Specific features and assembly of the plant mitochondrial complex I revealed by cryo-EM.

Mitochondrial complex I purification Cauliflower ( Brassica oleracea var. botrytis) mitochondria were purified as previously described 12 . Quickly, fresh cauliflower inflorescence tissue was blended in extraction buffer containing 0.3 M mannitol, 30 mM sodium pyrophosphate (10.H 2 O), 0.5% BSA, 0.8% (w/v) polyvinylpyrrolidone-25, 2 mM beta-mercaptoethanol, 1 mM EDTA, 20 mM ascorbate and 5 mM cysteine, pH 7.5. Lysate was filtered and clarified by centrifugation at 1500 × g , 10 min at 4 °C. Supernatant was kept and centrifuged at 18,000 × g , 15 min at 4 °C. Organelle pellet was re-suspended in wash buffer (0.3 M mannitol and 10 mM phosphate buffer, 1 mM EDTA, pH 7.5) and the precedent centrifugations were repeated once. The resulting organelle pellet was re-suspended in wash buffer and loaded on a single-step 30% Percoll gradient (in wash buffer without EDTA) and run for 1h30 at 40,000 × g . Mitochondria were retrieved, washed two times and flash frozen in liquid nitrogen. For complex I purification, mitochondria were re-suspended in Lysis buffer (20 mM HEPES-KOH, pH 7.6, 100 mM KCl, 1 mM DTT, 2% n-β-DDM, 1 mM EDTA, supplemented with proteases inhibitors (Complete)) incubated for 15 min in 4 °C. Lysate was clarified by centrifugation at 30,000 × g , 20 min at 4 °C. The supernatant was loaded on a 10–50% sucrose gradient buffer (20 mM HEPES-KOH, pH 7.6, 50 mM KCl, 1 mM DTT, 0.2% n-β-DDM, 1 mM EDTA, supplemented with proteases inhibitors (Complete)) and run for 16 h at 75,000 × g . The fraction corresponding to complex I was collected, pelleted and re-suspended in final resuspension buffer (same as sucrose gradient buffer without sucrose and only 0.1% n-β-DDM).

Grid preparation

Four microliter of the samples at a concentration of 1 µg/µl was applied onto Quantifoil R2/2 300-mesh holey carbon grid, coated with thin home-made continuous carbon film and glow-discharged (3 mA for 20 s). The sample was incubated on the grid for 30 s and then blotted with filter paper for 2.5 s in a temperature and humidity controlled Vitrobot Mark IV (T = 4 °C, humidity 100%, blot force 5) followed by vitrification in liquid ethane.

Single particle cryo-electron microscopy data collection

Data collection was performed on a Talos Arctica instrument (Thermofisher Company) at 200 kV using the SerialEM software for automated data acquisition. Data were collected at a nominal underfocus of −0.5 to −2.5 µm at a magnification of ×36,000 yielding a pixel size of 1.11 Å. Micrographs were recorded as movie stack on a K2 direct electron detector (GATAN Company), each movie stack were fractionated into 65 frames for a total exposure of 6.5 s corresponding to an electron dose of 45 ē/Å2.

Electron microscopy image processing

Drift and gain correction and dose weighting were performed using MotionCorr2 34 . A dose weighted average image of the whole stack was used to determine the contrast transfer function with the software Gctf 35 . The following process has been achieved using RELION 3.0 14 . Particles were picked using the general model of crYOLO 1.5 36 . Output box files from crYOLO were imported in RELION and particles were extracted. After reference-free 2D classification, 650,844 particles were extracted with a box size of 360 pixels and binned threefold, resulting in a 120 pixels box size and used for 3D classification into 6 classes (Supplementary Fig. 1 ). CryoSPARC 37 generated ab-initio cryo-EM map, low-pass filtered to 30 Å, was used as an initial reference for 3D classification. One subclass depicting high-resolution features have been selected for the mature complex I refinement with 65,018 particles. After Bayesian polishing a focused refinement has been performed using mask excluding the P D module yielding, respectively, 3.7 and 3.47 Å resolution. A second subclass containing the maturation factor GLDH composed of 155,644 particles was selected for further focused 3D classification, using a loose soft mask on the GLDH area. The 36,513 particles of GLDH enriched subclass were extracted for refinement and reached 3.8 Å. Determination of the local resolution of the final density map was performed using ResMap 38 .

Structure building and model refinement

The atomic model of the plant mitochondrial complex I was built into the high-resolution maps using Coot, Phenix, and Chimera. Atomic models from Y.lipolytica (6RFR) and M.musculus (6G2J) were used as starting points for protein identification and modelisation. Similarly to Waltz & Soufari 2019 12 , as the genome of cauliflower is not sequenced, and the closest fully sequenced member of the family ( Brassica oleracea subsp. oleracea) is poorly annotated, to facilitate comprehension and analysis we positioned Arabidopsis proteins in the cauliflower map. Still, protein sequence identities between members of the Brassicaceae family are higher than 90%, thus facilitating proteomics identification of cauliflower proteins. The online SWISS-MODEL service was used to generate initial models for bacterial and mitochondria conserved r-proteins. Models were then rigid body fitted to the density in Chimera 39 and all subsequent modeling was done in Coot 40 . Extensions were built has polyalanine and mutated to the adequate sequences. A combination of regularization and real-space refine was performed in Coot for each proteins. The global atomic model was then subjected to real space refinement cycles using phenix.real_space_refine PHENIX 41 function, during which protein secondary structure, Ramachandran and side chain rotamer restraints. Several rounds of refinement (manual in Coot and automated using the phenix.real_space_refine ) were performed to obtain the final models, which were validated using the built-in validation tool of PHENIX, based on MolProbity. For the assembly intermediate, the model was built with the mature complex as template. Refinement and validation statistics were summarized in Supplementary Table 2 .

Proteomic and statistical analyses of mitochondrial complex I composition

Mass spectrometry analyses of the complex

I fractions were performed at the Strasbourg-Esplanade proteomic platform and performed as previously 42 . In brief, proteins were trypsin digested, mass spectrometry analyses and quantitative proteomics were carried out by nanoLC-ESI-MS/MS analysis on a QExactive+ (Thermo) mass spectrometer. Data were searched against the TAIR A. thaliana database with a target-decoy strategy (release TAIRv10, 27281 forward protein sequences), and home-made Brassica oleracea database extracted from UniProtKB (Swissprot+TrEMBL) including Brassica sub-taxonomy (release 2017_10, 59050 forward protein sequences). Proteins were validated respecting FDR < 1% (False Discovery Rate) and quantitative label-free analysis was performed through in-house bioinformatics pipelines.

Figure preparation

Figures featuring cryo-EM densities, as well as atomic models were visualized with UCSF ChimeraX 43 and Chimera 39 . Reporting summary Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Supplementary information Supplementary Information Peer Review File Descriptions of Additional Supplementary Files Supplementary Data 1 Reporting Summary