Mechanism of lid closure in the eukaryotic chaperonin TRiC/CCT.

TRiC purification and specimen preparation

We purified TRiC from bovine testes essentially as described 49 , but incorporated an ultracentrifugation step through 20−60% (w/v) sucrose cushions before chromatographic purification. Copurifying bound substrates were removed from TRiC by incubation with ATP before chromatography. Most TRiC particles in our preparation were active, based on the stoichiometry of actin binding, the efficiency of actin binding (60−80% folding after 45 min at 30 °C) and the efficiency of conversion to the closed state in the presence of ATP and AlF x (∼100% as assayed by proteolytic digestion; see also ref. 17 and Supplementary Fig. 1 ). TRiC samples were prepared for cryo-EM studies by dilution of purified TRiC to 1 mg ml −1 in buffer A (20 mM HEPES-KOH, pH 7.4, 100 mM potassium acetate, 5 mM magnesium acetate). Nucleotide-free samples were used immediately, whereas ATP-AlF x samples were incubated in the presence of 1 mM ATP (Sigma), 5 mM Al(NO 3 ) 3 , and 30 mM NaF for 1 h at 30 °C before being frozen. Samples were embedded in vitreous ice as follows. We placed a 3 μl aliquot of TRiC sample onto a washed, glow-discharged, 200-mesh R2−1 Quantifoil continuous carbon grid (Quantifoil Micro Tools GmbH, Jena, Germany). The grid was blotted and flash frozen in liquid ethane using a Vitrobot (FEI). We stored the grids in liquid nitrogen until they were imaged.

Image collection

We acquired ice-embedded images at an effective magnification of 83,100× on a US4000 charge-coupled device (CCD) camera (Gatan Inc.) using a JEOL2010F electron microscope (JEOL Inc.) with a field emission gun operated at 200 kV. A custom software package, JAMES 50 , was used to acquire focal pairs of images semiautomatically, at a range of 0.8−2.5 μm underfocus for the first image and a range of 2−5 μm underfocus for the second image. Image preprocessing An initial three-dimensional model was created using the EMAN program startcsym ( http://ncmi.bcm.tmc.edu/software/ ) 51 by manually selecting all the particles in a few representative micrographs. We genereated references from the initial model and used them to select the particles from the far-from-focus CCD frames using the automatic reference-based particle selection method in the EMAN software program batchboxer 51 . The far-from-focus boxes were aligned to their close-to-focus counterparts using the EMAN program boxer 51 . The automatically selected particles were visually screened using the EMAN program boxer 51 such that every nonover-lapping TRiC particle was selected. Contrast transfer function (CTF) parameters were initially estimated using the EMAN program ctfit 51 . Final CTF parameters were determined by manually fitting a theoretical one-dimensional power spectrum to a one-dimensional power spectrum calculated from the boxed particles for each micrograph using the EMAN program ctfit 51 . Using these procedures, 7,129 particles of ATP-AlF x TRiC and 13,287 particles of nucleotide-free TRiC were prepared for the three-dimensional reconstruction.

Three-dimensional reconstruction

We carried out particle orientations and three-dimensional reconstruction as described previously 52 , 53 . Briefly, the entire data set of raw particles was subjected to a multirefine-based procedure as described previously 52 , using two starting models created using the EMAN program startcsym 51 that had different amounts of applied noise. After this step, the class with the most particles assigned to it was further refined using the standard EMAN iterative reconstruction algorithm 53 . Briefly, the iterative reconstruction consists of reference-based classification of particles, class-averaging with CTF correction and three-dimensional model construction. The iterative reconstruction process continued until convergence was achieved, as assessed by observing the iteration-to-iteration Fourier shell correlation (FSC) curve. We made no reference to previously determined structural information during the reconstruction process; the only assumption made was an imposition of C8 symmetry. The density maps were scaled for visualization so that an isosurface threshold of 1 would correspond to the approximate molecular weight of the TRiC complex. Resolution of the reconstruction was estimated using the 0.5 FSC criterion between two reconstructions generated from even- and odd-numbered particles, respectively. All visualization was done with the UCSF Chimera molecular visualization package ( http://www.cgl.ucsf.edu/chimera/ ) 54 or the PyMOL Molecular Graphics System ( http://www.pymol.org/ ). Comparative modeling of TRiC closed (complexed with ATP-AlFx) state We carried out homology or comparative protein-structure modeling using the program MODELLER 32 . Each of the eight bovine TRiC subunits (TIGR: TCPA_BOS, TCPB_BOS, TCPD_BOS, TCPE_BOS, TCPG_BOS, TCPH_BOS, TCPQ_BOS, TCPZ_BOS) was modeled using the standard multiple-template modeling protocol implemented in the ‘model’ module. For the TRiC closed state, the thermosome structure from Thermoplasma acidophilum (PDB 1A6E) and the N and C termini from the thermosome structure found in Thermococcus strain KS-1 (PDB 1Q3Q) were used as templates. Both templates have a sequence identity to the bovine sequences ranging from 35% to 41%. For each of the subunits, 50 models were generated by MODELLER, and the best one was selected by the multivariate model assessment score 55 . Inaddition, for loop regions with relatively low sequence similarity to the templates, 200 models for each of those loop regions were generated by the ‘loop’ module, and the best one was chosen by the DOPE statistical potential score 55 , in combination with the higher docking score to the cryo-EM density map 56 obtained by the EMAN program foldhunter 57 . We built a complete model of the eight-subunit ring using the subunit permutation previously proposed 33 . Finally, we subjected this complete model to another round of MODELLER refinement to improve its stereochemistry. The refined model of the eight-subunit structure was docked into the TRiC-closed cryo-EM map using the rigid body docking program from EMAN, foldhunter 57 , with the similarity score of 0.987 calculated by the EMAN program fh-stat.py. Although each subunit in the model was well docked within the density map, at this level of resolution, a similarly good fit would be obtained with a different subunit arrangement.

