Image processing

The 5576 movies of mLST8–CCT complex were aligned using MotionCorr2⁵⁷ program as part of the Scipion processing workflow⁵⁸. The MotionCorr2 output was subjected to CTF determination using CTFFIND4⁵⁹. Totally, 1,769,600 particles were automatically picked with Xmipp⁶⁰ and extracted with a downsampling factor of 3 (4.08 Å/pixel, 68 pixel box size). All the image processing steps were carried out without any symmetry imposition. A first 2D classification using Relion 2.0⁶¹ (unless otherwise stated Relion 2.0 was used in all subsequent steps) was performed to exclude bad particles and ice contamination. Some of the best 2D classes were used as a template to generate an initial model using both EMAN⁶² and RANSAC⁶³. In both cases a cylinder with the general dimensions similar to the CCT structure was obtained, which was subsequently used for the iteration process. One of the models was low-pass filtered to 60 Å and used for a 3D classification of 1,197,358 particles contained in the best 2D classes. The 3D classes that showed well-defined CCT features and a mass inside the cavity (860,453 particles) were subjected to refinement using 3D auto-refine, which generated a 9 Å map. The particles used in this refinement were re-extracted from the 1.36 Å/pixel micrograph to continue the processing with the original data. A new 3D classification was performed in which a mask was applied around the mass attributed to the substrate in order to favor the classification to the substrate contribution and prevent the CCT predominance. Those particles with a better-defined mLST8 mass (452,000) were finally subjected to auto-refine and a map was obtained with a final resolution of 4.35 Å. Subsequently, a postprocessing was performed masking the previous map and enhancing the high frequencies and the resolution improved to 4.0 Å, as estimated using the gold standard FSC criterion at 0.143⁶⁴. Local resolution in the 3D structure of the mLST8–CCT complex was estimated using MonoRes⁶⁵ from Xmipp package and ResMap⁶⁶. The statistical information is listed in Supplementary Table ¹.

The 3D reconstruction of substrate-free CCT was carried out following a similar procedure. A total of 2293 movies were aligned with MotionCor2 and CTF corrected. A total of 504,060 particles were automatically selected and extracted with a downsampling factor of 3 (3.18 Å/pixel, 80 pixel box size). Particles were 2D classified using Relion 2.0 and the best classes (139,819 particles) were subjected to 3D classification. Classes with the best structural features of CCT were further refined, the particles were re-extracted from the original micrographs (1.06 Å/pixel) and after 3D refinement, a final map at 7.5 Å was obtained (gold standard FSC).