Agonist Selectivity and Ion Permeation in the α3β4 Ganglionic Nicotinic Receptor.

Lead Contact and Materials Availability

Further information and requests for resources including α3 and β4 subunit expression constructs should be directed to and will be fulfilled by the Lead Contact, Ryan E. Hibbs ( ryan.hibbs@utsouthwestern.edu ). Experimental Model and Subject Details Cultured Cells HEK293S GnTI − cells were used for protein production and electrophysiology experiments. Cells were obtained from ATCC and were not further authenticated. Adherent cells were grown at 37°C with 5% CO 2 in DMEM medium (Corning) supplemented with 10% fetal bovine serum (EMD Millipore). Suspension cells were grown at 37°C with 8% CO 2 in FreeStyle 293 expression medium (Gibco) supplemented with 2% fetal bovine serum (EMD Millipore). Sf9 cells were used to produce baculovirus. Cells were obtained from ATCC and were not further authenticated. Cells were grown at 27°3 in SF900-III SFM expression medium (Gibco).

Method Details Construct design

Human α3 (UniProtKB P32297 ) and β4 ( P30926 ) genes were codon optimized, synthesized, and cloned into the pEZT-BM expression vector ( Morales-Perez et al., 2016a ). Enhanced green fluorescent protein (EGFP) was inserted into the M3-M4 loop of both subunits and a Strep-tag was placed at the C-terminus of the β4 subunit. Constructs were initially screened via co-transfection of HEK293S GnTI − cells (ATCC CRL-3022) with combinations of EGFP-tagged and untagged subunits via Lipofectamine 2000 (Invitrogen). Cells were pelleted, solubilized with 20 mM Tris, pH 7.4, 150 mM NaCl (TBS buffer), 40 mM n -dodecyl-β-D-maltopyranoside (DDM; Anatrace), and 1 mM phenylmethanesulfonyl fluoride (PMSF; Sigma-Aldrich), and analyzed by fluorescence-detection size-exclusion chromatography (FSEC) ( Kawate and Gouaux, 2006 ). Viable candidates without EGFP were then co-transfected in GnTI − cells in small-scale purification experiments (1-2 mL culture scale). Cells were pelleted, solubilized as before, and allowed to bind to high-capacity Strep-Tactin (IBA) affinity resin. Resin was washed with TBS containing 1 mM DDM, and protein was eluted with the same buffer supplemented with 5 mM desthiobiotin (Sigma-Aldrich) before being evaluated by FSEC monitoring tryptophan fluorescence. Screens revealed that robust expression and stable pentamer formation required deletions in the M3-M4 loop and were aided by the inclusion of soluble fusion partners in this region. Thus, in the final EM constructs, residues Asn348-Ser402 in α3 and Pro341-Ser395 in β4 were replaced with apocytochrome b(562)RIL (BRIL), a thermostabilized four-helical bundle that has been used to promote crystallization in G protein-coupled receptors ( Chun et al., 2012 ).

