Molecular mechanism of choline and ethanolamine transport in humans – Nature

Generation of inducible HEK293 stable cell lines

The complementary DNAs of full-length wild-type FLVCR1 (human SLC49A1, NCBI reference sequence NM_014053) and FLVCR2 (human SLC49A2, NCBI reference sequence NM_017791) were cloned into pcDNA5/FRT/TO (Invitrogen) vectors, respectively. The gene for both FLVCRs was modified by a C-terminal FLAG fusion tag. Further details are found in sequence data provided in Supplementary Tables 1 and 2. The recombinant Flp-In T-REx293-FLVCR1 and Flp-In T-REx293-FLVCR2 cell lines were generated by using a tetracycline-inducible and commercially available Flp-In T-REx293 host-cell line system from Invitrogen. Flp-In T-REx293 cells were cultured in high-glucose Dulbecco’s Modified Eagle’s Medium (DMEM; Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS; Gibco), 1% Pen/Strep (Gibco), 1 μg ml⁻¹ of Zeocin (Gibco) and 15 μg ml⁻¹ of blasticidin S hydrochloride (AppliChem) at 37 °C in an atmosphere of 5% CO₂. Cells were periodically tested negative for mycoplasma contamination. For stable integration, the pcDNA5/FRT-FLVCR1-FLAG and pcDNA5/FRT-FLVCR2-FLAG vectors were cotransfected with the Flp recombinase encoding expression vector pOG44 (Invitrogen) at a 1:13 mass ratio, respectively. All transfection procedures were performed with Lipofectamine 2000 reagent according to the manufacturer’s instructions (Invitrogen). To select for stable clones, transfected cells were cultivated with growth medium containing 100 μg ml⁻¹ of hygromycin B (AppliChem).

Transport assays in HEK293 cells

HEK293 cells were cotransfected with pcDNA3.1 plasmid and human FLVCR1 or FLVCR2 and human choline kinase A (CHKA) for choline transport assays or ethanolamine kinase 1 (ETNK1) for ethanolamine transport assays using Lipofectamine 2000 reagent (Invitrogen). Cells were periodically tested negative for mycoplasma contamination. After 24 h post-transfection, cells were incubated with DMEM containing 20 μM [³H]choline or 2.5 μM [¹⁴C]ethanolamine. The cells were incubated at 37 °C and 5% CO₂ for 1 h for uptake of the ligands. The cells were subsequently washed with ice-cold plain DMEM and lysed with RIPA buffer (Thermo Scientific) by shaking at room temperature for 30 min. The cell lysates were quantified by scintillation counter Tri-Carb (Perkin Elmer). Radioactive signals from cell lysates were normalized to total protein levels. For dose curve assays, indicated concentrations of choline and ethanolamine were incubated with the cells for 1 h at 37 °C. For time-course assays, the cells were incubated with 20 μM [³H]choline or with 2.5 μM [¹⁴C]ethanolamine. The transport assays were stopped at indicated time points by adding ice-cold plain DMEM. For testing transport activity of FLVCR1 mutants, 20 μM [³H]choline and 2.5 μM [¹⁴C]ethanolamine were used. For testing transport activity of FLVCR2 mutants, 100 μM [³H]choline and 2.5 μM [¹⁴C]ethanolamine were used. For transport assays of HEK293 cells overexpressing FLVCR1 or FLVCR2 without co-expressing with CHKA or ETNK1, 20 μM [³H]choline and 2.5 μM [¹⁴C]ethanolamine were used.

For transport assays under indicated pH conditions, the following buffers were used: pH 8.5 buffer (140 mM NaCl, 20 mM Tris-HCl pH 8.5, 2 mM CaCl₂, 1 g l⁻¹ of d-glucose), pH 6.5 buffer (140 mM NaCl, 20 mM MES pH 6.5, 2 mM CaCl₂, 1 g l⁻¹ of d-glucose) or pH 7.5 buffer (140 mM NaCl, 20 mM HEPES-NaOH, 2 mM CaCl₂, 1 g l⁻¹ of d-glucose). For sodium-free buffer, buffer containing 140 mM KCl, 20 mM HEPES-KOH pH 7.5, 2 mM CaCl₂, 1 g l⁻¹ of d-glucose was used. In these assays, 20 μM [³H]choline or 2.5 μM [¹⁴C]ethanolamine was used and the assays were stopped after 15 min of incubation with the ligands. Radioactive signals from cell lysates were normalized to total protein levels. Total protein was quantified using Pierce BCA Protein Assay Kit (Thermo Scientific).

