μM A77636. K d value of the wild-type DRD1 with [ 3 H]-SCH23390 is 0.28 ± 0.03 nM. For competition assays, membranes (~8 μg) were incubated for 45 min at 30°C shaking with 0.5 nM [ 3 H]-SCH23390 (PerkinElmer) and an increasing concentration of competing ligand concentrations in binding buffer (50 mM HEPES, pH 7.4, 10 mM MgCl 2 , 1 mM EDTA, 0.1% protease-free BSA, 0.01% ascorbic acid). In case of the dopamine competition assay in the presence of an allosteric regulator, different concentrations of LY3154207 (from 0.14 nM to 100 nM) were added. Cells were filtered over GF/C glass microfiber filters (Whatman) pre-soaked in ice-cold binding buffer supplemented 0.5% polyethylenimine using a 96-well filtermate harvester (Perkin Elmer) and rinsed 3 times with 2 mL ice-cold assay buffer to remove any free [ 3 H]-SCH23390. Afterward, filters were dried and subjected to liquid scintillation counting on a MicroBeta TriLux scintillation counter (Perkin Elmer). All data were analyzed by GraphPad Prism 7 (GraphPad Software Inc.). Each measurement was repeated in at least three independent experiments, each in triplicate. cAMP functional assays The coding sequence of wild-type human DRD1 and DRD2 was sub-cloned into the expression vector pcDNA3.1+ with a HA signal sequence followed by a Flag tag at the N terminus. Point mutations used in our study were generated using Q5 site-Directed Mutagenesis kit (NEB). The constructs were expressed in HEK293 cells using jetPRIME transfection reagent (Polyplus transfection, France) according to the manufacturer’s instruction. Cells were harvested 48 h post-transfection. Cell surface expression level of receptors were determined by flow cytometry with an anti-Flag M2-FITC antibody (Sigma, F4049). cAMP accumulation level was measured using a cAMP-Gs kit (Cisbio Bioassays, 62AM4PEB) or GloSensor cAMP assay (Promega) for DRD1/DRD5. In brief, for cAMP accumulation assays of DRD1/DRD5 activated by agonists, the harvested cells were seeded into 384-well plates and stimulated with the corresponding agonists and incubated for 1 h at 37°C with 5% CO 2 . After that, cAMP Eu Cryptate conjugate and anti-cAMP-d2 were added to 384-wells plates and incubated for 1 h at room temperature. Finally, plates were read on BIOTEK Cytation3 reader with excitation at 320 nm, emission at 620 nm and 665 nm. In addition, increase of cAMP level by positive allosteric modulator LY3154207 were measured in the presence and absence of EC 20 concentration of dopamine (10 nM). The allosteric regulation efficacy of LY3154207 on dopamine action on DRD1 was performed using different concentrations of LY3154207 from 0.14 nM to 100 nM. For GloSensor cAMP assay, HEK293 cells transfected with wild-type or mutant DRD1 and Glosensor plasmid were plated into 96-well plates and cultured for 48 h at 37°C with 5% CO 2 . Cells were stimulated with serum-free media containing 5% v/v dilution of the GloSensor cAMP reagent stock solution (Promega). Afterward, a range of concentrations of ligands were added, and the luminescence signal were counted on an EnVision multimode plate reader (Perkin Elmer). cAMP level was calculated according to a standard dose-response curve. EC 50 were calculated using nonlinear regression (curve fit) using GraphPad Prism 7 (GraphPad Software). To validate our results, each measurement was repeated in at least three independent experiments, each in triplicate. To measure cAMP inhibition by D2-like receptors (DRD2, DRD3 and DRD4) activating, cell-based luciferase activity assay was performed by pGL3-CRE-luciferase reporter systems (Promega), as described in a previous study ( Wu et al., 2019 ) . Briefly, cells expressing wild-type and mutant DRD2 were treated with dopamine diluted in 2 μM forskolin stimulating buffer. After 6 h incubation, 1 × Passive Lysis Buffer was added to the 96-well plate to measure luciferase assay in cell lysates as determined by the luciferase assay reagent kit (Promega). The luciferase activities in agonist treatment groups were first calculated as ratios to that of control group