Receptor expression and purification

Bacmam viruses for each subunit were produced as described for the α4β2 receptor ( Morales-Perez et al., 2016a ). Briefly, constructs in the pEZT-BM vector were transformed in DH10-Bac cells to produce recombinant bacmid. 2 mL of Sf9 cells (ATCC CRL-1711) were transfected with purified bacmid DNA to generate “P1” virus. 500 μL of this virus was then added to 1 L of Sf9 cells at a cell density of 1 × 10 6 cells/ml to produce a second generation of amplified virus (“P2”). Suspension cultures of GnTI − cells were grown at 37°C, 8% CO 2 and were transduced with α3 and β4 P2 viruses at a cell density of 4 × 10 6 cells/ml. At the time of transduction, 1 mM sodium butyrate (Sigma-Aldrich) was added to the culture and temperature was dropped to 30°C to boost protein expression. Af ter 72 hours, cells were harvested by centrifugation, resuspended in TBS and 1 mM PMSF, and disrupted using an Avestin Emulsiflex. Lysed cells were centrifuged for 20 minutes at 10,000 g, and the resulting supernatants were centrifuged for 2 hours at 186,000 g to isolate membranes. Membrane pellets were mechanically homogenized and solubilized for 1 hour at 4°C with 40 mM DDM in TBS. Solubilized membranes were centrifuged for 40 minutes at 186,000 g then passed over high capacity Strep-Tactin resin. The resin was washed with TBS, 1 mM DDM, 0.2 mM cholesteryl hemisuccinate (CHS, Anatrace), and 1 mM TCEP (Thermo Fisher Scientific), and protein was eluted in the same buffer containing 5 mM desthiobiotin. For the nicotine-bound structure, 1 mM nicotine (Sigma) was included during affinity purification and for the AT-1001-bound structure, 10 μM AT-1001 (generous gift from Astrea Therapeutics) was included. α3β4 receptors have been proposed to assemble in multiple subunit stoichiometries: (α3) 2 (β4) 3 and (α3) 3 (β4) 2 ( Covernton et al., 1994 ; Grishin et al., 2010 ; Krashia et al., 2010 ). To bias expression towards one stoichiometry, a fluorescence-based assay was used, as described previously for the α4β2 nicotinic receptor ( Morales-Perez et al., 2016a ). Briefly, bacmam viruses were made and titered for α3-mCherry and β4-EGFP constructs. Viruses were used to transduce one liter of GnTI- cells. Protein was purified as described above, and molar concentrations of each subunit were calculated by measuring absorbance at the maxima for the two fluorophores (GFP, 488 nm; mCherry, 587 nm) and dividing by their respective extinction coefficients (GFP, 56,000 M −1 cm −1 ; mCherry, 72,000 M −1 cm −1 ). Protein was analyzed by FSEC measuring GFP and mCherry fluorescence, allowing scale factors to be calculated relating known molar concentrations to fluorescence signal. This medium-scale purification was followed by small-scale experiments where different viral ratios were used to transduce 1 mL of GnTI-cells. Cells were solubilized and analyzed by FSEC, and previously calculated scale factors were used to determine molar ratios of α3 and β4 subunits. A 1:1 ratio of α3 and β4 viruses was found to produce a homogenous population of pentamers containing two α3 and three β4 subunits, and thus this ratio of viruses was used to produce protein for structural studies, reconstitutions for electrophysiology, and for binding assays. Saposin nanodisc reconstitution The Saposin A expression plasmid was provided by Salipro Biotech AB. Reconstitution of α3β4 into saposin nanoparticles was modified from Nguyen et al. ( Nguyen et al., 2018 ) The reaction contained a 1:20:100 molar ratio of α3β4: saposin: soy polar lipid extract (Avanti). Lipids and saposin were mixed in TBS and 14 mM DDM and allowed to rotate at 4°C for 1 hour. Affinity-purified α3β4 was concentrated to ~20 μM, added to the saposin/lipid mixture, and rotated for 30 min at 4°C. 200 mg/mL Bio-Beads SM-2 (BioRad) were added to the mixture and rotation was continued overnight. The following morning, Bio-Beads were removed and replaced with 150 mg/mL fresh Bio-Beads for 2 hours. Generation of monoclonal antibodies and Fab fragments The 4G9 monoclonal antibody (mAb) (IgG2b, κ) was raised using standard methods following immunization of mice with α3β4 in detergent (Monoclonal Core, Vaccine and Gene Therapy Institute, Oregon Health & Science University). High affinity and specificity of the antibody for properly folded receptor was assayed by FSEC with EGFP-tagged receptor (shift in elution volume) and western blot (no binding). Fab fragments were generated by papain cleavage of whole antibody at a final concentration of 0.5 mg/ml for 2 hours at 37°C in 50 mM NaPO 4 , pH 7.0, 1 mM EDTA, 10 mM cysteine and 1:10 (w/w) papain. Digestion was quenched using 30 mM iodoacetamide at 25°C for 30 min. Fab w as purified by anion exchange using a HiTrap Q HP (GE Healthcare) column in 10 mM Tris, pH 8.0 and a NaCl gradient elution. Cloning and sequencing of Fab antibody regions were performed from mouse hybridoma cells.

Cryo-EM sample preparation and data collection

Affinity-purified α3β4 receptors reconstituted in nanodiscs were mixed with 4G9 Fab in a 1:1 (w/w) ratio and injected over a Superose 6 Increase 10/300 GL column (GE Healthcare) equilibrated in TBS, 1 mM TCEP, and ligand (1 mM nicotine or 50 μM AT-1001). Receptors purified in detergent followed the same protocol, but the buffer included 1 mM DDM and 0.2 mM CHS. Peak fractions were evaluated by analytical SEC, monitoring tryptophan fluorescence, and concentrated to an A280 of ~6. Samples in nanodiscs were supplemented with 1 mM Fos-Choline-8, fluorinated (Anatrace) immediately prior to freezing to induce random orientations in the grid holes. Protein sample (3 μL) was applied to glow-discharged gold R1.2/1.3 300 mesh holey carbon grids (Quantifoil) and immediately blotted for 4 s at 100% humidity and 4°C before being plunge-frozen into liquid ethane cooled by liquid nitrogen using a Vitrobot Mark IV (FEI). Cryo-EM data were collected on a 300 kV Titan Krios microscope (FEI) equipped with a K2 Summit direct electron detector (Gatan) and a GIF quantum energy filter (20 eV) (Gatan) using EPU (FEI) and a 200 kV Talos Arctica (FEI) equipped with a K3 direct electron detector (Gatan) using Serial EM ( Mastronarde, 2005 ). Sample-specific details are included in Table S1 .