Immunofluorescent staining

HEK293 cells were seeded onto 24-well plates with coverslips and maintained in DMEM (Gibco) supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin. HEK293 cells were cotransfected with FLVCR1 or FLVCR2 with membrane expressing GFP (Addgene: catalogue no. 14757) using Lipofectamine 2000 reagent (Invitrogen). The inducible HEK293 stable cell lines overproducing FLVCR1 or FLVCR2 were seeded onto Millicell EZ SLIDE eight-well glass slides (Millipore), respectively. The stable cell lines were induced at 80% confluence by adding a final concentration of 2 μg ml⁻¹ of doxycycline hydrochloride. The protein overproduction was carried out for 24 h. For permeabilization and staining, cells were washed with PBS twice and fixed in 4% PFA for 15 min at room temperature, followed by washing with PBS twice and permeabilized in PBST (PBS with 0.5% Triton-X) for 15 min at room temperature. For immunofluorescent staining, the HEK293 cells were subsequently washed with PBS and blocked in 5% normal goat serum for 1 h before staining with FLVCR1 and FLVCR2 polyclonal antibodies at 1:250 dilutions for 1 h and then with Alexa Fluor 555 (A-21428, Invitrogen) as secondary antibody at 1:500 dilutions for 1 h. The cells were counter-stained with DAPI (Thermo Scientific) and imaged with a laser confocal microscope (Zeiss LSM710). The overproduction stable cells were treated with the same protocol but stained with monoclonal ANTI-FLAG M2-FITC (F4049, Sigma-Aldrich) at 10 μg ml⁻¹ in TBS at room temperature for 1 h against their FLAG-tags and MitoTracker Red CMXRos (M7512, Invitrogen) for mitochondria localization. The cells were mounted using ProLong Diamond mounting medium with DAPI (P36966, Invitrogen) and imaged with laser confocal microscope (Confocal Microscope Leica STELLARIS 5).

Structure-guided mutagenesis

To generate the mutant plasmids for FLVCR1 and FLVCR2, an overlapping PCR approach was used. The mutated cDNA of FLVCR1 or FLVCR2 was cloned into pcDNA3.1 for overexpression. The mutations were validated by Sanger sequencing. To test the transport activity of these mutants, the mutant plasmid was either cotransfected with CHKA for choline transport assay or ETNK1 for ethanolamine transport assay. After 24 h of post-transfections, cells were washed with DMEM and incubated with DMEM containing 20 μM [³H]choline or 2.5 μM [¹⁴C]ethanolamine for FLVCR1 mutants and 100 μM [³H]choline or 2.5 μM [¹⁴C]ethanolamine for FLVCR2 mutants. The assays were stopped after 1 h of incubation at 37 °C. Radioactive signal of each mutant was normalized to the total protein levels.

Choline export assay

To examine the export function, FLVCR1 and FLVCR2 plasmids were expressed in HEK293 cells without cotransfection with CHKA or ETNK1. The cells were then incubated with 200 μM [³H]choline or 100 μM [¹⁴C]ethanolamine for 2 h to prepack the cells with the ligand. Subsequently, the cells were washed to remove the ligands left over in the medium and incubated with choline/ethanolamine-free medium for 1 h at 37 °C for the release of prepacked ligand. The cells were washed and collected for quantification of radioactive signals. Samples after 2 h of incubation with the radioactive ligand were collected to determine the levels of radioactive levels before the release and used for control.

Metabolomic analysis

Adult livers (aged 3–6 months) from controls (FLVCR1^f/f and FLVCR1^f/+-Mx1-Cre) and conditional FLVCR1-knockout (FLVCR1^f/f-Mx1-Cre) mice were used for metabolomic analysis. All mice were maintained under specific pathogen-free conditions with free access to food and water with 12 h dark–light cycle. Briefly, the mice were perfused with PBS to remove blood before organ collection. Liver samples were snap-frozen before being shipped for metabolomics by Metabolon. The levels of metabolites were expressed as relative amount. Studies involving mice were reviewed and approved by the University of Washington Institutional Animal Care and Use Committee under protocol number 2001-13.