and normalized by wild-type receptor. To validate our function results, all experiments were repeated at least three times independently, each in triplicate. β-arrestin recruitment assay To measure β-arrestin recruitment by DRD1 upon PW0464 or A77636 activation, intracellular bioluminescence resonance energy transfer (BRET) assay was performed as previously described ( Yang et al., 2018 , 2020 ). Briefly, Flag-DRD1-WT-eYFP, β-arrestin-2-Rluc plasmids were co-transfected into HEK293 cells using PEI and cultured at 37°C with 5% CO 2 for 24 h. Afterward, the transfected cells were seeded (50,000 cells per well) into clear-bottom 96-well plate and cultured at 37°C for another 24 h. The cells were washed 3 times with 200 μL PBS and incubated with increasing concentrations of PW0464, A77636 or vehicle for 10 min. BRET signal was determined, as the ratio of the light emitted by YFP (530 nm) and Rluc (485 nm), after addition of 5 μM of luciferase substrate coelenterazine 400a (Cayman) using a Mithras LB940 microplate reader (Berthold Technologies). Molecule docking The GLIDE program (version 5.6, Schrödinger, LLC, New York, NY, 2010) ( Friesner et al., 2006 ; Halgren et al., 2004 ) was used for docking studies. Prior to docking, ligands were prepared using “Ligand Prep” module. For GLIDE docking, complex structures of DRD1 was preprocessed according to the protein preparation procedure recommended and docking site was defined in terms of ligands. All docking calculations were run in the “Standard Precision” (SP) mode ( Friesner et al., 2004 ) with default values for all parameters. Five poses per prepared ligand were saved for each docking calculation. The best-docked structure was chosen using the Glide Score ( Jorgensen et al., 1996 ). Molecular dynamics (MD) simulation Prior to running MD simulations, the receptor-ligand complex model was further prepared in Protein Preparation Wizard in Maestro (Schrödinger) to fix the missing side chains, add hydrogen atoms and cap the receptor chain termini while the residues D70 and D120 were protonated in the active-state simulations, all other titratable residues were left in their dominate protonation state at pH 7.0. The structures were inserted into a pre-equilibrated POPC bilayer and solvated with 0.15 M NaCl in explicitly water using dabble ( Betz, 2018 ). Parameters for the MD simulations were generated using the CHARMM-GUI web interface ( Jo et al., 2008 ; Lee et al., 2016 ; Wu et al., 2014 ) using the CHARMM36 forcefield ( Huang and MacKerell, 2013 ) with CGenFF ( Vanommeslaeghe et al., 2010 ) for ligands. Simulations were performed CUDA enabled version of PMEMD in AMBER16 ( Le Grand et al., 2013 ). The system was heated from 0 to 100 K in the NVT ensemble using the Langevin thermostat over 12.5 ps with harmonic restraints of 10.0 kcal mol −1 Å −2 on the non-hydrogen atoms of lipid, protein, and ligand. Then the system was heated to 310 K over 125 ps in the NPT ensemble with semi-isotropic pressure coupling and a pressure of 1 bar. Further equilibration was performed at 310K with harmonic restraints on the protein and ligand starting at 5.0 kcal mol −1 Å −2 and reduced by 1.0 kcal mol −1 Å −2 in a stepwise fashion every 2 ns, for a total of 10 ns of additional restrained equilibration. Multiple independent simulations were initialized from the final snapshot of the restrained equilibration. These simulations were conducted in the NPT ensemble at 310K and 1 bar, using a Langevin thermostat and Monte Carlo barostat under periodic boundary conditions. A time step of 2.0 fs with hydrogen mass repartitioning was used, bond lengths to hydrogen atoms were constrained using SHAKE ( Ryckaert et al., 1977 ). Non-bonded interactions were cut off at 9.0 Å, and long-range electrostatic interactions were computed using the particle mesh Ewald method. The simulation frames were written every 100 ps. Root-mean-square deviation (RMSD) was calculated using the CPPTRAJ module ( Roe and Cheatham, 2013 ) to monitor the fluctuation of the protein and ligand during the simulations.