Cryo-EM data processing

All datasets were processed using the same general workflow in RELION 3.0 ( Zivanov et al., 2018 ). Dose-fractionated images were gain normalized, 2 x Fourier binned, aligned, dose-weighted, and summed with MotionCor2 ( Zheng et al., 2017 ). Contrast transfer function correction and defocus value estimation were done with GCTF ( Zhang, 2016 ). Several hundred particles were manually picked and subjected to 2D classification to generate templates for auto-picking. Auto-picked particles were subjected to 2D classification to remove false positives. Ab initio models were generated in RELION and used for 3D classification. 3D classes with strong ICD density were selected for 3D refinement. An initial round of 3D refinement using the best 3D class as an initial model (low-pass filtered to 60 A) was followed by a second round with finer angular sampling using the map from the first refinement low-pass filtered to 10 Å as the initial model. Next, per-particle CTF refinement and beam tilt estimation were performed before another round of 3D classification with no image alignment/angular searches. Particles from the best classes were selected, polished, and used for 3D refinement to generate the final maps. In the AT-1001 detergent dataset, particles with defocus values greater than −3.0 μm were removed from the final reconstruction to improve resolution. Local resolution was estimated with ResMap ( Kucukelbir et al., 2014 ). Model building, refinement, and validation A homology model for the α3β4 receptor was generated from the cryo-EM structure of the 2α:3β assembly of the α4β2 nicotinic acetylcholine receptor (PDB ID:6CNJ) ( Walsh et al., 2018 ) via Swiss-Model ( Schwede et al., 2003 ). A homology model for the Fab fragment was made using PDB entry 4WFE ( Brohawn et al., 2014 ) for the light chain and 3MXV ( Maun et al., 2010 ) for the heavy chain. The receptor and one copy of the Fab were docked into the density map using UCSF-Chimera ( Pettersen et al., 2004 ). Manual adjustments of the models were then done in Coot ( Emsley et al., 2010 ). The ECD and TMD of each individual subunit, as well as the variable domain of the Fab fragment were rigid body fitted into the density map. The density in the region of the constant domain of the Fab was not sufficiently ordered to allow accurate building of an atomic model; therefore, only the variable domain was included in the final model. Once the variable domain was rebuilt into the density, it was copied into the second site and manually adjusted. MA helices, which were not included in the α4β2 structure, were built de novo, and M4 helices, which are continuous with MA, were rebuilt into the map. Well-ordered N-linked glycans were built along the surface of the ECD. In many cases, density for these glycans was clearer in a 6 Å low-pass filtered map ( Fig. S4C - D ), and this map was used to assist building. Likewise, the water molecule in the ligand-binding pocket was apparent in the full map but showed stronger density in a 5 Å high-pass filtered map, which was used for accurate placement of this water. A portion of the M3-M4 loop (including the bril fusion protein) in both subunits was unresolved and the following residues were not modeled: 328-409 in α3 and 329-399 in β4. Additionally, residues 1-3 were not modeled in chain C (β4) due to weak density. After manual building in Coot, global real space coordinate and B-factor refinement were performed in Phenix ( Adams et al., 2010 ). Sequences used in alignments were retrieved from the UnitProtKB database ( UniProt Consortium, 2018 ). Sequence alignments were made using PROMALS3D ( Pei et al., 2008 ). Pore radius profiles and hydrophobicity plots were made using CHAP ( Klesse et al., 2019 ). Structural figures were made using UCSF-Chimera and PyMOL (Schrodinger, LLC). Structural biology software packages were compiled by SBGrid ( Morin et al., 2013 ).

Electrophysiology

Whole cell voltage-clamp recordings were made from cells transiently transfected with the constructs used in structural analysis. For the patch-clamp experiments, adherent HEK293S GnTI- cells were transiently transfected with pEZT-based plasmids 2-3 days before recording. Each 35 mm dish of cells was transfected with the DNA of α3 and β4 subunits in a 1:1 ratio. Upon transfection, cells were moved to 30°C. On the day of recording, cells were washed with bath solution, which contained (in mM): 140 NaCl, 2.4 KCl, 4 MgCl 2 , 4 CaCl 2 , 10 HEPES pH 7.3, and 10 glucose. Borosilicate pipettes were pulled and polished to a resistance of 2-4 MΩ. The pipette solution contained (in mM): 150 CsCl, 10 NaCl, 10 EGTA, and 20 HEPES pH 7.3. Cells were clamped at −75 mV. The recordings were made with an Axopatch 200B amplifier, sampled at 5 kHz, and low-pass filtered at 2 kHz using a Digidata 1440A (Molecular Devices) and analyzed with pClamp 10 software (Molecular Devices). The nicotine and AT-1001 solutions were prepared in bath solution from concentrated stocks. A stock solution of 1 M nicotine was prepared in water and the stock solution of 100 mM AT-1001 was prepared in DMSO. Solution exchange was achieved using a gravity driven RSC-200 rapid solution changer (Bio-Logic). Cell-attached single channel recordings were made from cells 1-2 days post-transient transfection, following the same procedure for transfection as above. On the day of recording, cells were washed with bath solution containing (in mM): 142 KCl, 5.4 NaCl, 1.8 CaCl 2 , 1.7 MgCl 2 , and 10 HEPES pH 7.4 (adjusted with KOH) ( Mukhtasimova et al., 2016 ). Borosilicate pipettes were pulled and polished to initial resistances of 8-12 MΩ. The pipette solution contained (in mM): 80 KF, 20 KCl, 40 potassium aspartate, 2 MgCl 2 , 1 EGTA, 10 HEPES 7.4 (adjusted with KOH), and 0.05 nicotine ( Mukhtasimova et al., 2016 ). Currents were recorded at 100 mV, sampled at 50 kHz, and filtered at 10 kHz. For proteoliposome patch-clamp experiments, receptors were first affinity-purified in DDM/CHS, as described above. Soy polar lipids in chloroform (Avanti) were dried in a test tube under a stream of argon while rotating the tube to make a homogeneous lipid film. The lipid film was further dried under vacuum for 2 hours and resuspended to 10 mg/mL with TBS. To make uniform lipid vesicles, the lipid resuspension solution was sonicated for 15 min. Purified receptors (4 μg) were added into lipid vesicles in a protein to lipid mass ratio of 1:500 (w/w). The mixture was rotated at room temperature for 1 hour to allow the protein to incorporate into lipid vesicles. Detergent was removed by incubating with Bio-Beads SM-2 and the resultant liposomes were collected by ultracentrifugation, 4°C, 30 min at 186,000 g. The pellet was resuspended in 6 μL TBS buffer. 2 μL of the suspension was spotted on a glass coverslip, and then desiccated overnight under vacuum at 4°C. Desi ccated liposomes were rehydrated with 5 μL of buffer (320 mM sucrose, 10 mM KCl, and 2 mM MgCl 2 , 5 mM Hepes pH 8.0) for at least 2 hours at 4°C, and then used for patch-clamp recordi ng. Channel activity of α3β4 was examined in excised liposome patches. Data were acquired at 70 mV at a sampling rate of 50 kHz with a 10 kHz filter. The bath solution contained (in mM): 200 KCl, 40 MgCl 2 , 2 CaCl 2 , and 5 HEPES pH 7.3. Pipettes were filled with the same bath solution and initial pipette resistances ranged from 4-8 MΩ. After a stable baseline was observed, 10 mM nicotine in bath solution was added to the bath to achieve a final concentration of ~0.5 mM. Radioligand binding Experiments to measure binding of [ 3 H]-epibatidine (PerkinElmer) to the α3β4 receptor were performed with protein purified in TBS with 1 mM DDM, 0.2 mM CHS, and 1 mM TCEP in the absence of agonists. The concentration of binding sites was 0.3 nM. For the binding experiments in the presence of Fab, Fab was added in large excess (1 μM). In addition to the receptor, the binding assay conditions included 20 mM Tris pH 7.4, 150 mM NaCl, 1 mM DDM, and 1 mg/mL streptavidin-YiSi scintillation proximity assay beads (SPA; GE Healthcare Life Sciences). Non-specific signal was determined in the presence of 1 mM [ 1 H]-nicotine. All data shown are from background-subtracted measurements. For radioligand competition experiments, binding site concentration was also 0.3 nM and the concentration of [ 3 H]-epibatidine was 1 nM.