Production and purification of the human FLVCR1 and FLVCR2

For protein production, the Flp-In T-REx293-FLVCR1 and Flp-In T-REx293-FLVCR2 cell lines were cultured in roller bottles (Greiner Bio-One) in growth media containing 100 μg ml⁻¹ of hygromycin B for 14 d under the above-mentioned conditions. Gene expression was induced at 100% confluence by adding a final concentration of 2 μg ml⁻¹ of doxycycline hydrochloride. After 72 h, cells were harvested with Accutase solution (Sigma-Aldrich) and stored at −80 °C until further use. Harvested cells were suspended in cold lysis buffer containing 25 mM Tris pH 7.4, 150 mM NaCl and 0.1 g ml⁻¹ of SigmaFast ethylenediaminetetraacetic acid (EDTA)-free protease inhibitor (Sigma-Aldrich) and disrupted by stirring under high-pressure nitrogen atmosphere (750 MPa) for 45 min at 4 °C in a cell-disruption vessel (Parr Instrument). The cell lysate was centrifuged at 8,000g at 4 °C for 15 min. Subsequently, the low-velocity supernatant was centrifuged at 220,000g at 4 °C for 60 min. Pelleted membranes were resuspended and stored in a storage buffer containing 25 mM Tris pH 7.4, 150 mM NaCl, 10% glycerol (v/v) and 0.1 g ml⁻¹ of SigmaFast EDTA-free protease inhibitor (Sigma-Aldrich).

All purification steps of both FLVCRs were performed at 4 °C. Isolated membranes were solubilized with 1% (w/v) lauryl maltose neopentyl glycol (LMNG; GLYCON Biochemicals) with gentle stirring for 1 h. The insoluble membrane fraction was removed through ultracentrifugation at 220,000g for 1 h. Subsequently, the supernatant was incubated with ANTI-FLAG M2 Affinity Gel resin (Millipore) for 1 h. The resin was pre-equilibrated with a buffer containing 50 mM Tris pH 7.4, 150 mM NaCl and 0.02% LMNG (w/v). The washing step was performed using 20 column volumes of wash buffer (50 mM Tris pH 7.4, 150 mM NaCl, 5% (v/v) glycerol and 0.02% LMNG). The protein was eluted from the M2 resin with 10 column volumes of the same buffer supplemented with 4 mM FLAG Peptide (Millipore). The eluted sample was concentrated and subjected to a Superdex 200 Increase 10/300 column (Cytiva) equilibrated with size exclusion chromatography buffer (50 mM Tris pH 7.4, 150 mM NaCl and 0.001% (w/v) LMNG). Peak fractions were pooled, concentrated to 1.5 mg ml⁻¹ using an Amicon 50 kDa cut-off concentrator (Millipore) and stored for further analysis.

Immunoblotting

Affinity-purified proteins were subjected to SDS–polyacrylamide gel electrophoresis and immunoblotting. FLAG-tagged FLVCR1 and FLVCR2 were detected using anti-FLAG (F3165, Sigma-Aldrich) at 1:1,000 dilution. Anti-mouse IgG conjugated with alkaline phosphatase (A9316, Sigma-Aldrich) was used as secondary antibody at 1:5,000 dilution. Native FLVCR1 and FLVCR2 proteins were detected by polyclonal FLVCR1 and FLVCR2 antibodies raised in-house at 1:1,000 dilution. GAPDH antibody (sc-32233, Santa Cruz) was used as loading control at 1:4,000 dilution. IRDye 680LT (926-32212, Li-COR Biosciences) was used as secondary antibody for detection.

Tryptophan fluorescence measurement

Tryptophan fluorescence measurements were carried out using Prometheus Panta (NanoTemper Technologies). Purified protein samples were diluted with dilution buffer containing 50 mM HEPES pH 7.4, 150 mM NaCl and 0.001% (w/v) LMNG to 1 μM. Buffers with different concentrations of choline or betaine were prepared by serial dilutions in dilution buffer containing 4 mM of the compounds. The protein samples were mixed with an equal volume of dilution buffer or the compound-containing buffer with a final protein concentration of 0.5 μM and then incubated at room temperature for 15 min. A volume of 10 μl of mixed solution was used per Prometheus high-sensitivity capillary (NanoTemper Technologies). Recorded F₃₅₀/F₃₃₀ was analysed by using Python libraries including pandas, numpy, scipy and seaborn in Visual Studio Code (Microsoft). Three technical replicates were recorded for data analysis. The custom python code used for data analysis is publicly available through https://doi.org/10.5281/zenodo.10938397.