QUANTIFICATION AND STATISTICAL ANALYSIS

The resolutions of all cryo-EM maps were estimated with the gold-standard Fourier shell correlation 0.143 criterion indicated in figure legends. The detailed can be found in Figure S1 , supplemental information Data S1 , legends and method. Statistical analysis of radioligand binding assays, cAMP function assays, BRET assays and receptor expression level assays were processed using Prism 7.0 (GraphPad). Data in figures and tables are presented as the mean ± standard error of the mean (SEM) of three independent experiments performed in triplicate indicated in figure and table legends. The “n” value in tables means the numbers of independent experiment. The details can be found in method and Figures 5C , 5D , S3A , S3B , S3G , S3I , S4A , S5F – S5H , S6B – S6H , S7E , S7H , and S7I , and Tables S2 – S5 and legends.

Materials availability

The plasmids generated from this study may be obtained directly from the lead contact.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Spodoptera frugiperda ( Sf9 ) cells were purchased from Expression system and were cultured in ESF921 medium (Expression system) at 27°C with 140 rpm. HEK293T cells were obtained from American Type Culture Collection (CRL-11268) and were grown in DEME medium (GIBCO, #11995) supplemented with 10% fetal bovine serum (FBS) (CELL-BOX, #SAG-01U-02), 1% penicillin/streptomycin (GIBCO, #15140122) at 37°C with 5% CO 2 .

METHOD DETAILS Constructs

The human wild-type DRD1 gene was cloned into pFastBac1 with an N-terminal FLAG tag (DYKDDDDA). To facilitate expression and purification, the DRD1 signal peptide was substituted with that of hemagglutinin (HA). Human DNGα s , human Gβ1 with an N-terminal 6 × Histidine tag and human Gγ2 were also sub-cloned into pFastBac1 vector. The same constructs were used for the expression and functional assays in this study. Expression of DRD1, Gα s /Gβγ heterotrimer Recombinant baculovirus was generated using the Bac-to-Bac Baculovirus Expression System (Invitrogen). Briefly, FuGENE HD transfection reagent (Promega) was used to prepare baculovirus. Sf9 suspension cells were seeded in ESF921 medium at a density of 3 × 10 6 cells/ml and infected with DRD1 baculovirus. After 48h incubation at 27°C, shaking at 110 rpm, cells were collected by centrifugation, flash-frozen in liquid nitrogen and stored at −80°C. Gs heterotrimer was co-expressed via infection with virus of Gs and Gβ1γ2. Infected cells were also cultured at 27°C, 110 rpm for 48 h. The harvested cells were collected by centrifugation and the cell pellets were stored at −80°C.

DRD1-Gs complex formation and purification

Sf9 cell pellets infected with DRD1 and Gs heterotrimer were resuspended in lysis buffer (20 mM HEPES, pH 7.4, 100 mM NaCl, 3 mM MgCl 2 , 5 mM CaCl 2 , 2.5 mg/ml leupeptin and 0.2 mg/ml benzamidine). The DRD1-Gs complex was prepared in membranes, and the mixture was incubated for 2 h at room temperature by adding either 10 μM fenoldopam, A77636 and SKF83959 , 10 mg/ml Nb35, and 25 mU/ml Apyrase. The complex was then solubilized in a buffer containing 20 mM HEPES, pH 7.4, 100 mM NaCl, 3 mM MgCl 2 , 5 mM CaCl 2 , 2.5 mg/ml leupeptin, 0.2 mg/ml benzamidine, 0.5% (w/v) lauryl maltose neopentylglycol (LMNG, Anatrace), 0.1% (w/v) cholesteryl hemisuccinate TRIS salt (CHS) for 3 h at 4°C. The supernatant was collected by centrifugation at 25,000 rpm for 30 min, and the solubilized complex was incubated with M1 anti-FLAG resin for 2 h at 4°C. The complex was immobilized on Flag-M1 resin, loaded on a Flag-M1 column and washed with 20 column volumes of 20 mM HEPES, pH 7.4, 100 mM NaCl, 3 mM MgCl 2 , 5 mM CaCl 2 , 0.01% (w/v) LMNG, 0.002% (w/v) CHS, 10 μM ligand, 2.5 mg/ml leupeptin, 0.2 mg/ml benzamidine. The DRD1-Gs complex was eluted with 20 mM HEPES, pH 7.4, 100 mM NaCl, 0.01% (w/v) LMNG, 0.002% (w/v) CHS, 10 μM ligand, 5 mM EGTA and 0.2 mg/ml FLAG peptide. The complex was collected and concentrated, then loaded onto a Superdex 6 Increase 10/300 GL column (GE Healthcare) with buffer containing 20 mM HEPES, pH 7.4, 100 mM NaCl, 0.00075% (w/v) LMNG, 0.00025% GDN, 0.0002% (w/v) CHS, 10 μM ligand. The complex fractions were concentrated with a 100 kDa MWCO Millipore concentrator for electron microscopy. The DRD1-Gs-Nb35 in complex with PW0464, or dopamine and LY3154207 were prepared with similar procedure as fenoldopam bound DRD1 complex. All samples were concentrated over 5mg/ml for making cryo grid.