Molecular dynamics simulations

Deposited coordinates for the cryo-electron microscopy structure of the α3β4 receptor bound to nicotine in nanodiscs were used as a starting model for molecular dynamics simulations. Nicotine and AT-1001 parameters were generated using STaGE ( Lundborg and Lindahl, 2015 ), and virtual sites were added. Nicotine, CHS, and ions and water resolved in the channel pore were placed as in the deposited structure. To resolve instabilities observed in the partially resolved intracellular domain, 1-palmitoyl 2-oleoyl phosphatidylcholine (POPC) was docked using AutoDock Vina ( Trott and Olson, 2010 ) in a 37.5 Å × 36 Å × 39.75 Å box surrounding the MA helical bundle. For simulations with AT-1001, the partial agonist was substituted for nicotine at both binding sites. The Amber99sb-ildn force field ( Lindorff-Larsen et al., 2010 ) was used to describe each protein, which was embedded in a bilayer of 300 POPC molecules modeled with Slipids-extended force field parameters ( Jämbeck and Lyubartsev, 2012 , 2013 ). Each system was solvated in a cubic box using CHARMM-GUI ( Jo et al., 2008 ; Wu et al., 2014 ) and the TIP3P water model ( Jorgensen et al., 1983 ), and NaCl was added to bring the system to neutral charge and an ionic strength of 0.15 M. All simulations were performed with GROMACS 2018 ( Abraham et al., 2015 ). Each system was energy-minimized with a velocity rescaling thermostat ( Bussi et al., 2007 ) set to 300 K, then equilibrated for 50 ps, both with a constant number of particles, volume, and temperature. Virtual interaction sites were used for hydrogens to enable 5-fs time steps. Each was then equilibrated with a constant number of particles, pressure, and temperature for at least 60 ns, during which the position restraints on the protein were gradually released. Agonists, CHS, and resolved ions and water in the channel pore were restrained until the final 15 ns of equilibration. For each equilibrated system containing POPC in the intracellular domain, three replicates of 500-ns unrestrained simulations were generated. An additional 200-ns unrestrained simulation was performed for the nicotine-bound model in the absence of intracellular POPC. Parrinello-Rahman pressure coupling ( Parrinello and Rahman, 1980 ) ensured constant pressure, the particle mesh Ewald algorithm ( Essmann et al., 1995 ) was used for long-range electrostatic interactions, and bond lengths were constrained using the LINCS algorithm ( Hess, 2008 ). Analyses were performed using VMD ( Humphrey et al., 1996 ), CHAP, and MDTraj ( McGibbon et al., 2015 ).

Quantification and Statistical Analysis

Statistical analyses were performed using Prism 8 (GraphPad). To quantify differences in single channel currents between EM and mutant constructs, mean and standard deviations were calculated from three independent patches for each group. Statistical significance was determined with an unpaired t-test. To quantify differences in binding affinities and hill slopes, each set of binding reaction experiments was performed three to five times. For each independent experiment, measurements were taken in triplicate. From these triplicate measurements, mean and standard error were calculated, and K d and n H values were determined by nonlinear regression. For competition experiments, IC 50 and n H values were determined by nonlinear regression and K i values were calculated using experimental K d values for [ 3 H]-epibatidine. Statistical significance between +Fab and −Fab groups was determined with unpaired t-tests.

