Computational protocol: Structure of the Mammalian Ribosome-Sec61 Complex to 3.4 Å Resolution

Similar protocols

Protocol publication

[…] Particle picking was performed using EMAN2 (), contrast transfer function parameters were estimated using CTFFIND3 (), and all 2D and 3D classifications and refinements were performed using RELION (). The resulting density maps were corrected for the modulation transfer function (MTF) of the detector and sharpened as previously described (). [...] The porcine 80S ribosome was built using the moderate resolution model for the human ribosome (), while the Sec61 channel bound to both the idle and translating ribosome were built using the crystal structure of the archaeal SecY () and the models of the canine Sec61 bound to the ribosome (). All models were built in COOT (), and refined using REFMAC v5.8 (). Secondary structure restraints for the Sec61 channel were generated in ProSMART (). To test for overfitting, we performed a validation procedure similar to that described previously (). The final models for the 40S and 60S subunits were rigid-body fitted into the maps for the remaining classes, and refined. Figures were generated using Chimera () and PyMOL (). Extended Experimental Procedures Sample Preparation Porcine pancreatic microsomes were prepared as previously described (), resuspended in membrane buffer (50 mM HEPES, 250 mM sucrose, 1 mM DTT) and flash frozen for long-term storage at −80°C. Microsome flotation and sucrose gradient experiments showed that all ribosomes in the preparation were membrane bound and primarily in polysomes (data not shown). Details of additional characterization are shown in . To convert polysomes into monosomes, microsomes were adjusted to 1 mM CaCl2 and 150 U/ml micrococcal nuclease, incubated at 25°C for 7 min, adjusted to 2 mM EGTA, and flash frozen in single-use 50 ul aliquots. A 50 ul aliquot of nuclease-digested microsomes (at an A280 of 90) was adjusted with an equal volume of ice cold 2X solubiization buffer (3.5% digitonin, 100 mM HEPES [pH 7.5], 800 mM KOAc, 20 mM MgOAc2, 2 mM DTT) and incubated 10 min on ice. Earlier reports using similar solubilization conditions at both higher and lower salt observed no loss of nascent chains from the mammalian Sec61 channel as judged by protease protection assays (). Samples were spun for 15 min at 20,000 x g to remove insoluble material, and the solubilized material was fractionated by gravity flow over a 1 ml Sephacryl-S-300 column pre-equilibrated in column buffer (50 mM HEPES [pH 7.5], 200 mM KOAc, 15 mM MgOAc2, 1 mM DTT, and 0.25% digitonin). Roughly 100 ul fractions were manually collected and the void fraction containing ribosome-translocon complexes was identified by A260 measurements. The sample was centrifuged again as above to remove any potential aggregates before using immediately to prepare and freeze cryo-EM grids. It is worth noting that we efficiently recovered nontranslating ribosome-Sec61 complexes despite using 400 mM KOAC during solubilization. Two reasons probably contributed to this high recovery: (i) solubilization and fractionation at a very high sample concentration, favoring an otherwise weak interaction; (ii) re-association of ribosomes with Sec61 when the salt concentration was reduced upon entering the gel filtration resin. Note for example that dissociation of translocon compoments was greater using sucrose gradient sedimentation than the more rapid Sephacryl-S-300 separation (D). A second question is why a large proportion of our ribosomes contained no tRNA, some of which have eEF2. We do not know for certain, but differences from most earlier ribosome preparation protocols include isolating ribosomes from a microsomal subcellular fraction derived from a native tissue, the nuclease digestion reaction, and the specific conditions used for solubilisation and purification. Most earlier ribosome purification protocols are typically from total cell lysates, do not employ high detergent concentrations, often involve greater fractionation, and are bound to Stm1-like proteins. Grid Preparation and Data Collection Ribosome-Sec61 complexes were diluted in column buffer to a concentration of 40 nM, and were applied to glow-discharged holey carbon grids (Quantifoil R2/2) which had been coated with a ∼50–60 Å thick layer of continuous amorphous carbon. After application of 3 μl of sample, the grids were incubated at 4°C for 30 s, blotted for 9 s, and flash-cooled in liquid ethane using an FEI Vitrobot. Data were collected on an FEI Titan Krios operating at 300 kV, using FEI’s automated single particle acquisition software (EPU). Images were recorded using a back-thinned FEI Falcon II detector at a calibrated magnification of 104,478 (pixel size of 1.34 Å), using defocus values between 2.5–3.5 μm. Videos from the detector were recorded using a previously described system at a speed of 17 frames/s (). Image Processing Semi-automated particle picking was performed using EMAN2 (), which resulted in selection of 83,839 particles from 1,410 micrographs. A smaller second data set (referred to as data set 2 below) was collected later from another grid containing the same sample in order to increase the number of particles containing tRNAs and nascent chain. Data set 2 contained 726 micrographs that led to the selection of 37,061 particles. Contrast transfer function parameters were estimated using CTFFIND3 for both data sets (), and any micrographs that showed evidence of astigmatism or drift were