Cryo-EM reconstruction and functional characterization of VMAT2-tetrabenazine complex.

a, Predicted structural elements of VMAT2. The neurotransmitter substrate is bound at the central site (yellow, triangle). The red and blue triangles depict the pseudo two-fold symmetric repeat comprised of TM1-6 and 7–12, respectively. A disulfide bond (purple line) is predicted between extracellular loop 1 (EL1) and EL4, N-linked glycosylation sites in EL1 are shown as red ‘Y’ shapes. b, Occluded map of VMAT2-tetrabenazine complex (3.3 Å resolution, contour level 0.076). The mVenus and GFP-Nb fiducial is not shown for clarity. c, Left panel, plots of [3H]dihydrotetrabenazine saturation binding to wild-type VMAT2 (black, circles) and chimera (red, squares). Symbols show the mean derived from n=3 technical replicates. Error bars show the s.e.m. Right panel, graphs of competition binding of 3H-dihydrotetrabenazine with unlabeled reserpine, error bars show the s.e.m. d, Plots of transport into vesicles using 1 and 10 µM 3H-serotonin for wild-type VMAT2 (grey bars) and chimera (red bars). The bars show the means and points show the value for each technical replicate. Error bars show the s.e.m.

VMAT2 conformation and residues involved in gating.

a, ‘slice view” through an electrostatic surface representation of the VMAT2-tetrabenazine (TBZ) complex. TBZ is shown in light green sticks. b, Cartoon representation showing the extracellular gating residues and the intracellular gating residues in pink sticks. c, Variations of gating residue poses captured in molecular dynamics simulations. d, Cartoon representation of polar network ‘one’. Blue and red sticks denote residues in the N– and C-terminal half respectively. e, Polar network ‘two’ (top) and ‘three’ (bottom), residues colored in red.

Tetrabenazine recognition and binding.

a, Chemical structure of tetrabenazine (TBZ). The blue dotted circle indicates the position of the hydroxyl group in dihydrotetrabenazine. b, Binding site of TBZ, residues which are involved in binding are shown in tan sticks. TBZ is shown in light green sticks and the associated density in dark grey mesh. c, 2D cartoon of the TBZ binding site showing only highlighted residues. Green, red, and blue indicate hydrophobic, negative, or positively charged properties of the side chains. d, Plots of 3H-dihydrotetrabenazine saturation binding to wild-type (black line) and mutant VMAT2. F135A (light blue), R189A (red), V232L (dark green), E312Q (salmon), W318A (light purple), F429A (dark blue), and Y433A (light green). Symbols show the mean derived from n=3 technical replicates. Error bars show the s.e.m.

Mechanism of tetrabenazine inhibition, gating mechanisms, and polar networks.

a, Cartoon depicting tetrabenazine binding to VMAT2. Tetrabenazine (green hexagon) binds to the luminal-facing state and induces conformational change to a high-affinity occluded conformation which is the resolved cryo-EM structure reported in this work. b, Cartoon model of the VMAT2 – tetrabenazine complex highlighting significant features including both cytosolic (slashes) and luminal gates (triangle), the three polar networks (numbered circles) and relative location of the tetrabenazine binding site (green hexagon).

Biochemical characterization, construct design, and sequence conservation of VMAT2.

a, SDS-PAGE gel showing purified VMAT2 chimera which migrates as a ∼75 kDa species. b, Fluorescence-detection-size exclusion chromatography of purified VMAT2 chimera. The yellow trace is the fluorescence of mVenus and the black trace is of Trp. c, Time course accumulation of serotonin in vesicles using 1 µM 3H-serotonin for wild-type (black trace) and chimera (red trace). d, Sequence of VMAT2 colored by sequence variation across 150 VMAT2 sequences from different species, using Consurf server91. The position of mVenus and the GFP-Nb are indicated with arrows. Residues in EL1 that are not resolved in the cryo-EM map are also noted. e, VMAT2 model colored by sequence conservation.

Cryo-EM data processing of the VMAT2-tetrabenazine complex.