Data and Code Availability

The structures and EM density maps generated in this study are available in the PDB and EMDB respectively. The accession numbers for the reported data are PDB: 6PV7, EMDB: EMD-20487 (α3β4 EM -Nicotine complex); PDB: 6PV8, EMDB: EMD-20488 (α3β4 EM -AT-1001 complex in DDM); EMDB: EMD-20489 (α3β4 EM -Nicotine complex without CHS); EMDB: EMD-20490 (α3β4 EM -AT-1001 complex in nanodiscs).

Lead Contact and Materials Availability

Further information and requests for resources including α3 and β4 subunit expression constructs should be directed to and will be fulfilled by the Lead Contact, Ryan E. Hibbs ( ryan.hibbs@utsouthwestern.edu ).

Experimental Model and Subject Details Cultured Cells HEK293S GnTI − cells were used for protein production and electrophysiology experiments. Cells were obtained from ATCC and were not further authenticated. Adherent cells were grown at 37°C with 5% CO 2 in DMEM medium (Corning) supplemented with 10% fetal bovine serum (EMD Millipore). Suspension cells were grown at 37°C with 8% CO 2 in FreeStyle 293 expression medium (Gibco) supplemented with 2% fetal bovine serum (EMD Millipore). Sf9 cells were used to produce baculovirus. Cells were obtained from ATCC and were not further authenticated. Cells were grown at 27°3 in SF900-III SFM expression medium (Gibco).

Method Details Construct design

Human α3 (UniProtKB P32297 ) and β4 ( P30926 ) genes were codon optimized, synthesized, and cloned into the pEZT-BM expression vector ( Morales-Perez et al., 2016a ). Enhanced green fluorescent protein (EGFP) was inserted into the M3-M4 loop of both subunits and a Strep-tag was placed at the C-terminus of the β4 subunit. Constructs were initially screened via co-transfection of HEK293S GnTI − cells (ATCC CRL-3022) with combinations of EGFP-tagged and untagged subunits via Lipofectamine 2000 (Invitrogen). Cells were pelleted, solubilized with 20 mM Tris, pH 7.4, 150 mM NaCl (TBS buffer), 40 mM n -dodecyl-β-D-maltopyranoside (DDM; Anatrace), and 1 mM phenylmethanesulfonyl fluoride (PMSF; Sigma-Aldrich), and analyzed by fluorescence-detection size-exclusion chromatography (FSEC) ( Kawate and Gouaux, 2006 ). Viable candidates without EGFP were then co-transfected in GnTI − cells in small-scale purification experiments (1-2 mL culture scale). Cells were pelleted, solubilized as before, and allowed to bind to high-capacity Strep-Tactin (IBA) affinity resin. Resin was washed with TBS containing 1 mM DDM, and protein was eluted with the same buffer supplemented with 5 mM desthiobiotin (Sigma-Aldrich) before being evaluated by FSEC monitoring tryptophan fluorescence. Screens revealed that robust expression and stable pentamer formation required deletions in the M3-M4 loop and were aided by the inclusion of soluble fusion partners in this region. Thus, in the final EM constructs, residues Asn348-Ser402 in α3 and Pro341-Ser395 in β4 were replaced with apocytochrome b(562)RIL (BRIL), a thermostabilized four-helical bundle that has been used to promote crystallization in G protein-coupled receptors ( Chun et al., 2012 ).

