Cryo-EM structures of human TMEM120A and TMEM120B

TMEM120A (Transmembrane protein 120A) was recently identified as a mechanical pain sensing ion channel named as TACAN, while its homologue TMEM120B has no mechanosensing property1. Here, we report the cryo-EM structures of both human TMEM120A and TMEM120B. The two structures share the same dimeric assembly, mediated by extensive interactions through the transmembrane domain (TMD) and the N-terminal coiled coil domain (CCD). However, the nearly identical structures cannot provide clues for the difference in mechanosensing between TMEM120A and TMEM120B. Although TMEM120A could mediate conducting currents in a bilayer system, it does not mediate mechanical-induced currents in a heterologous expression system, suggesting TMEM120A is unlikely a mechanosensing channel. Instead, the TMDs of TMEM120A and TMEM120B resemble the structure of a fatty acid elongase, ELOVL7, indicating their potential role of an enzyme in lipid metabolism.


Protein sample preparation
The codon optimized gene of human TMEM120A and TMEM120B were synthesized and subcloned into pCAG vector with a N-terminal FLAG and mCherry tag. HEK shaker. The insoluble fraction was precipitated by ultracentrifugation at 255,700 g for 1 h, and the supernatant was applied to anti-Flag G1 affinity resin (GenScript) by gravity. The resin was washed sequentially to remove unbound proteins with wash buffer containing 25 mM Tris-HCl (pH 8.0), 400 mM NaCl, 0.015% (w/v) glyco diosgenin (GDN, Anatrace), and protease inhibitor cocktail. The target protein was eluted by the elution buffer containing 25 mM Tris-HCl (pH 8.0), 400 mM NaCl, 0.015% (w/v) GDN, and protease inhibitor cocktail plus 250 μg/ml FLAG peptide (GenScript). The eluent was then concentrated using a 30-kDa cut-off Centricon (Millipore) and further purified by size exclusion chromatography (Superose 6 Increase 10/300 GL column, GE Healthcare) in 25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 0.015% (w/v) GDN, and protease inhibitor cocktail. Peak fractions containing the target protein were collected and concentrated to about 6.5 mg/mL for cryo-EM analysis. The wild type and mutant proteins for electrophysiological experiments in lipid bilayer were prepared in a similar procedure except for the detergent used in the gel filtration was replaced by 0.05% Lauryl dimethylamine-N-Oxide (LDAO, Anatrace) and 0.01% CHS.

Bilayer recording
Electrophysiological recordings of wild type TMEM120A, TMEM120B, and TMEM120A mutants were performed on the planar bilayer lipid membrane formed in a cup (1 mL) and a chamber (1 mL) (Warner instrument), with the buffer conditions of cis-: 500 mM NaCl, 10 mM HEPES, pH 7.5 and trans-: 100 mM NaCl, 10 mM HEPES, pH 7.5. E.coli Polar Lipid Extract (AVANTI) was dissolved in decane with a final concentration of 25 mg/mL. The aperture (150 μm) of the cup was precoated with 0.8 μL lipid solution. After the bilayer lipid membrane was formed, 1 μL of purified protein (about 6 mg/mL) was added into the cis-chamber. Current trace was recorded by HEKA EPC 10 USB (HEKA) and Axonpatch 200B (Axon Instruments) with 10 kHz sampling frequency.

Cells and transfections
HEK293T-P1KO cells used for whole-cell electrophysiology is a gift from Dr.
Bailong Xiao and HEK293T-P1KO cells used for cell-attached electrophysiology were generated using CRISPR-Cas9 technique as previously described 1,2 . The

Whole cell electrophysiology
The whole-cell currents were recorded using an EPC-10 amplifier with Mechanical stimulation utilized a fire-polished glass pipette (tip diameter 3-4 mm) positioned at an angel of 80° relative to the cell being recorded as described 3 .
The probe was driven by Patchmaster-controlled piezoelectric crystal microstage (E625 LVPZT Controller/Amplifier; Physik Instrument). The probe velocity was 1 μm/ms during the upward and downward movement, and the stimulus was kept constant for 200 ms. A series of mechanical steps in 1μm increments was applied every 10 s and currents were recorded at a holding potential of -70 mV. inter-sweep duration of 30s. Currents were sampled at 20 kHz and filtered at 2 kHz or 10 kHz. Leak currents before mechanical stimulations were subtracted off-line from the current traces. All experiments were done at room temperature. All data was analyzed in Clampfit 11.2 and data was plotted using GraphPad Prism.