Cryo-EM sample preparation

To collect cryo-EM data of FLVCR1 and FLVCR2 in different sample conditions, different combinations of FLVCR proteins and putative substrate molecules were prepared. For both as-isolated samples of FLVCRs, the protein concentration was adjusted to approximately 1.5 mg ml⁻¹ and subjected to plunge freezing. For samples supplemented with choline, purified proteins were adjusted to 1.5 mg ml⁻¹ and choline was added at a final concentration of 1 mM. For FLVCR1 samples supplemented with ethanolamine, purified proteins were adjusted to 1.5 mg ml⁻¹ and ethanolamine was added at a final concentration of 1 mM. The samples were incubated for 10 min at room temperature before plunge freezing. Identical plunge freezing conditions were applied for all samples: 300 mesh R1.2/1.3 copper grids (Quantifoil) were washed in chloroform and subsequently glow-discharged with a PELCO easiGlow device at 15 mA for 90 s. A volume of 4 µl sample was applied to a grid and blotting was performed for 4 s at 4 °C, 100% relative humidity with nominal blot force 20 immediately before freezing in liquid ethane, using a Vitrobot Mark IV device (Thermo Scientific).

Cryo-EM image recording

For each cryo-EM sample, a dataset was recorded in energy-filtered transmission electron microscopy mode using either a Titan Krios G3i or a Krios G4 microscope (Thermo Scientific), both operated at 300 kV. Electron-optical alignments were adjusted with EPU software 3.0–3.4 (Thermo Scientific). Images were recorded using automation strategies of EPU 3.0–3.4 in electron counting mode with either a Gatan K3 (installed on Krios G3i) or a Falcon4 (installed on Krios G4) direct electron detector. For Gatan K3 detector, a nominal magnification of 105,000, corresponding to a calibrated pixel size of 0.837 Å was used and dose fractionated videos (80 frames) were recorded at an electron flux of approximately 15 e⁻ pixel⁻¹ s⁻¹ for 4 s, corresponding to a total dose of about 80 e⁻ A⁻². For Falcon4 detector, a nominal magnification of 215,000, corresponding to a calibrated pixel size 0.573 Å was used, dose fractionated videos were recorded in electron-event representation format at an electron flux of approximately 4 e⁻ pixel⁻¹ s⁻¹ for 5 s, corresponding to a total dose of about 50 e⁻ A⁻². Images were recorded between −1.1 and −2.0 µm nominal defocus. Data collection quality was monitored through EPU v.3.0-3.4 and CryoSparc Live (v.3.0 and 4.0)³³.

Cryo-EM image processing

For each acquired dataset, the same cryo-EM image processing approach was applied: MotionCor2 was used to correct for beam-induced motion and to generate dose-weighted images³⁴. Gctf was used to determine the contrast transfer function (CTF) parameters and perform correction steps³⁵. Images with estimated poor resolution (more than 4 Å) and severe astigmatism (more than 400 Å) were removed at this step. Particles were picked by TOPAZ and used for all further processing steps³⁶. Two-dimensional classification, initial model generation, three-dimensional (3D) classification, CTF refinement, Bayesian polishing, 3D sorting and final map reconstructions were performed using RELION (v.3.1 and 4.0) or cryoSPARC (v.3.0 and 4.0)^33,37,38. In the data processing pipeline, 3D autorefine jobs were conducted following each 3D classification or 3D sorting round for all resulted classes, to carefully assess the resulting density maps for quality and resolution through both metrics and visual inspection. Data processing was only proceeded with those maps that seemed promising for further refinement stages. Fourier shell correlation (FSC) curves and local-resolution estimation were generated in RELION or cryoSPARC for individual final maps. A schematic overview of our processing workflow and a summary of map qualities are shown and Supplementary Figs. 3–5.

Model building and geometry refinement

The first atomic models of FLVCR1 and FLVCR2 were built into the respective electron microscopy density maps of the as-isolated state in Coot (v0.8) or ISOLDE within ChimeraX (v.1.5 and 1.6)^39,40,41, using the AlphaFold predicted structures as initial templates⁴². After manual backbone tracing and docking of side chains, real-space refinement in Phenix was performed (v.1.18)⁴³. Refinement results were manually inspected and corrected if required. These models were used as templates to build all subsequent atomic models. The finalized models were validated by MolProbity implemented in Phenix⁴⁴. Map-to-model cross-validation was performed in Phenix (v.1.18). FSC_0.5 was used as cut-off to define resolution. The comprehensive information on the Cryo-EM data collection, refinement and validation statistics is shown in Extended Data Table 1. The finalized models of the two FLVCR proteins in different states were visualized using ChimeraX and used as starting structures for molecular dynamics simulations.