Receptor expression and purification

Bacmam viruses for each subunit were produced as described for the α4β2 receptor ( Morales-Perez et al., 2016a ). Briefly, constructs in the pEZT-BM vector were transformed in DH10-Bac cells to produce recombinant bacmid. 2 mL of Sf9 cells (ATCC CRL-1711) were transfected with purified bacmid DNA to generate “P1” virus. 500 μL of this virus was then added to 1 L of Sf9 cells at a cell density of 1 × 10 6 cells/ml to produce a second generation of amplified virus (“P2”). Suspension cultures of GnTI − cells were grown at 37°C, 8% CO 2 and were transduced with α3 and β4 P2 viruses at a cell density of 4 × 10 6 cells/ml. At the time of transduction, 1 mM sodium butyrate (Sigma-Aldrich) was added to the culture and temperature was dropped to 30°C to boost protein expression. Af ter 72 hours, cells were harvested by centrifugation, resuspended in TBS and 1 mM PMSF, and disrupted using an Avestin Emulsiflex. Lysed cells were centrifuged for 20 minutes at 10,000 g, and the resulting supernatants were centrifuged for 2 hours at 186,000 g to isolate membranes. Membrane pellets were mechanically homogenized and solubilized for 1 hour at 4°C with 40 mM DDM in TBS. Solubilized membranes were centrifuged for 40 minutes at 186,000 g then passed over high capacity Strep-Tactin resin. The resin was washed with TBS, 1 mM DDM, 0.2 mM cholesteryl hemisuccinate (CHS, Anatrace), and 1 mM TCEP (Thermo Fisher Scientific), and protein was eluted in the same buffer containing 5 mM desthiobiotin. For the nicotine-bound structure, 1 mM nicotine (Sigma) was included during affinity purification and for the AT-1001-bound structure, 10 μM AT-1001 (generous gift from Astrea Therapeutics) was included. α3β4 receptors have been proposed to assemble in multiple subunit stoichiometries: (α3) 2 (β4) 3 and (α3) 3 (β4) 2 ( Covernton et al., 1994 ; Grishin et al., 2010 ; Krashia et al., 2010 ). To bias expression towards one stoichiometry, a fluorescence-based assay was used, as described previously for the α4β2 nicotinic receptor ( Morales-Perez et al., 2016a ). Briefly, bacmam viruses were made and titered for α3-mCherry and β4-EGFP constructs. Viruses were used to transduce one liter of GnTI- cells. Protein was purified as described above, and molar concentrations of each subunit were calculated by measuring absorbance at the maxima for the two fluorophores (GFP, 488 nm; mCherry, 587 nm) and dividing by their respective extinction coefficients (GFP, 56,000 M −1 cm −1 ; mCherry, 72,000 M −1 cm −1 ). Protein was analyzed by FSEC measuring GFP and mCherry fluorescence, allowing scale factors to be calculated relating known molar concentrations to fluorescence signal. This medium-scale purification was followed by small-scale experiments where different viral ratios were used to transduce 1 mL of GnTI-cells. Cells were solubilized and analyzed by FSEC, and previously calculated scale factors were used to determine molar ratios of α3 and β4 subunits. A 1:1 ratio of α3 and β4 viruses was found to produce a homogenous population of pentamers containing two α3 and three β4 subunits, and thus this ratio of viruses was used to produce protein for structural studies, reconstitutions for electrophysiology, and for binding assays. Saposin nanodisc reconstitution The Saposin A expression plasmid was provided by Salipro Biotech AB. Reconstitution of α3β4 into saposin nanoparticles was modified from Nguyen et al. ( Nguyen et al., 2018 ) The reaction contained a 1:20:100 molar ratio of α3β4: saposin: soy polar lipid extract (Avanti). Lipids and saposin were mixed in TBS and 14 mM DDM and allowed to rotate at 4°C for 1 hour. Affinity-purified α3β4 was concentrated to ~20 μM, added to the saposin/lipid mixture, and rotated for 30 min at 4°C. 200 mg/mL Bio-Beads SM-2 (BioRad) were added to the mixture and rotation was continued overnight. The following morning, Bio-Beads were removed and replaced with 150 mg/mL fresh Bio-Beads for 2 hours. Generation of monoclonal antibodies and Fab fragments The 4G9 monoclonal antibody (mAb) (IgG2b, κ) was raised using standard methods following immunization of mice with α3β4 in detergent (Monoclonal Core, Vaccine and Gene Therapy Institute, Oregon Health & Science University). High affinity and specificity of the antibody for properly folded receptor was assayed by FSEC with EGFP-tagged receptor (shift in elution volume) and western blot (no binding). Fab fragments were generated by papain cleavage of whole antibody at a final concentration of 0.5 mg/ml for 2 hours at 37°C in 50 mM NaPO 4 , pH 7.0, 1 mM EDTA, 10 mM cysteine and 1:10 (w/w) papain. Digestion was quenched using 30 mM iodoacetamide at 25°C for 30 min. Fab w as purified by anion exchange using a HiTrap Q HP (GE Healthcare) column in 10 mM Tris, pH 8.0 and a NaCl gradient elution. Cloning and sequencing of Fab antibody regions were performed from mouse hybridoma cells.

Cryo-EM sample preparation and data collection

Affinity-purified α3β4 receptors reconstituted in nanodiscs were mixed with 4G9 Fab in a 1:1 (w/w) ratio and injected over a Superose 6 Increase 10/300 GL column (GE Healthcare) equilibrated in TBS, 1 mM TCEP, and ligand (1 mM nicotine or 50 μM AT-1001). Receptors purified in detergent followed the same protocol, but the buffer included 1 mM DDM and 0.2 mM CHS. Peak fractions were evaluated by analytical SEC, monitoring tryptophan fluorescence, and concentrated to an A280 of ~6. Samples in nanodiscs were supplemented with 1 mM Fos-Choline-8, fluorinated (Anatrace) immediately prior to freezing to induce random orientations in the grid holes. Protein sample (3 μL) was applied to glow-discharged gold R1.2/1.3 300 mesh holey carbon grids (Quantifoil) and immediately blotted for 4 s at 100% humidity and 4°C before being plunge-frozen into liquid ethane cooled by liquid nitrogen using a Vitrobot Mark IV (FEI). Cryo-EM data were collected on a 300 kV Titan Krios microscope (FEI) equipped with a K2 Summit direct electron detector (Gatan) and a GIF quantum energy filter (20 eV) (Gatan) using EPU (FEI) and a 200 kV Talos Arctica (FEI) equipped with a K3 direct electron detector (Gatan) using Serial EM ( Mastronarde, 2005 ). Sample-specific details are included in Table S1 .