Cryo-EM sample preparation and data collection
0.1% Fos-Choline-8, Fluorinated (Anatrace) was added into the protein sample before grid preparation. Aliquots (3.5 μL) of purified TMEM120A or TMEM120B with concentration about 6.5 mg/ml were loaded onto glow-discharged holey carbon grids (Quantifoil Au R1.2/1.3) for cryo-EM data collection. The grids were blotted for 3.5 s and immersed in liquid ethane using Vitrobot (Mark IV, Thermo Fisher Scientific) in condition of 100 % humidity and 8 ℃. The imaging system comprised of Titan Krios operating at 300 kV, Gatan K3 Summit detector, and GIF Quantum energy filter with a 20 eV slit width. Movie stacks were automatically acquired in super-resolution mode (81,000× magnification) using AutoEMation 4 , with a defocus range from -0.5 μm to -3.0 μm. Each stack was exposed for 2.56 s with 0.08 s per frame, resulting in 32 frames and approximately 50 e -/Å 2 of total dose.

Cryo-EM data processing
For data processing of TMEM120A, 8168 movie stacks were motion-corrected with 2-fold binning by MotionCor2 5 . Patch-based CTF parameters of the dose-weighted micrographs (1.087 Å per pixel) were determined by cryoSPARC v2 6 , and around three million particles were automatically picked applying Topaz pipeline with a pretrained neural network model (ResNet8, 32 units) 7 . Three rounds of 2D classification enriched images of good classes, resulting in a total of 279,942 particles being selected with box size of 220 pixels. Following an Ab-Initio reconstruction for initial map generation, homogeneous and non-uniform refinement jobs 8 pushed the resolution to 4.3 Å and 3.8 Å, respectively, under C2 symmetry. To improve map quality, we exported the particles of cryoSPARC refinement jobs, and performed a 3D classification without orientation searching using RELION 3.0 (--skip-align flag) 9,10 .
The best class out of five that exhibits the most intact density with high resolution details, were selected for subsequent homogeneous and non-uniform refinements back to cryoSPARC v2, pushing the resolution to 3.7 Å. Local refinements with masks for TMD and CCD, further improved the local maps at 3.4 Å and 3.8 Å resolution, respectively.
The dataset of TMEM120B was processed in a similar way. Using 2D projections of TMEM120A reconstruction as templates, about 3.7 million particles were autopicked. After 3 rounds of 2D classifications, 560 K particles were selected for subsequent non-uniform refinement with C2 symmetry, using the lowpass filtered TMEM120A map (30 Å) as initial model. We then performed a 3D classification job with 5 classes using RELION 3.0 (--skip-align flag). One class of 138 K particles giving rise to the most intact reconstruction was transformed back to cryoSPARC for further refinement, yielding a final map at 4.0 Å resolution. Map resolutions were determined by the gold-standard Fourier shell correlation (FSC) 0.143 criterion using Phenix.mtriage 11 .

Model building and refinement
The structure of human TMEM120A was de novo built in Coot 12 . Sequence assignment was guided by bulky residues such as Phe, Tyr and Trp. Most of the side chains were assigned except for peripheral region in the CCD due to limited local resolution. The model of TEME120B was built using a homolog modeling template generated by the structure of TMEM120A. The model was then manually adjusted in Coot.
Subsequently, the models were refined against the corresponding maps by PHENIX 11 in real space (phenix.real_space_refine) with secondary structure, C2 symmetry and geometry restraints generated by ProSMART 13 . Overfitting of the overall models was monitored by refining the models in one of the two independent half maps from the gold-standard refinement approach and testing the refined model against the other map 14 . Statistics of 3D reconstruction and model refinement can be found in Table S1. The primary sequences of six TMEM120As from different species and human TMEM120B are aligned using Clustal W 17 . The invariant residues are shaded dark green and the conserved residues are colored orange. Secondary structural elements of human TMEM120A/B are presented above the sequence alignment. The residues that mediate interactions in TMEM120A/B dimer interface are indicated by blue triangles. Based on the alignment, a color scheme of sequence conservation was mapped on a TMEM120A protomer by ConSurf server 18