Cryo-grid preparation and EM data collection

Forcryo-EM grid preparation, 3.5 μL aliquots of concentrated ligands bound DRD1-Gs-Nb35 complexes were loaded onto glow-discharged holey carbon grids (Quantifoil Au R1.2/1.3, 300 mesh). Grids were blotted for 3.0 s and plunge-frozen in liquid ethane cooled by liquid nitrogen using a Vitrobot Mark IV (Thermo Fisher) at 4°C and with 100% humidity. Grids were then transferred to a Titan Krios electron microscope (Thermo Fisher) operating at 300 kV and equipped with spherical aberration (Cs) image corrector. Micrographs were recorded using a K2 Summit counting camera (Gatan Company) in super-resolution mode with a nominal magnification of 22,500 × , resulting in a physical pixel size of 1 Å. Movie stacks were obtained with a defocus range of −1.0 to −2.0 μm ( Table S1 ) using SerialEM with a set of customized scripts enabling automated low-dose image acquisition. Each movie stack was recorded for a total of 8 s with 0.2 s exposure per frame and exposure dose set to 8 electrons per pixel per second. For micrographs recording of SKF83959 bound DRD1 complex, the magnification is 105,000 × and pixel size is 0.85 Å, movie stacks were obtained with a defocus range of −1.0 to −2.5 μm ( Table S1 ).

Image processing and 3D reconstructions

For the fenoldopam-DRD1-Gs-Nb35 complex, a total 2316 movie stacks were collected and all 40 frames in each stack were summed and motion-corrected using MotionCorr2 ( Roh et al., 2017 ) ( Figure S1 ), with twofold binned to a pixel size of 1Å per pixel and dose weighting was performed. Each micrograph was manually inspected to remove bad pictures that were contaminated by crystalline ice or other forms of visible contamination, and CTF parameters were estimated by GCTF ( Zhang, 2016 ). After sorting, micrographs with maximum estimated resolution beyond 4.0 Å were discarded. A total of 2,386,124 particles were auto-picked by Gautomatch ( Zhang, 2016 ) for the fenoldopam-DRD1-Gs-Nb35 sample with a box size of 216 pixels. Particle picking was performed using low-pass filtered templates to 40 Å to limit reference bias. Particles extracted from the dataset were downscaled 4 times and subjected to reference-free 2D classification to remove false picks and obvious junk classes leaving behind approximately 1,581,300 particles. The map generated in cryoSPARC ( Punjani et al., 2017 ) low passed filtered to 60 Å was used as an initial reference model for maximum-likelihood-based three-dimensional classifications. The total good 1,581,300 particles were 3D classified with 6 classes in RELION3.0 ( Zivanov et al., 2018 ). Two good classes (887,222 particles) were selected for another round of 3D classification. A dataset of 539,327 particles were subjected to 3D auto-refinement, resulting in an initial 3.3 Å density map. Further Bayesian polishing of these particle projections was performed in RELION3.0, followed by another round of auto-refinement, which generated a final 3.2 Å map determined by gold standard Fourier shell correlation using the 0.143 criterion. The density maps were corrected for the modulation transfer function (MTF) of the K2 summit direct detector and then sharpened by applying a temperature factor that was estimated using post-processing in RELION3.0. Local resolution estimation was performed in RELION3.0 using the unfiltered half map. For A77636-DRD1-Gs-Nb35 complex, a total 3509 movie stacks were collected ( Data S1A ). 720,719 particles were extracted for further 3D classification and we obtained four different conformational subclasses. 391,771 particle projections of the best class were further applied for final homogeneous refinement in RELION3.0 and obtained the best density map with global l resolution of 3.5 Å. For SKF83959 -DRD1-Gs-Nb35 complex, a total 3610 movie stacks were collected ( Data S1D ). 140,000 particle projections (38.9%) of the best class were further applied for final homogeneous refinement in RELION3.0 and obtained the best density map with global resolution of 3.3 Å. For the PW0464-DRD1-Gs-Nb35 complex, a total of 4,690 movie stacks were collected ( Data S1B ). 720,786 particles were extracted for further 3D classification and we obtained four different conformational subclasses. 672,890 particle projections of the best class were further applied for final homogeneous refinement in RELION3.0 and obtained the best density map with global resolution of 3.1 Å. For dopamine-LY3154207-DRD1-Gs-Nb35 complex ( Data S1C ), a total of 4,475 movie stacks were collected. 1,206,885 particles were extracted for further 3D classification and obtained four different conformational subclasses. 584,166 particle projections of the best class were further applied for final homogeneous refinement in RELION3.0 and obtained the best density map with global resolution of 3.1 Å.

