1. Structural Biology and Molecular Biophysics

A Bayesian approach to single-particle electron cryo-tomography in RELION-4.0

  1. Jasenko Zivanov
  2. Joaquín Otón
  3. Zunlong Ke
  4. Andriko von Kügelgen
  5. Euan Pyle
  6. Kun Qu
  7. Dustin Morado
  8. Daniel Castaño-Díez
  9. Giulia Zanetti
  10. Tanmay AM Bharat
  11. John AG Briggs  Is a corresponding author
  12. Sjors HW Scheres  Is a corresponding author
  1. MRC Laboratory of Molecular Biology, United Kingdom
  2. Laboratory of Biomedical Imaging (LIB), Switzerland
  3. BioEM lab, Biozentrum, University of Basel, Switzerland
  4. ALBA Synchrotron, Spain
  5. Max Planck Institute of Biochemistry, Germany
  6. Sir William Dunn School of Pathology, University of Oxford, United Kingdom
  7. Institute of Structural and Molecular Biology, Birkbeck College, United Kingdom
  8. Instituto Biofisika, Spain
https://doi.org/10.7554/eLife.83724
Share this article
Cite this article
  1. Jasenko Zivanov
  2. Joaquín Otón
  3. Zunlong Ke
  4. Andriko von Kügelgen
  5. Euan Pyle
  6. Kun Qu
  7. Dustin Morado
  8. Daniel Castaño-Díez
  9. Giulia Zanetti
  10. Tanmay AM Bharat
  11. John AG Briggs
  12. Sjors HW Scheres
(2022)
A Bayesian approach to single-particle electron cryo-tomography in RELION-4.0
eLife 11:e83724.
https://doi.org/10.7554/eLife.83724
  2. Download BibTeX
  3. Download .RIS
3 figures, 2 tables and 1 additional file

Figures

Figure 1 with 3 supplements
Download asset Open asset
Subtomogram averaging of the HIV-1 immature capsid.

(a) Fourier Shell Correlation (FSC) for resolution estimation of iteratively improved reconstructions using the new RELION-4.0 workflow. (b) Representative region of reconstructed density in the final map. (c) The same density as in (b), together with the published atomic model 5L93, which has not been additionally refined in the density.

Figure 1—figure supplement 1
Download asset Open asset
Iterative map improvement.

Representative region of reconstructed densities at several stages of the iterative refinement process. The four panels on the left show the four stages with the same labels as in Figure 1a; the fifth panel shows the same region from the final map that was calculated from all 43 tomograms.

Figure 1—figure supplement 2
Download asset Open asset
Comparison with emClarity.

(a) FSC for the RELION-4.0 map, which used 12,910 particles from the five tomogram subset (blue); the reported FSC for emClarity, which used 15,460 particles from the five tomogram subset (orange); and the FSC calculated from the deposited half-maps of emClarity, using the same mask as used for the RELION-4.0 FSC curve (yellow). (b) Representative region of the map calculated by RELION-4.0 from the five tomogram subset. (c) The same region of the map calculated by emClarity.

Figure 1—figure supplement 3
Download asset Open asset
Comparison with M/RELION-3.1.

(a) FSC for the RELION-4.0 map, which used 144,275 particles from the full 43 tomogram data set (blue); the reported FSC for M/RELION-3.1, which used 130,658 particles from the full 43 tomogram data set (orange); and the FSC calculated from the deposited half-maps of M, using the same mask as used for the RELION-4.0 FSC curve (yellow). (b) Representative region of the map calculated by RELION-4.0 from the full 43 tomogram data set. (c) The same region of the map calculated by M/RELION-3.1.

Figure 2 with 1 supplement
Download asset Open asset
Subtomogram averaging of the C. crescentus S-layer from cell stalks.

(a) FSC for resolution estimation of iteratively improved reconstructions using the new RELION-4.0 workflow, tested on the S-layer inner domain. (b) Densities for the previously identified lipopolysaccharide (LPS) (cyan and orange) and Ca2+ ions (green) in prior electron cryo-microscopy (cryo-EM) single-particle analyses are resolved. (c, d) The final map shows two densities for bound LPS O-antigen chains. Panel (c) shows only the S-layer protein as blue ribbon and (d) shows LPS O-antigen as orange and cyan sugars corresponding to the N-acetyl-perosamine and mannose moieties, respectively.

Figure 2—figure supplement 1
Download asset Open asset
Comparison of subtomogram averaging (STA) and single-particle analysis (SPA) reconstructions of the C. crescentus RsaA.

(a) Bottom view of STA map of C. crescentus RsaA with lipopolysaccharide (LPS) O-antigen binding site as orange and cyan sugars corresponding to the N-acetyl-perosamine and mannose moieties, respectively, at a threshold of 9 σ. (b) Same view of the SPA map (EMD-10389) with no second LPS O-antigen. (c) FSC between the original refined model (PDB-ID: 6T72) and the SPA map (blue), as well as the same model after a rigid body fit (orange solid line) and refinement with a 4 Å resolution cut-off (orange dashed line) into the STA map. (d) Overlay of the refined model in the STA map (orange) and the original model (light blue, PDB-ID: 6T72) with an overall RMSD of 0.62 Å.

Figure 3
Download asset Open asset
Subtomogram averaging of the COP-II inner layer.