TRiC open-state comparative modeling and domain docking

The thermosome structure from T. acidophilum (PDB 1A6D) and the N and C termini of the thermosome from the Thermococcus strain KS-1 (PDB 1Q3R) were chosen as templates. These PDB structures were used because they corresponded to the nucleotide-free state. Both of the templates have a sequence identity ranging from 35% to 41% to the bovine sequences. We carried out modeling of each of the subunits as described above. The best model of each subunit was docked into the cryo-EM density of the TRiC open state according to the proposed subunit arrangement 33 . This initial docking was refined by breaking the subunit into three domains ( Fig. 3a ). Domain boundaries were initially assigned analogously to those in the thermosome 25 and GroEL 30 and verified through normal mode analysis of the comparative models 39 using the Molecular Modeling Toolkit ( http://dirac.cnrs-orleans.fr/MMTK/ ). We performed the docking of each of the three domains to the TRiC open-state cryo-EM map independently, resulting in a similarity score of 0.963, 0.980 and 0.971 for the α subunit apical, intermediate and equatorial domains, respectively, as assessed by the EMAN program fh-stat.py 57 . These three rigid-body docked domains were then used together as a template, for a new round of model generation with MODELLER, to refine the structure and reestablish the broken chemical bonds in the hinge regions between the domains. Finally, we applied the same loop refinement and quaternary model building procedures described above. The optimized eight-subunit comparative model was docked into the TRiC open-state density map using foldhunter 57 , resulting in the similarity score of 0.969 as evaluated by the EMAN program fh-stat.py. Comparing comparative models for the open and closed states The trajectory of transition between the two docked conformations was modeled by linear interpolation between these two states using the morph.tcl script in the VMD software package ( http://www.ks.uiuc.edu/Research/vmd/ ). The buried surface area between the apical and intermediate domains of GroEL and the thermosome was calculated using the g_sas module from GROMACS ( http://www.gromacs.org/ ) 58 . Normal mode analysis on the TRiC open (nucleotide-free) state We carried out NMA 36 – 38 , 40 , 41 on the comparative model of the apo TRiC (open state) to investigate the motions of TRiC in its native state. The same procedure was adopted as described previously 37 . Low-frequency modes from NMA represent large global conformational changes observed in biological systems 41 . The lowest-frequency nondegenerate normal mode, mode 1, is shown in Supplementary Videos 3 and 4 , in which only Cα atoms were illustrated in the space-filling style. The top view of mode 1 ( Supplementary Video 3 ) clearly shows a rotation motion of the apical domains around the eight-fold rotational axis. Meanwhile, the side view of this mode ( Supplementary Video 4 ) shows that the intermediate domains rotate together with the apical domains in the same direction, whereas the equatorial domains maintain their original position. The motions interpreted from mode 1, the lowest-frequency nondegenerate normal mode, are highly consistent with the mechanism of lid closure proposed in this study on the basis of our open and closed TRiC cryo-EM reconstructions and corresponding homology models.

Supplementary Material Supp File Supp Vid 1 Supp Vid 2 Supp Vid 3 Supp Vid 4