discarded at this stage. All 2D and 3D classifications and refinements were performed using RELION as described below (). Unsupervised 2D class averaging was used to discard any nonribosome or defective particles, which resulted in a final data set of 80,019 80S particles, and an additional 36,696 particles from data set 2. Each data set was individually refined against a map of the S. cerevisiae ribosome filtered to 60 Å resolution, utilizing statistical movie processing in RELION as described previously (). As the 40S subunit was in several distinct conformations, a mask that included only the 60S subunit was used during refinement of the complete initial data set. This resulted in a final resolution of 3.35 Å using 80,019 particles for the 60S subunit as judged by the gold-standard FSC criterion (). In parallel, 3D classification of the initial 80,019 80S particles was performed using angular sampling of 1.8 degrees and local angular searches (using angles determined from the 3D refinement). Ten possible classes were allowed and resulted in the following. Class 4: representing ∼13% of the data set (10,555 particles), contained ribosomes in a ratcheted conformation with A/P- and P/E-site tRNAs and a nascent polypeptide in the ribosomal tunnel. Classes 6,7,9: three identical classes, together representing ∼46% of the data set (36,667 particles), contained ribosomes with eEF2 in a partially ratcheted orientation. Classes 2 and 3: together representing ∼19% of the data set, contained eEF2 bound to a ribosome in the canonical, unratcheted conformation. Classes 1,5,8,10: together comprising ∼22% of particles, contained empty ribosomes, without tRNAs or translation factors. Given the large percentage of particles that contain eEF2 in this sample (∼65%), weak density was observed for the factor in refinements using the complete data sets. However, continuous density for the factor could only be observed in refinements of appropriately grouped particles, described below. The apparent composition of the sample was similar for data set 2, from which 4,168 particles (∼11%) containing hybrid state tRNAs and a nascent peptide chain were combined with class 4 above to produce a larger set of particles for the translating ribosome-Sec61 structure. Particles identified from 3D classification were combined according to biological state, and subjected to a final 3D refinement resulting in the density maps used for model building as described in . The idle ribosome-Sec61 map was obtained using the 69,464 particles (Classes 1-3 and 5-10) that did not contain tRNAs or a nascent peptide, refined using a 60S mask to 3.4 Å resolution. The map of the translating ribosome-Sec61 complex, obtained by combining the first and second data sets, resulted in 14,723 particles that produced a 3.9 Å density map. The 40S subunit and the ribosomal stalk were best resolved in the 36,667 particles containing eEF2 (Classes 6,7,9) and the 40S subunit in a defined orientation, which extended to 3.5 Å resolution. All maps were corrected for the modulation transfer function (MTF) of the detector, and then sharpened using a negative B-factor (as described in ), which was estimated using previously reported procedures (). Local resolution of the final unsharpened maps was calculated using ResMap (). Model Building and Refinement The porcine 80S ribosome was built using the moderate resolution model for the human ribosome (PDB IDs 3j3a, 3j3b, 3j3d, and 3j3f; ). Sec61 bound to both the idle and translating ribosome was built using the crystal structure of the archaeal SecY (PDB ID 1RH5; ) and the low-resolution model of the canine Sec61 bound to the ribosome (PDB IDs 4CG7 and 4CG5; ). All models were built in COOT (), and refined using REFMAC v5.8 () as previously described (). Registry and other errors to the ribosomal proteins were corrected manually, and each chain was refined individually against an appropriately cut map. Secondary structure restraints were generated in ProSMART (), and nucleic acid base-pairing and stacking restraints were generated as before () and were maintained throughout refinement to prevent overfitting. Ramachandran restraints were not applied, such that backbone dihedral angles could be used for subsequent validation of the refined models. To test for overfitting, we performed a validation procedure similar to that described previously (). In brief, the final model was refined against an unsharpened density map calculated from only one half of the data using empirically determined chemical restraints. The resulting model was then used to calculate FSC curves for both halves of the data, one of which had been used during the refinement, and the other which had not (). The two FSC curves nearly overlap, and we observe significant correlation beyond the resolution used for refinement (indicated by a vertical dashed line in B and S3D), demonstrating that the model has predictive power and has not been overfitted. The models for the 60S and 40S subunits were then refined using these same restraints against the highest resolution sharpened maps for each subunit ( Map2 and Map 5, respectively). The resulting models for the 60S subunit, the 40S body, and 40S head were individually rigid-body fitted into the maps for the remaining classes. All figures were generated using Chimera () and PyMOL (). […]

Pipeline specifications

Software tools EMAN, CTFFIND, RELION, Coot, PyMOL, ResMap
Applications cryo-EM, Protein structure analysis
Organisms Homo sapiens