(a) FSC for resolution estimation of iteratively improved reconstructions using the new RELION-4.0 workflow, tested on the COP-II inner layer. (b) Reconstructed density for the inner layer. (c) Zoomed-in region of the final map (in transparent grey) with the refined atomic model (blue).

Tables

Appendix 2—table 1
Electron cryo-tomography (Cryo-ET) data collection, refinement, and validation statistics.
HIV-1 Gag(EMD-16207 /EMD-16209)S-layer_inner_domain(EMD-16183)(PDB 8BQE)COPII inner coat(EMD-15949)(PDB 8BSH)
Data collection
MicroscopeTitan KriosTitan KriosTitan Krios
DetectorK2 (Gatan)K2 (Gatan)K2 (Gatan)
SoftwareSerialEM (Mastronarde, 2003)SerialEM (Mastronarde, 2003)SerialEM
(Mastronarde, 2003)
Voltage (kV)300300300
Slit width (eV)202020
Defocus range ( μm)–1.5 to –5.0–1.5 to –5.0–1.5 to –4.5
Pixel size (Å)1.351.351.33
Total exposure (e−/\AA2)120–145∼140∼120
Exposure per tilt ( e−/\AA2)3.0–3.53.42.9
Total number of tilts414141
Frames per tilt-movie8–101010
Tilt increment±3∘±3∘±3∘
Tilt-series schemeDose-symmetricalDose-symmetricalDose-symmetrical
Tilt range±60∘±60∘±60∘
Tilt-series (no.)5/43110137
Data processing
Software tilt-series alignmentIMOD (Kremer et al., 1996)IMOD (Kremer et al., 1996)Dynamo (Castaño-Díez et al., 2012)
Software CTF estimationCTFPLOTTER (Xiong et al., 2009)CTFFIND4 (Rohou and Grigorieff, 2015)CTFFIND4 (Rohou and Grigorieff, 2015)
Particle images (no.)12,910/144,27542,990106,533
Pre-cropped box-size (pix)512600512
Final box-size (pix)192180196
Pixel size final rec. (Å)1.351.351.33
Symmetry imposedC6C6C1
Map resolution (Å)3.2/3.03.53.8
FSC threshold0.1430.1430.143
Map resolution range (Å)3.2–4.3/3.0–3.53.5–4.83.8–7.2
Map sharpening B factor (Å 2)–85 / –95–75–106
Model refinement
Initial model used (PDB code)6T726GNI
SoftwarePHENIX (Afonine et al., 2018)Isolde (Croll, 2018) and PHENIX (Afonine et al., 2018)
Model resolution (Å)3.64.0
FSC threshold0.50.5
Model composition
Non-hydrogen atoms (no.)11,27413,635
Protein residues (no.)14521729
R.m.s. deviations
Bond lengths (Å)0.0010.004
Bond angles (°)0.3220.858
Validation
MolProbity score1.132.01
Clashscore3.4214.71
Poor rotamers (%)00.66
C β outliers (%)00.00
CABLAM outliers (%)0.842.33
Ramachandran plot
Favoured (%)98.895.1
Allowed (%)1.24.8
Disallowed (%)00.1
Appendix 2—table 2
Computational costs and hardware specifics.
HIV-1 GagS-layer inner domainCOPII inner coat
Tilt-series (no.)5/43110137
Final particle images (no.)12,910/144,27542,990106,533
Pre-cropped box-size (pix)512600512
Final box-size (pix)192180196
Computational costs
Pseudo-subtomogram
Compute time21 min/40 min34 min67 min
Number of CPU nodes1/111
Disk space343 GB/3.8 TB777 GB3.1 TB
Refine3D
Compute time18 hr*/33 hr14 hr57 hr
Number of GPU nodes1/111
Ctf refinement
Compute time15 min*/35 min2 hr2 hr
Number of CPU nodes1/111
Disk space32 MB/247 MB621 MB673 MB
Frame alignment
Compute time2 hr*/12 hr2 hr6 hr
Number of CPU nodes1/111
Disk space383 MB/4.1 GB1.9 GB2.2 GB
Hardware specifics
CPU nodes
CPU model2x Intel Xeon E5-2698 v42x Intel Xeon 6258R1x AMD EPYC 7H12
CPU memory512 GB754 GB256 GB
GPU nodes
CPU model2x Intel Xeon Silver 41162x Intel Xeon E5-2667 v41x AMD EPYC 7H12
CPU memory384 GB256 GB256 GB
GPU model2x Nvidia Quadro RTX 50004x Nvidia GeForce GTX 1080 Ti4x Nvidia RTX A6000
  1. *

    These calculations were performed using the same hardware as for the S-layer inner domain.

Additional files

MDAR checklist
https://cdn.elifesciences.org/articles/83724/elife-83724-mdarchecklist1-v3.docx

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Jasenko Zivanov
  2. Joaquín Otón
  3. Zunlong Ke
  4. Andriko von Kügelgen
  5. Euan Pyle
  6. Kun Qu
  7. Dustin Morado
  8. Daniel Castaño-Díez
  9. Giulia Zanetti
  10. Tanmay AM Bharat
  11. John AG Briggs
  12. Sjors HW Scheres
(2022)
A Bayesian approach to single-particle electron cryo-tomography in RELION-4.0
eLife 11:e83724.
https://doi.org/10.7554/eLife.83724