A representative micrograph (defocus –1.3 µm) is shown (scale bar equals 80 nm). The workflow depicts the data processing scheme used to reconstruct VMAT2. Two datasets were collected comprising 7,742 and 17,133 micrographs respectively. Movies were corrected for drift using patch motion correction and resultant micrographs were used to estimate defocus and pick particles. Blob picking followed by template picking was utilized to select approximately 5 million particles from each dataset. 2D classification was used to sort particles and the sorted particles were subjected to ab-initio reconstructions to obtain initial reference. Next, all of the particles picks from each dataset were subjected to multiple rounds of heterogeneous classification/refinement with the ab-initio VMAT2 map and two ‘decoy’ classes (yellow, a spherical blob and red, empty detergent micelle) starting with a box size of 128 pixels, followed by subsequent rounds of classification at box sizes of 256 and 384 pixels (full box size). This resulted in approximately 500k particles after combining both datasets. Particles underwent non-uniform refinement and further rounds of 2D classification and heterogeneous refinement to select particles with higher resolution features. Local refinements with a mask that excluded the detergent micelle further improved the resolution of the reconstruction. Non-uniform refinements, local refinement and CTF refinements with a mask focused on VMAT2 further improved the resolution. Bayesian polishing was utilized to correct local particle motion followed by further rounds of 2D classification, heterogeneous refinements, and CTF refinement. The final stack of 92k particles was then subjected to local refinement to produce the final unsharpened map. DeepEMhancer was used to locally sharpen the map for interpretation.

Cryo-EM maps and interpretation of VMAT2 reconstruction.

a, Cryo-EM density colored by local resolution estimation. b, FSC curves for cross-validation, the unmasked map (blue), loose mask (green), tight mask (red) and the reported corrected (purple) curves. The dotted line indicates an FSC value of 0.143. c, FSC curves for model versus half map 1 (working, red), half map 2 (free, blue) and model versus final map (black). d, Angular distribution of particles used in the final reconstruction. e, Cryo-EM density segments of TM1 to TM12.

Tetrabenazine docking and molecular dynamics simulations.

a, Time evolution of Cα root mean square deviations (RMSD) from the cryo-EM-resolved VMAT2 structure; and b, computed RMSD of TBZ, in three different runs. c, Histogram of TBZ binding affinities, summarized over all three runs. Binding affinities were calculated using PRODIGY-LIG applied to 800 evenly collected snapshots between 20 ns to 100 ns from each run. d, The TBZ binding poses and variations of W318 captured in the MD simulation runs 1-3, with a snapshot taken every 4 ns. The ligand conformations are shown in cyan sticks with blue stick illustrating cryo-EM resolved binding pose. The variations of W318 are displayed in purple sticks with dark purple showing the cryo-EM-resolved orientation. Docking simulations identified e, the most favorable (–9.7 kcal/mol) binding pose of TBZ, captured by run2. f, The TBZ pose (–9.3 kcal/mol) that closely resembles the cryo-EM-resolved structure (captured by runs 1,3). The RMSD from the resolved TBZ pose is 3.0 to 0.4 Å in e and f, respectively.

Point mutants in tetrabenazine binding site.

a, Binding site showing the positions of L37 and V232 which are a phenylalanine and a leucine in VMAT1 respectively. b, Plots of binding of 60 nM of [3H]dihydrotetrabenazine. The bars show the means and points show the value for each technical replicate. Error bars show the s.e.m. c, Sequence alignment of VMAT1, VMAT2, VAChT, and VPAT. The positions of mutated residues are shown boxed and in red. The positions of human variants are shown in blue boxes. d, Human variants of VMAT2, P316A, P237H and P387L localize to EL4 and the luminal ends of TM5 and TM9 respectively.

Comparison of the VMAT2-TBZ structure with the predicted Alphafold structure and with other MFS transporters.

a, Comparison of the cryo-EM structure (tan) vs. Alphafold (grey). The position of TMHs 1, 7, 8, 11 and EL4 in the cryo-EM structure show the most substantial differences and are shown for clarity. b, Comparison with the outward-open VGLUT2 structure (PDB code 8SBE) shown in blue. c, Comparison with the inward-open GLUT4 structure (7WSM) shown in magenta.

Docking-predicted binding poses of dopamine and serotonin.

a, Docking of dopamine and b, serotonin to the cryo-EM-resolved VMAT2. The most energetically favorable pose is shown; residues within 4 Å of the ligand are showed in sticks. Dopamine and serotonin are displayed in violet and cyan as van der waals (VDW) surfaces. The amine group of dopamine and serotonin is in close contact with E312 and their respective hydroxyl group(s) interact with R189.