Cryo-EM data processing

All datasets were processed using the same general workflow in RELION 3.0 ( Zivanov et al., 2018 ). Dose-fractionated images were gain normalized, 2 x Fourier binned, aligned, dose-weighted, and summed with MotionCor2 ( Zheng et al., 2017 ). Contrast transfer function correction and defocus value estimation were done with GCTF ( Zhang, 2016 ). Several hundred particles were manually picked and subjected to 2D classification to generate templates for auto-picking. Auto-picked particles were subjected to 2D classification to remove false positives. Ab initio models were generated in RELION and used for 3D classification. 3D classes with strong ICD density were selected for 3D refinement. An initial round of 3D refinement using the best 3D class as an initial model (low-pass filtered to 60 A) was followed by a second round with finer angular sampling using the map from the first refinement low-pass filtered to 10 Å as the initial model. Next, per-particle CTF refinement and beam tilt estimation were performed before another round of 3D classification with no image alignment/angular searches. Particles from the best classes were selected, polished, and used for 3D refinement to generate the final maps. In the AT-1001 detergent dataset, particles with defocus values greater than −3.0 μm were removed from the final reconstruction to improve resolution. Local resolution was estimated with ResMap ( Kucukelbir et al., 2014 ). Model building, refinement, and validation A homology model for the α3β4 receptor was generated from the cryo-EM structure of the 2α:3β assembly of the α4β2 nicotinic acetylcholine receptor (PDB ID:6CNJ) ( Walsh et al., 2018 ) via Swiss-Model ( Schwede et al., 2003 ). A homology model for the Fab fragment was made using PDB entry 4WFE ( Brohawn et al., 2014 ) for the light chain and 3MXV ( Maun et al., 2010 ) for the heavy chain. The receptor and one copy of the Fab were docked into the density map using UCSF-Chimera ( Pettersen et al., 2004 ). Manual adjustments of the models were then done in Coot ( Emsley et al., 2010 ). The ECD and TMD of each individual subunit, as well as the variable domain of the Fab fragment were rigid body fitted into the density map. The density in the region of the constant domain of the Fab was not sufficiently ordered to allow accurate building of an atomic model; therefore, only the variable domain was included in the final model. Once the variable domain was rebuilt into the density, it was copied into the second site and manually adjusted. MA helices, which were not included in the α4β2 structure, were built de novo, and M4 helices, which are continuous with MA, were rebuilt into the map. Well-ordered N-linked glycans were built along the surface of the ECD. In many cases, density for these glycans was clearer in a 6 Å low-pass filtered map ( Fig. S4C - D ), and this map was used to assist building. Likewise, the water molecule in the ligand-binding pocket was apparent in the full map but showed stronger density in a 5 Å high-pass filtered map, which was used for accurate placement of this water. A portion of the M3-M4 loop (including the bril fusion protein) in both subunits was unresolved and the following residues were not modeled: 328-409 in α3 and 329-399 in β4. Additionally, residues 1-3 were not modeled in chain C (β4) due to weak density. After manual building in Coot, global real space coordinate and B-factor refinement were performed in Phenix ( Adams et al., 2010 ). Sequences used in alignments were retrieved from the UnitProtKB database ( UniProt Consortium, 2018 ). Sequence alignments were made using PROMALS3D ( Pei et al., 2008 ). Pore radius profiles and hydrophobicity plots were made using CHAP ( Klesse et al., 2019 ). Structural figures were made using UCSF-Chimera and PyMOL (Schrodinger, LLC). Structural biology software packages were compiled by SBGrid ( Morin et al., 2013 ).