Molecular dynamics simulations

All molecular dynamics simulations were performed using the GROMACS 2022.4 (ref. ⁴⁵) software. The protein structures were embedded in a lipid bilayer with 75% POPE and 25% POPG with CHARMM-GUI⁴⁶ and solvated in TIP3P water with 150 mM NaCl. The CHARMM36m force field⁴⁷ was used with the improved WYF parameters for cation–π interactions, in particular of the choline and ethanolamine ligands⁴⁸. The systems were minimized for 5,000 steepest-descent steps and equilibrated for 250 ps of molecular dynamics in an NVT ensemble and for 1.625 ns in an NPT ensemble. Position restraints of 4,000 and 2,000 kJ mol⁻¹ nm⁻² in the backbone and side chain heavy atoms, respectively, were gradually released during equilibration. The z-positions of membrane phosphates, as well as lipid dihedrals, were initially restrained with force constants of 1,000 kJ mol⁻¹ nm⁻², which were gradually released during equilibration. The initial time step of 1 fs was increased to 2 fs during NPT equilibration. Long-range electrostatic interactions were treated with particle-mesh Ewald⁴⁹ with a real-space cut-off of 1.2 nm. Van-der-Waals interactions were cut-off beyond a distance of 1.2 nm. The LINCS algorithm⁵⁰ was used to constrain the bonds involving hydrogen atoms. During equilibration, a constant temperature of 310 K was maintained with the Berendsen thermostat⁵¹, using a coupling constant of 1 ps. Constant pressure of 1 bar was established with a semi-isotropic Berendsen barostat and a coupling constant of 5 ps. In the production runs, a Nosé–Hoover thermostat⁵² and a Parrinello–Rahman barostat were used⁵³.

We used our cryo-EM structures as initial models for simulations of as-isolated and choline-bound inward-facing FLVCR1, as-isolated and choline-bound inward-facing FLVCR2 and as-isolated outward-facing FLVCR2. We set the protonation state of each residue as predicted for pH 7.0 using the PROPKA server⁵⁴. An initial structure of choline-bound outward-facing FLVCR2 was generated by aligning the as-isolated outward-facing FLVCR2 to the choline-bound inward-facing FLVCR2 and maintaining choline in the cavity. In choline entry simulations, the as-isolated structures were used with 380 mM choline in solution. For simulations of ethanolamine-bound FLVCR1, the choline within the cavity of the cryo-EM structure was replaced by this ligand. Simulations with deprotonated ethanolamine were performed as well and results are included in Supplementary Figs. 11 and 12. Choline and ethanolamine release simulations were interrupted after ligand exit from the cavity and hence have variable duration. For all other systems, each replica was run for 1 µs. A summary of all simulations performed in this study is provided in Supplementary Table 3 (table with technical information). Time-resolved distance calculations for all replicas not included in the main figures are shown in Supplementary Fig. 11. Minimum atom-pair distances were calculated as the minimum distance over all pairs of atoms in two stated groups, including hydrogens (for example, ligand and certain defined side chains).

Alanine substitution mutations were introduced using PyMol⁵⁵ and simulated with identical parameters as those applied in the corresponding wild-type simulations. In FLVCR1 and FLVCR2, alanine mutations were introduced in the cavity residues W125^FLVCR1 and W102^FLVCR2, respectively.

For the MM/PBSA calculations, we used gmx_MMPBSA⁵⁶ with dielectric constants of 7.0, 80.0 and 4.0 for the membrane, solvent and protein, respectively, and the default surface tension of 0.0072 kcal mol⁻¹ nm⁻². We estimated entropies using the interaction entropy method⁵⁷. The contributions of W125^FLVCR1 and W102^FLVCR2 to the binding energy were estimated by alanine scanning.

Visual molecular dynamics⁵⁸ and MDAnalysis⁵⁹ were used to visualize and analyse the trajectories, respectively. An assessment of the reliability and reproducibility of our simulations is provided in Supplementary Table 4.

Interior tunnels and cavities

Tunnels and cavities were mapped with MOLE v.2.5 (ref. ⁶⁰) with a bottleneck radius of 1.2 Å, bottleneck tolerance 3 Å, origin radius 5 Å, surface radius 10 Å, probe radius 5 Å and an interior threshold of 1.1 Å.

We calculated the volume of the cavity using CASTp⁶¹ with a bottleneck radius of 1.4 Å. Residues 297–320 and 512–516 were removed from the FLVCR1 model to avoid the misattribution of the volume between internal loops to the cavity volume. Analogously, residues 272–296 and 487–502 were not included in the cavity volume calculation of FLVCR2.

Sequence alignments

Multiple sequence alignments of FLVCR1 and FLVCR2 from Homo sapiens, Felis catus, Mus musculus and Sus scrofa were performed using Clustal Omega⁶².

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.