Model building and structure refinement

The crystal structure of β2 receptor with Gs protein and NB35 complex (PDB code: 3SN6) ( Rasmussen et al., 2011 ) was used as initial model for model rebuilding and refinement against EM density map. The model was docked into the EM density map using UCSF Chimera ( Pettersen et al., 2004 ), followed by manual adjustment and rebuilding in COOT ( Emsley and Cowtan, 2004 ). Real space refinement was carried out in Phenix programs ( Adams et al., 2010 ). The model statistics was validated using MolProbity ( Williams et al., 2018 ). The final structures showed a good model geometry, and the detailed refinement statistics are provided in Table S1 . The extent of any model overfitting during refinement was measured by refining the final model against one of the half-maps and by comparing the resulting map versus model FSC curves with the two half-maps and the full model. Structural figures were prepared in PyMoL ( https://pymol.org/2/ ) and UCSF Chimera.

Radioligand binding assay

Binding assays were performed using wild-type or mutant human DRD1 transfected into HEK293 cell. Briefly, wild-type or mutant DRD1 was transfected into HEK293 cells using Lipofectamine 2000 (Thermo Fisher) according to the manufacturer’s instructions, and cultured at 37°C with 5% CO 2 for 48 h. Cells were then harvested with a cell scraper and dounced to homogenize in ice-cold lysis buffer containing 50 mM HEPES, pH 7.4, 2 mM EDTA. Cell membranes were collected by centrifugation at 25,000 g for 45 min and homogenized in ice-cold buffer containing 75 mM Tris pH 7.4, 1 mM EDTA, and 12.5 mM MgCl 2 . The total membrane surface protein expression was determined by Bradford protein assay. For saturation binding assay, membranes (8~10 μg) were incubated for 45 min at 30°C while shaking with 0.075-5 nM [ 3 H]-SCH23390 (Perkin Elmer). Non-specific binding was determined in the presence of 2 μM A77636. K d value of the wild-type DRD1 with [ 3 H]-SCH23390 is 0.28 ± 0.03 nM. For competition assays, membranes (~8 μg) were incubated for 45 min at 30°C shaking with 0.5 nM [ 3 H]-SCH23390 (PerkinElmer) and an increasing concentration of competing ligand concentrations in binding buffer (50 mM HEPES, pH 7.4, 10 mM MgCl 2 , 1 mM EDTA, 0.1% protease-free BSA, 0.01% ascorbic acid). In case of the dopamine competition assay in the presence of an allosteric regulator, different concentrations of LY3154207 (from 0.14 nM to 100 nM) were added. Cells were filtered over GF/C glass microfiber filters (Whatman) pre-soaked in ice-cold binding buffer supplemented 0.5% polyethylenimine using a 96-well filtermate harvester (Perkin Elmer) and rinsed 3 times with 2 mL ice-cold assay buffer to remove any free [ 3 H]-SCH23390. Afterward, filters were dried and subjected to liquid scintillation counting on a MicroBeta TriLux scintillation counter (Perkin Elmer). All data were analyzed by GraphPad Prism 7 (GraphPad Software Inc.). Each measurement was repeated in at least three independent experiments, each in triplicate. cAMP functional assays The coding sequence of wild-type human DRD1 and DRD2 was sub-cloned into the expression vector pcDNA3.1+ with a HA signal sequence followed by a Flag tag at the N terminus. Point mutations used in our study were generated using Q5 site-Directed Mutagenesis kit (NEB). The constructs were expressed in HEK293 cells using jetPRIME transfection reagent (Polyplus transfection, France) according to the manufacturer’s instruction. Cells were harvested 48 h post-transfection. Cell surface expression level of receptors were determined by flow cytometry with an anti-Flag M2-FITC antibody (Sigma, F4049). cAMP accumulation level was measured using a cAMP-Gs kit (Cisbio Bioassays, 62AM4PEB) or GloSensor cAMP assay (Promega) for DRD1/DRD5. In brief, for cAMP accumulation assays of DRD1/DRD5 activated by agonists, the harvested cells were seeded into 384-well plates and stimulated with the corresponding agonists and incubated for 1 h at 37°C with 5% CO 2 . After that, cAMP Eu Cryptate conjugate and anti-cAMP-d2 were added to 384-wells plates and incubated for 1 h at room temperature. Finally, plates were read on BIOTEK Cytation3 reader with excitation at 320 nm, emission at 620 nm and 665 nm. In addition, increase of cAMP level by positive allosteric modulator LY3154207 were measured in the presence and absence of EC 20 concentration of dopamine (10 nM). The allosteric regulation efficacy of LY3154207 on dopamine action on DRD1 was performed using different concentrations of LY3154207 from 0.14 nM to 100 nM. For GloSensor cAMP assay, HEK293 cells transfected with wild-type or mutant DRD1 and Glosensor plasmid were plated into 96-well plates and cultured for 48 h at 37°C with 5% CO 2 . Cells were stimulated with serum-free media containing 5% v/v dilution of the GloSensor cAMP reagent stock solution (Promega). Afterward, a range of concentrations of ligands were added, and the luminescence signal were counted on an EnVision multimode plate reader (Perkin Elmer). cAMP level was calculated according to a standard dose-response curve. EC 50 were calculated using nonlinear regression (curve fit) using GraphPad Prism 7 (GraphPad Software). To validate our results, each measurement was repeated in at least three independent experiments, each in triplicate. To measure cAMP inhibition by D2-like receptors (DRD2, DRD3 and DRD4) activating, cell-based luciferase activity assay was performed by pGL3-CRE-luciferase reporter systems (Promega), as described in a previous study ( Wu et al., 2019 ) . Briefly, cells expressing wild-type and mutant DRD2 were treated with dopamine diluted in 2 μM forskolin stimulating buffer. After 6 h incubation, 1 × Passive Lysis Buffer was added to the 96-well plate to measure luciferase assay in cell lysates as determined by the luciferase assay reagent kit (Promega). The luciferase activities in agonist treatment groups were first calculated as ratios to that of control group and normalized by wild-type receptor. To validate our function results, all experiments were repeated at least three times independently, each in triplicate. β-arrestin recruitment assay To measure β-arrestin recruitment by DRD1 upon PW0464 or A77636 activation, intracellular bioluminescence resonance energy transfer (BRET) assay was performed as previously described ( Yang et al., 2018 , 2020 ). Briefly, Flag-DRD1-WT-eYFP, β-arrestin-2-Rluc plasmids were co-transfected into HEK293 cells using PEI and cultured at 37°C with 5% CO 2 for 24 h. Afterward, the transfected cells were seeded (50,000 cells per well) into clear-bottom 96-well plate and cultured at 37°C for another 24 h. The cells were washed 3 times with 200 μL PBS and incubated with increasing concentrations of PW0464, A77636 or vehicle for 10 min. BRET signal was determined, as the ratio of the light emitted by YFP (530 nm) and Rluc (485 nm), after addition of 5 μM of luciferase substrate coelenterazine 400a (Cayman) using a Mithras LB940 microplate reader (Berthold Technologies). Molecule docking The GLIDE program (version 5.6, Schrödinger, LLC, New York, NY, 2010) ( Friesner et al., 2006 ; Halgren et al., 2004 ) was used for docking studies. Prior to docking, ligands were prepared using “Ligand Prep” module. For GLIDE docking, complex