Computational protocol: Ensemble cryo-EM elucidates the mechanism of translation fidelity

Similar protocols

Protocol publication

[…] Data for the cognate complex and near-cognate complex were collected on a Titan Krios electron microscope (FEI) operating at 300 kV and equipped with K2 Summit direct electron detector (Gatan Inc.) using 0.5- to 2.2-µm underfocus. For the cognate complex, a dataset of 800,367 particles from 3028 videos was collected automatically using SerialEM. 50 frames per video were collected at 1 e−/Å2 per frame for a total dose of 50 e−/Å2 on the sample. For the near-cognate complex, a dataset of 572,417 particles from 1773 videos of 30 frames each was collected. The videos for the near-cognate complex were taken with 1 e−/Å2 per frame for a total dose of 30 e−/Å2 on the sample. For both datasets, the super-resolution pixel size was 0.82 Å on the sample. [...] Particles were extracted from aligned video sums as follows. Videos were processed using IMOD to decompress frames and apply the gain reference. Videos were drift-corrected using unblur. Magnification anisotropy of the video sums was corrected with mag_distortion_estimate and mag_distortion_correct. CTFFIND3 was used to determine defocus values. Particles were automatically picked from 10×-binned images using Signature with a ribosome reference (18 representative reprojections of the EM databank map 1003, which was low-pass filtered to 50 Å). 480 × 480-pixel boxes with particles were extracted from super-resolution-aligned and magnification-anisotropy-corrected images, and the stack and FREALIGN parameter file were assembled in EMAN2. To speed up processing, binned image stacks were prepared using resample.exe, which is part of the FREALIGN distribution. [...] FREALIGN v9 (versions 9.07–9.11) was used for all steps of refinement and reconstruction (). A 6×-binned image stack was initially aligned to a ribosome reference (EM databank map 1003,) using three rounds of mode 3 (global search) alignment, including data in the resolution range from 300 Å to 30 Å. Next, the 2×-binned, and later the unbinned image stacks were successively aligned against the common reference using mode 1 (local refinement), including data up to a high-resolution limit (6 Å for the cognate ternary complex or 8 Å for near-cognate ternary complex), whereupon the resolution of the common reference stopped improving. Subsequently, the refined parameters were used for classification of 4×-binned stacks into 6 classes in 50 rounds using a spherical (60-Å radius) focus mask around EF-Tu and A/T tRNA, including resolutions from 300 to 8 Å during classification. This procedure yielded three EF-Tu-containing classes for the cognate complex and one for the near-cognate complex ().Further processing of the cognate complex was as follows. The Structure III map at 3.2-Å resolution was obtained from the 6-model classification described above at 1× binning. A related map, Structure IIIb, also had a closed 30S conformation and activated EF-Tu at the SRL, but a disordered L11 stalk. This class consisted of 72,533 particles and was not used for structure modeling and refinements. Finally, 50,667 particles belonging to the open-30S class were extracted using merge_classes.exe, including particles with >50% occupancy and scores >0. The resulting substack was subjected to further classification with a focused mask (30-Å radius) around the DC. Using three classes separated Structure I and Structure II from a third class in which the anticodon was disordered. The final maps for the Structures I and II were prepared from these classes, using 50% of particles with highest scores.Further processing of the near-cognate complex proceeded as follows. 37,341 particles belonging to the single class bound with EF-Tu from the 6-model classification described above, were extracted using merge_classes.exe using thresholds of >90% occupancy and scores >10. The particles were classified again for 50 rounds using the same 60-Å-wide focus mask around EF-Tu and A/T tRNA, including resolutions from 300 to 8 Å during classification. This classification separated the near-cognate GTPase-activated state (Structure III-nc) from two maps with an open 30S subunit. Particles belonging to Structure III-nc were extracted using merge_classes.exe using thresholds of >75% occupancy and scores >10 and 50% of them with the highest score were used to prepare the final Structure III-nc map. Particles belonging to the classes with the open 30S subunit were extracted using merge_classes.exe using thresholds of >75% occupancy and scores >10. The resulting substack of 23,078 particles was subjected to a 2-model classification with a focused mask (30-Å radius) around the DC. The final maps for Structures I-nc and II-nc were prepared from these classes, using 50% of particles with highest scores.We report the percentages of the particles that belong to Structures I, II or III or Structures