Bayesian reweighting of biomolecular structural ensembles using heterogeneous cryo-EM maps with the cryoENsemble method

Abstract

Cryogenic electron microscopy (cryo-EM) has emerged as a powerful method for the determination of structures of complex biological molecules. The accurate characterisation of the dynamics of such systems, however, remains a challenge. To address this problem, we introduce cryoENsemble, a method that applies Bayesian reweighting to conformational ensembles derived from molecular dynamics simulations to improve their agreement with cryo-EM data, thus enabling the extraction of dynamics information. We illustrate the use of cryoENsemble to determine the dynamics of the ribosome-bound state of the co-translational chaperone trigger factor (TF). We also show that cryoENsemble can assist with the interpretation of low-resolution, noisy or unaccounted regions of cryo-EM maps. Notably, we are able to link an unaccounted part of the cryo-EM map to the presence of another protein (methionine aminopeptidase, or MetAP), rather than to the dynamics of TF, and model its TF-bound state. Based on these results, we anticipate that cryoENsemble will find use for challenging heterogeneous cryo-EM maps for biomolecular systems encompassing dynamic components.

Keywords: Cryo-EM; Molecular dynamics simulations; Statistical inference; Structural biology; Trigger factor.

Figure 1

Schematic illustration of the standard cryoENsemble method and its iterative extension. The input includes a structural ensemble (depicted in grey), typically obtained from molecular dynamics simulations, and a cryo-EM map of the biological system under investigation. Each model from the prior structural ensemble is fitted into the reference cryo-EM density before the cryoENsemble calculations. Subsequently, density is generated for each structure and starting weights (w_i) are assigned. In the standard mode, Bayesian reweighting generates new weights (w′_i), indicated in the figure by varying shades of grey, for each structure in the ensemble (posterior structural ensemble) as well as posterior density for the system (see Methods). In the iterative mode, following the standard reweighting, a sub-ensemble of structures that meets weight criteria is selected for another round of Bayesian reweighting, which generates new weights (w″_i) for the sub-ensemble. This process is repeated until the agreement with the experimental data decreases (reflected by an increase in ${\chi }^2$). The iterative mode returns the minimal set of structures that maintains good agreement with the experimental cryo-EM data and posterior density.

Figure 2

CryoENsemble reweighting of ADK. (A) Open state populations calculated from cryoENsemble reweighting of the ADK dataset. The open state populations obtained after structural ensemble reweighting for each ADK dataset are shown in orange for the standard mode and green for the iterative mode, along with the target values (black circle). (B) Correlations between reference maps and posterior maps upon cryoENsemble reweighting of the ADK dataset in standard (orange) and iterative mode (green), including values for the prior ensemble and the best single structure. The datasets vary in resolution, noise level, and reference populations of the open state.

Figure 3

CryoENsemble reweighting of the FLN5-6 nascent chain dataset. (A) Correlations between the reference maps and posterior maps upon standard and iterative cryoENsemble reweighting of the FLN5-6 nascent chain dataset. Correlation coefficients calculated between the FLN5-6 nascent chain reference density maps and maps obtained before and after the reweighting, as well as the maps derived from the best single structure fitted into the reference density map. The 100 reference density maps (at resolutions of 3, 6, and 10 Å, and with noise levels of 1% and 10%) were generated based on ten randomly selected structures from the MD ensemble. (B) Weights obtained upon standard cryoENsemble reweighting of the FLN5-6 nascent chain dataset. Examples of the reweighting process for FLN5-6 nascent chain based on the reference map (at resolutions of 3, 6, and 10 Å, and with noise levels of 1% and 10%). Weights are calculated with different theta (θ) values ranging from 0 to 10⁷, and with black lines, we depict optimal weights selected based on the L-curve analysis. Additionally, weights corresponding to the ten models used to generate the reference map are circled.

Figure 4

CryoENsemble reweighting can retrieve local dynamics from cryo-EM maps. The root-mean-square deviations calculated between the RMSF of the target ensemble and the RMSFs from the prior ensemble (in grey), standard reweighting (in orange), and iterative reweighting (in green).

Figure 5

CryoENsemble iterative reweighting of the TF dataset. (A) Analysis of the effect of reweighting on the weights of each cluster obtained from the MD simulations. The green circle size is proportional to the population of the cluster upon reweighting. (B) The structural model with the highest weight selected by cryoENsemble (Supplementary Fig. 30) is visualised in two different orientations inside the cryo-EM map. (C) The three main states obtained upon reweighting corresponding to the cluster_1, cluster_2 and cluster_14, along with their populations. (D) RMSF calculated on the reweighted ensemble and mapped on the structure of TF. (E) Visualisation of the principal mode PC1 that captures the most dominant motion (indicated by blue arrows) within the reweighted ensemble.

Figure 6

CryoENsemble iterative reweighting captures conformational and compositional variability in cryoEM maps. (A) The cryo-EM map (EMDB: 30611) with unaccounted density coloured in green. (B) The outcome of fitting the MetAP structure (PDB ID: 1MAT) into the unaccounted density (from (A)), presented along the 70S-Trigger factor structure (PDB ID: 7D80). (C) The main cluster of TF structures, with a population of 10%, obtained after the reweighting of combined structural ensembles (TF and TF + MetAP). (D) The main cluster of TF + MetAP structures, with a population of 12%, obtained after the reweighting of combined structural ensembles (TF and TF + MetAP).