Electrophysiology

Whole cell voltage-clamp recordings were made from cells transiently transfected with the constructs used in structural analysis. For the patch-clamp experiments, adherent HEK293S GnTI- cells were transiently transfected with pEZT-based plasmids 2-3 days before recording. Each 35 mm dish of cells was transfected with the DNA of α3 and β4 subunits in a 1:1 ratio. Upon transfection, cells were moved to 30°C. On the day of recording, cells were washed with bath solution, which contained (in mM): 140 NaCl, 2.4 KCl, 4 MgCl 2 , 4 CaCl 2 , 10 HEPES pH 7.3, and 10 glucose. Borosilicate pipettes were pulled and polished to a resistance of 2-4 MΩ. The pipette solution contained (in mM): 150 CsCl, 10 NaCl, 10 EGTA, and 20 HEPES pH 7.3. Cells were clamped at −75 mV. The recordings were made with an Axopatch 200B amplifier, sampled at 5 kHz, and low-pass filtered at 2 kHz using a Digidata 1440A (Molecular Devices) and analyzed with pClamp 10 software (Molecular Devices). The nicotine and AT-1001 solutions were prepared in bath solution from concentrated stocks. A stock solution of 1 M nicotine was prepared in water and the stock solution of 100 mM AT-1001 was prepared in DMSO. Solution exchange was achieved using a gravity driven RSC-200 rapid solution changer (Bio-Logic). Cell-attached single channel recordings were made from cells 1-2 days post-transient transfection, following the same procedure for transfection as above. On the day of recording, cells were washed with bath solution containing (in mM): 142 KCl, 5.4 NaCl, 1.8 CaCl 2 , 1.7 MgCl 2 , and 10 HEPES pH 7.4 (adjusted with KOH) ( Mukhtasimova et al., 2016 ). Borosilicate pipettes were pulled and polished to initial resistances of 8-12 MΩ. The pipette solution contained (in mM): 80 KF, 20 KCl, 40 potassium aspartate, 2 MgCl 2 , 1 EGTA, 10 HEPES 7.4 (adjusted with KOH), and 0.05 nicotine ( Mukhtasimova et al., 2016 ). Currents were recorded at 100 mV, sampled at 50 kHz, and filtered at 10 kHz. For proteoliposome patch-clamp experiments, receptors were first affinity-purified in DDM/CHS, as described above. Soy polar lipids in chloroform (Avanti) were dried in a test tube under a stream of argon while rotating the tube to make a homogeneous lipid film. The lipid film was further dried under vacuum for 2 hours and resuspended to 10 mg/mL with TBS. To make uniform lipid vesicles, the lipid resuspension solution was sonicated for 15 min. Purified receptors (4 μg) were added into lipid vesicles in a protein to lipid mass ratio of 1:500 (w/w). The mixture was rotated at room temperature for 1 hour to allow the protein to incorporate into lipid vesicles. Detergent was removed by incubating with Bio-Beads SM-2 and the resultant liposomes were collected by ultracentrifugation, 4°C, 30 min at 186,000 g. The pellet was resuspended in 6 μL TBS buffer. 2 μL of the suspension was spotted on a glass coverslip, and then desiccated overnight under vacuum at 4°C. Desi ccated liposomes were rehydrated with 5 μL of buffer (320 mM sucrose, 10 mM KCl, and 2 mM MgCl 2 , 5 mM Hepes pH 8.0) for at least 2 hours at 4°C, and then used for patch-clamp recordi ng. Channel activity of α3β4 was examined in excised liposome patches. Data were acquired at 70 mV at a sampling rate of 50 kHz with a 10 kHz filter. The bath solution contained (in mM): 200 KCl, 40 MgCl 2 , 2 CaCl 2 , and 5 HEPES pH 7.3. Pipettes were filled with the same bath solution and initial pipette resistances ranged from 4-8 MΩ. After a stable baseline was observed, 10 mM nicotine in bath solution was added to the bath to achieve a final concentration of ~0.5 mM. Radioligand binding Experiments to measure binding of [ 3 H]-epibatidine (PerkinElmer) to the α3β4 receptor were performed with protein purified in TBS with 1 mM DDM, 0.2 mM CHS, and 1 mM TCEP in the absence of agonists. The concentration of binding sites was 0.3 nM. For the binding experiments in the presence of Fab, Fab was added in large excess (1 μM). In addition to the receptor, the binding assay conditions included 20 mM Tris pH 7.4, 150 mM NaCl, 1 mM DDM, and 1 mg/mL streptavidin-YiSi scintillation proximity assay beads (SPA; GE Healthcare Life Sciences). Non-specific signal was determined in the presence of 1 mM [ 1 H]-nicotine. All data shown are from background-subtracted measurements. For radioligand competition experiments, binding site concentration was also 0.3 nM and the concentration of [ 3 H]-epibatidine was 1 nM.

Molecular dynamics simulations

Deposited coordinates for the cryo-electron microscopy structure of the α3β4 receptor bound to nicotine in nanodiscs were used as a starting model for molecular dynamics simulations. Nicotine and AT-1001 parameters were generated using STaGE ( Lundborg and Lindahl, 2015 ), and virtual sites were added. Nicotine, CHS, and ions and water resolved in the channel pore were placed as in the deposited structure. To resolve instabilities observed in the partially resolved intracellular domain, 1-palmitoyl 2-oleoyl phosphatidylcholine (POPC) was docked using AutoDock Vina ( Trott and Olson, 2010 ) in a 37.5 Å × 36 Å × 39.75 Å box surrounding the MA helical bundle. For simulations with AT-1001, the partial agonist was substituted for nicotine at both binding sites. The Amber99sb-ildn force field ( Lindorff-Larsen et al., 2010 ) was used to describe each protein, which was embedded in a bilayer of 300 POPC molecules modeled with Slipids-extended force field parameters ( Jämbeck and Lyubartsev, 2012 , 2013 ). Each system was solvated in a cubic box using CHARMM-GUI ( Jo et al., 2008 ; Wu et al., 2014 ) and the TIP3P water model ( Jorgensen et al., 1983 ), and NaCl was added to bring the system to neutral charge and an ionic strength of 0.15 M. All simulations were performed with GROMACS 2018 ( Abraham et al., 2015 ). Each system was energy-minimized with a velocity rescaling thermostat ( Bussi et al., 2007 ) set to 300 K, then equilibrated for 50 ps, both with a constant number of particles, volume, and temperature. Virtual interaction sites were used for hydrogens to enable 5-fs time steps. Each was then equilibrated with a constant number of particles, pressure, and temperature for at least 60 ns, during which the position restraints on the protein were gradually released. Agonists, CHS, and resolved ions and water in the channel pore were restrained until the final 15 ns of equilibration. For each equilibrated system containing POPC in the intracellular domain, three replicates of 500-ns unrestrained simulations were generated. An additional 200-ns unrestrained simulation was performed for the nicotine-bound model in the absence of intracellular POPC. Parrinello-Rahman pressure coupling ( Parrinello and Rahman, 1980 ) ensured constant pressure, the particle mesh Ewald algorithm ( Essmann et al., 1995 ) was used for long-range electrostatic interactions, and bond lengths were constrained using the LINCS algorithm ( Hess, 2008 ). Analyses were performed using VMD ( Humphrey et al., 1996 ), CHAP, and MDTraj ( McGibbon et al., 2015 ).

Supplementary Material 2 3 Video S1: Density map for nicotine and water, related to Figure 3 . 4 Video S2: Density map for AT-1001 at Site 1, related to Figure 3 . 5 Video S3: Density map for AT-1001 at Site 2, related to Figure 3 .