Molecular dynamics simulations of co-translational protein folding intermediates on the ribosome

This repository contains input and output files of molecular dynamics (MD) simulations presented in the manuscript "Structures of protein folding intermediates on the ribosome" by Chan et al. (2025). The main outputs of this work and dataset are all-atom structural ensembles of co-translational folding intermediates of FLN5 on the ribosome (I1 and I2), which were validated by 19F NMR, cryo-EM and cross-linking experiments (described in the paper). Briefly, proline free energy calculations (in "proline_metad_fln5") show that the isomerisation state of Pro742 near the C-terminus of FLN5 is determined by the structure of the C-terminal strand, and therefore, that I1 and I2 contain a cis-Pro742 and trans-Pro742, respectively. Structural models of I1 were predicted using ratchet-and-pawl folding simulations (rMD_* folders), while structural models for I2 a previous NMR structure of an intermediate sampled by isolated, truncated FLN5 was used ("Systematic mapping of free energy landscapes of a growing filamin domain during biosynthesis". Waudby et al., 2018, PNAS). Both intermediates were then subjected to extensive (15 x 1 microsecond using 5 different starting structures), unbiased MD simulations on the ribosome with the DES-Amber-3.20 force field (unbiased_I1 and unbiased_I2 folders). The "proline_metad_fln5" folder contains well-tempered metadynamics MD trajectories of unfolded, truncated (delta6) and full-length (FL) FLN5. These simulations were performed with GROMACS version 2020.4 and PLUMED version 2.6.2 to simulate the proline isomerisation free energy landscape of Pro742. All constructs were simulated in triplicate and each subfolder contains the input tpr files, the output .gro files, the .log files of the simulations, values of the collective variables used, and the metadynamics bias potential data (COLVAR and HILLS files). An input plumed.dat and .pdb file of the protein is also provided along with the processed trajectories containing protein coordinates (to be loaded together with the processed.pdb file). For all simulations the DES-Amber protein force field (with a charge scaling factor of 0.9) was used. All trajectories are 1 microsecond long. The "rMD_isolated_FLN5" folder contains ratchet-and-pawl MD simulations (rMD) of isolated, full-length FLN5 where folding was simulated from unfolded starting structures. These simulations were performed with GROMACS version 2020.4 and PLUMED version 2.6.2. The folder contains the input .tpr files and concatenated .xtc files (all.xtc) of 321 successful folding transitions, along with a subdirectory containing analysis files including calculations of RMSD relative to the native structure (em_protein.pdb) and the fraction (rmsd*.xvg) of native contacts (COLVAR). The input plumed and .mdp files for the simulations are also included. Each individual trajectory is 5 ns long. The "rMD_FLN5_47_RNC" folder contains ratchet-and-pawl MD simulations (rMD) of the FLN5+47 RNC where folding was simulated from unfolded starting structures. These simulations were performed with GROMACS version 2020.4 and PLUMED version 2.6.2. The folder contains the input .tpr files and concatenated .xtc files (all.xtc) of 226 successful folding transitions, along with a subdirectory containing analysis files including calculations of RMSD relative to the native structure (em_protein.pdb) and the fraction (rmsd*.xvg) of native contacts (COLVAR). The input plumed and .mdp files for the simulations are also included. Each individual trajectory is 5 ns long. The concatenated trajectory file contains all atoms corresponding to the nascent chain of FLN5+47 (nc.pdb), and a pdb file corresponding to the position-restrained ribosome during the simulations is also provided (ribosome.pdb). The "unbiased_I1_tprfiles" and "unbiased_I2_tprfiles" folders contain input .tpr files for all unbiased MD simulations of I1 and I2 on the ribosome (3x independent simulations with different starting velocities starting from 5x different input conformations referred to as conf1 to conf5). The folders also contain the input coordinates of the RNC complex as .pdb files and an input mdp file. These simulations were performed with GROMACS version 2023.1. The "unbiased_I1_ensemble" and "unbiased_I2_ensemble" contain the structural ensembles of I1 and I2 obtained from the unbiased MD simulations (15 microseconds and 75,000 frames total per intermediate state). The nascent chain coordinates can be loaded with nc.pdb and trajall.xtc and the ribosome coordinates are in ribosome_sim.pdb. These folders additionally contain source data of analyses presented in the manuscript (e.g., calculated distances between residue pairs for all frames stored in *.txt files). These also include clustering analyses (Clustering subdirectories), FLN5-ribosome distance calculations (FLN5_ribo_distance_data subdirectory), native contact analyses (Native_contact_analysis subdirectory), rotational autocorrelation data calculated for the FLN5 domain on the ribosome stored in numpy files as python dictionaries (ROT_TAUC_data_900ns_2ns subdirectory), backbone dynamics analyses (RMSF_analysis subdirectory), and calculated ribosome interaction probabilities (Ribosome_interaction analysis subdirectory). The "19F_simulations_ * " folders contain additional simulations of isolated, fluorinated protein variants that aid in the interpretation of the 19F NMR data in the manuscript. All simulations were performed with GROMACS version 2023.1. The subdirectories for each protein variant/simulation contain the input .tpr files, processed trajectory files (.xtc) with a protein topology file (.pdb). Source data of analysed RMSD, RMSF, fluorinated sidechain SASA values, relevant distances between fluorinated amino acid and other residues are also included in each variant/simulation folder (triplicate for all simulations, see Chan et al (2025) for details). The "19F_simulations_1" folder contains three subdirectories: "19F_MD_732TAG" contains simulations of the FLN5 732tfmF variant (to analyse the interaction between 732tfmF and 692Glu sidechains with two force fields - C36m and ff15ipq). The "19F_sims_Istates_mimic" contain all simulations of fluorinated FLN5 variants simulated in the I1 and I2 conformations (ff15ipq force field). The "19F_Titin_deltaA_MD_10us" contains simulations of fluorinated variants of titin I27 lacking the A-strand (residues 1-7) deleted (10 microseconds per simulation, ff15ipq force field). Finally, the "19F_MD_buried_sites" directory contains simulations of fluorinated FLN5 variants whose sidechains are buried in the core of the native state, and some exposed residues as controls (3-6 repeats per variant lasting 1 microsecond each with the ff15ipq force field). The subdirectory "HBOND_DATA" contains analyses of hydrogen bonds formed between fluorine atoms and the protein or solvent molecules (including their approximate lifetimes and populations as described in the manuscript). For analyses of fluorine-solvent hydrogen bonds, the .xtc files contain all water molecules of the simulation box.