structures of DRD1 was preprocessed according to the protein preparation procedure recommended and docking site was defined in terms of ligands. All docking calculations were run in the “Standard Precision” (SP) mode ( Friesner et al., 2004 ) with default values for all parameters. Five poses per prepared ligand were saved for each docking calculation. The best-docked structure was chosen using the Glide Score ( Jorgensen et al., 1996 ). Molecular dynamics (MD) simulation Prior to running MD simulations, the receptor-ligand complex model was further prepared in Protein Preparation Wizard in Maestro (Schrödinger) to fix the missing side chains, add hydrogen atoms and cap the receptor chain termini while the residues D70 and D120 were protonated in the active-state simulations, all other titratable residues were left in their dominate protonation state at pH 7.0. The structures were inserted into a pre-equilibrated POPC bilayer and solvated with 0.15 M NaCl in explicitly water using dabble ( Betz, 2018 ). Parameters for the MD simulations were generated using the CHARMM-GUI web interface ( Jo et al., 2008 ; Lee et al., 2016 ; Wu et al., 2014 ) using the CHARMM36 forcefield ( Huang and MacKerell, 2013 ) with CGenFF ( Vanommeslaeghe et al., 2010 ) for ligands. Simulations were performed CUDA enabled version of PMEMD in AMBER16 ( Le Grand et al., 2013 ). The system was heated from 0 to 100 K in the NVT ensemble using the Langevin thermostat over 12.5 ps with harmonic restraints of 10.0 kcal mol −1 Å −2 on the non-hydrogen atoms of lipid, protein, and ligand. Then the system was heated to 310 K over 125 ps in the NPT ensemble with semi-isotropic pressure coupling and a pressure of 1 bar. Further equilibration was performed at 310K with harmonic restraints on the protein and ligand starting at 5.0 kcal mol −1 Å −2 and reduced by 1.0 kcal mol −1 Å −2 in a stepwise fashion every 2 ns, for a total of 10 ns of additional restrained equilibration. Multiple independent simulations were initialized from the final snapshot of the restrained equilibration. These simulations were conducted in the NPT ensemble at 310K and 1 bar, using a Langevin thermostat and Monte Carlo barostat under periodic boundary conditions. A time step of 2.0 fs with hydrogen mass repartitioning was used, bond lengths to hydrogen atoms were constrained using SHAKE ( Ryckaert et al., 1977 ). Non-bonded interactions were cut off at 9.0 Å, and long-range electrostatic interactions were computed using the particle mesh Ewald method. The simulation frames were written every 100 ps. Root-mean-square deviation (RMSD) was calculated using the CPPTRAJ module ( Roe and Cheatham, 2013 ) to monitor the fluctuation of the protein and ligand during the simulations.

QUANTIFICATION AND STATISTICAL ANALYSIS

The resolutions of all cryo-EM maps were estimated with the gold-standard Fourier shell correlation 0.143 criterion indicated in figure legends. The detailed can be found in Figure S1 , supplemental information Data S1 , legends and method. Statistical analysis of radioligand binding assays, cAMP function assays, BRET assays and receptor expression level assays were processed using Prism 7.0 (GraphPad). Data in figures and tables are presented as the mean ± standard error of the mean (SEM) of three independent experiments performed in triplicate indicated in figure and table legends. The “n” value in tables means the numbers of independent experiment. The details can be found in method and Figures 5C , 5D , S3A , S3B , S3G , S3I , S4A , S5F – S5H , S6B – S6H , S7E , S7H , and S7I , and Tables S2 – S5 and legends.

Supplementary Material Data S1 Data S2 3