The building of initial models into the experimental density maps was guided by the available crystal structure of Elp2 [Protein Data Bank (PDB) ID, 5M2N], the crystal structure of the Elp1 CTD (PDB ID, 5CQS), as well as the homology models of Elongator subunits Elp1 and Elp3 described previously (26). For initial model building, the models were placed into the apoElp123 lobe map at a resolution of 3.3 Å by rigid body fitting in Chimera (46), and this initial model was used for the first round of LocScale (see below). The fitted structures of individual subunits were used to aid de novo model building and rebuilding reference models in the LocScale maps using Coot (47). In detail, the WD40 domains of Elp1 were only used for the generation of the first LocScale map, as they were regarded as low quality with the confident assignment of the fold but uncertain sequence register. Thus, both WD40 domains of Elp1 were built de novo, aided by secondary structure prediction and side-chain densities in this region of the apoElp123 lobe map (fig. S2). The fit of the Elp1 CTD was optimized by flexible fitting of the crystal structure (PDB 5CQS) using iMODfit (48) until convergence of cross-correlation below a threshold of 0.00001 over 5000 steps (with maximum 100,000 steps) and refined in Coot in regions with good enough map quality. Before the fitting, missing side chains were added with Modeller (49), and the stereochemistry was refined with energy minimization using GROMACS (50). To prevent overfitting during the flexible fitting, constraints on the secondary structure were applied. Moreover, overfitting was controlled by comparing the resulting models with the rigid body fragments of the starting crystal structure of the Elp1 CTD (PDB ID, 5CQS), ensuring that relative arrangements of secondary structure elements are locally preserved and that side chains are similar to the starting structure. Moreover, we extended the Elp1 structure with the “link loop” (residues 811 to 853), which was predicted on a secondary structure level and perfectly followed a prominent density. In addition, the Elp3 structure was extended with the helix (residues, 392 to 403) that was clearly visible in the cryo-EM density (apoElp123 lobe map) showing side-chain density and was also predicted on a secondary structure level. Our current model of apoElp123 accounts for most features observed in the experimental EM density maps (Elp123 lobe, Elp123 lobe DD, and full Elp123), including two strong densities located close to the active site that corresponds to the [4Fe4S] cluster and the 5′dA molecule.

tRNA-bound models were generated by flexible fitting of the apoElp123 model using iMODFit until convergence (with the convergence achieved in less than 500,000 steps). To identify the optimal fit of the tRNA and to select the template for comparative modeling of tRNAAlaUGC, fitting 802 tRNA chains available in the RNA Bricks database (51) to the segment of the EM map corresponding to the apparent tRNA density. For each structure, the fitting was performed using the University of California, San Francisco (UCSF) Chimera (46) global search with 10,000 random initial positions and normalized cross-correlation as the fitting metric. The statistical significance of the fits was assessed as a P value calculated from the normalized cross-correlation scores. To calculate the P values, the cross-correlation scores were first transformed to z scores (Fisher’s z transform) and centered, from which two-sided P values were computed using standard deviation derived from an empirical null distribution [derived from all obtained fits and fitted using fdrtool (52) from the R package] (53, 54). All P values were corrected for multiple testing using the Benjamini-Hochberg procedure, with the number of tests corresponding to all resulting fits from all 802 structures (i.e., treating each fitting attempt from a random initial position as an independent test). From the top fits, the structure of tRNA-Glu that bound to RlmN methyltransferase (PDB ID, 5HR6, chain D) was selected as the modeling template. This structure and the corresponding fit ranked first according to the normalized cross-correlation score (adjusted P = 6 × 10−7). The comparative model based on this template was built using ModeRNA (55). The G:U base pair 50–64 of the model was further refined using SimRNA (56) to recapitulate appropriate hydrogen bond interactions, and the final models were minimized using GROMACS (50). Note that because of limitations of comparative modeling and limited resolution of the tRNA density, the comparative model of the tRNA-Ala should be regarded with caution with deviations from the true structure expected, particularly in the regions of D- and T-loop and the acceptor stem. These deviations are not expected to affect the overall fit and the anticodon loop, where the quality of the density is sufficient for de novo tracing. The comparative model was fitted into the Elp123-tRNA density map at a resolution of 4.4 Å by superposition onto the template fitted as described above and local optimization using “Fit in Map” tool of UCSF Chimera. The anticodon loop was built de novo by manual building in Coot.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.