I-nc, II-nc, or IIInc in . The percentages were calculated using all particles assigned to the corresponding classes shown in (Structure III comprises particles assigned to both 30S-domain-closed classes III and IIIb, which only differ in the L11 stalk occupancy, as described above).The maps used for structure refinements were sharpened by applying negative B-factors of up to −100 Å2 using bfactor.exe (included with the FREALIGN distribution). FSC curves were calculated by FREALIGN for even and odd particle half-sets. frealign_calc_stats was used to derive the number of particles assigned to each class. Blocres was used to assess local resolution of unfiltered and unmasked volumes using a box size of 60 pixels, step size of 10 pixels, and resolution criterion of FSC value at 0.143. [...] The high-resolution cryo-EM structure of the 70S•tRNA•EF-Tu•GDP•kirromycin complex (PDB: 5AFI), excluding EF-Tu, A/T-,P-, and E-site tRNAs, was used as a starting model for structure refinements. The starting structural models for fMet-tRNAfMet in the P and E sites were adopted from the 70S•RF2•tRNA crystal structure. We could not distinguish the identity of E-site tRNA (tRNAfMet or tRNAPhe for cognate complex or tRNAfMet or tRNALys for near-cognate complex) due to lower than average resolution of this part of the cryo-EM maps, likely due to conformational flexibility suggested by further classification. Since tRNAfmet was used in the absence of EF-Tu and is likely to bind the E site upon deacylation, we modeled the E-site tRNA as tRNAfMet. The starting model for A/T Phe-tRNAPhe in Structures II and III was taken from PDB: 5AFI. The starting model for Phe-tRNAPhe in Structure I was from the crystal structure of the isolated T. aquaticus ternary complex (PDB: 1TTT). For the near-cognate structures, the starting model for Lys-tRNALys was from the crystal structure of the ribosome with a near-cognate tRNALys in the A site (PDB: 5IB8). The T. aquaticus ternary complex (PDB: 1TTT) was used for homology modeling of E. coli EF-Tu employing SWISS-PROT and deriving the initial structure of GDPCP. A homology model was similarly created for E. coli L1 utilizing the crystal structure of the isolated T. thermophilus L1 stalk (PDB: 3U4M).All Structures were domain-fitted using Chimera and refined using real-space simulated-annealing refinement using RSRef, against corresponding maps. Atomic electron scattering factors were used during refinement. Local structural elements that differed between structures, such as the DC, were manually fitted into cryo-EM maps prior to refinement. Refinement parameters, such as the relative weighting of stereochemical restraints and experimental energy term, were optimized to produce the optimal structure stereochemistry, real-space correlation coefficient and R-factor, which report on the fit of the model to the map. Secondary-structure restraints, comprising hydrogen-bonding restraints for ribosomal proteins and base-pairing restraints for RNA molecules were employed as described. The structures were next refined using phenix.real_space_refine followed by a round of refinement in RSRef applying harmonic restraints to preserve protein backbone geometry,. Ions were modeled as Mg2+ in Structure III, filling the difference-map peaks (using CNS) residing next to oxygen atoms. Phenix was used to refine B-factors of the models against their respective maps. The resulting structural models have good stereochemical parameters, characterized by low deviation from ideal bond lengths and angles and agree closely with the corresponding maps as indicated by high correlation coefficients and low real-space R factors (). Structure quality was validated using MolProbity.The cryo-EM maps for Structure I and Structure II-nc did not allow unambiguous visual assignment of the G530 conformation. To interpret a predominant conformation, we prepared two ribosome models with G530 in the alternative conformations, syn and anti, and refined the complete ribosome structures independently against corresponding maps. Following the refinements, a preferred fit was assessed based on the local real-space correlation coefficient (CC, calculated only for G530 non-hydrogen atoms). The local CC in Structure I suggests G530-syn and G530-anti fit nearly equally well (CC=0.67 vs. CC=0.66, respectively, ). The local CC in Structure II-nc suggests a better fit for G530-syn (CC=0.63), whereas a refined G530-anti yields the moderately lower CC of 0.57. In Structure II, whose density unambiguously shows G530-anti (), this preferred conformation yields the CC of 0.71, whereas a refined G530-syn fits poorly and exhibits the CC of 0.44.Figures were prepared in Chimera and Pymol,. […]

Pipeline specifications

Software tools SerialEM, IMOD, Unblur, mag_distortion_estimate, CTFFIND, Frealign, EMAN, bfactor, PHENIX, CNS, MolProbity, PyMOL
Databases UniProt
Applications cryo-EM, Protein structure analysis
Chemicals Guanosine Triphosphate