Optimising the tilt increment for in situ cryo-electron tomography
- Department of Molecular Sociology, Max Planck Institute of Biophysics, Germany
- Cluster of Excellence SubCellular Architecture of Life (SCALE), Goethe University Frankfurt, Germany
- Institute of Biochemistry, Goethe University Frankfurt, Germany
Figures
Figure 1 with 13 supplements
Representative tilt images and tomograms at varying tilt increment.
(A) Single projection images of the sample at effective zero-tilt from a tilt series acquired with a tilt increment of 1° (i); 2° (ii); 3° (iii); 5° (iv); and 10° (v). (B) Central XY slices through the tomograms reconstructed from the tilt series shown in (A). (C) Central XZ slices through the tomograms shown in (B). White arrowheads point to membranes that appear to extend outside of the lamella. (D) Power spectra of representative tomograms for all five conditions. Yellow circles represent the Crowther criterion for the tomograms shown in (B) and (C), with the numerical value shown as text in yellow. Scalebar is 200 nm and applies to all relevant panels.
Figure 1—figure supplement 1
Overview of acquired, removed and remaining tilt images for each dataset.
For each dataset, the percentage of tilt images is shown per category: images contributing to the final tomogram (green), images that were removed due to drift, obstruction or other causes (red), and images that were not acquired due to SerialEM’s automated abort criteria (orange).
Figure 1—figure supplement 2
Overview of distinct features in tilt-images and tomogram slices.
(A–E) Columns represent the 1° (A), 2° (B), 3° (C), 5° (D), and 10° (E) datasets. Within each column, the central tilt-image is shown on the left and a slice through the tomogram at the same position is shown on the right. The rows represent different features, from top to bottom: (i) ribosomes, (ii) mitochondria outer membrane, (iii) microtubule, (iv) nuclear pore complex, and (v) VAULT protein. Scalebar is 50 nm and applies to both panels.
Figure 1—figure supplement 3
Artefacts in tomograms acquired with 10° tilt increment.
(A) XY slice of a tomogram, near the lamella surface, reconstructed from a tilt series with a 10° increment. Arrowheads point to artefacts in the reconstruction in the region close to strongly curved mitochondrial membranes. (B) Central XZ slice of the same tomogram shown in (A), reconstructed with extended Z-height. The region where the lamella is positioned is indicated with yellow lines. Membranes that extend far beyond the lamella are indicated with white arrowheads. Scalebar is 200 nm and applies to both panels.
Figure 1—video 1
Representative tilt series acquired with a tilt increment of 1°, as depicted in Figure 1.
Figure 1—video 2
Representative tilt series acquired with a tilt increment of 2°, as depicted in Figure 1.
Figure 1—video 3
Representative tilt series acquired with a tilt increment of 3°, as depicted in Figure 1.
Figure 1—video 4
Representative tilt series acquired with a tilt increment of 5°, as depicted in Figure 1.
Figure 1—video 5
Representative tilt series acquired with a tilt increment of 10°, as depicted in Figure 1.
Figure 1—video 6
Representative tomographic volume of data acquired with a tilt increment of 1°, as depicted in Figure 1.
Figure 1—video 7
Representative tomographic volume of data acquired with a tilt increment of 2°, as depicted in Figure 1.
Figure 1—video 8
Representative tomographic volume of data acquired with a tilt increment of 3°, as depicted in Figure 1.
Figure 1—video 9
Representative tomographic volume of data acquired with a tilt increment of 5°, as depicted in Figure 1.
Figure 1—video 10
Representative tomographic volume of data acquired with a tilt increment of 10°, as depicted in Figure 1.
Figure 2 with 2 supplements
Assessment of signal-to-noise ratio (SNR), tilt-alignment quality, and CTF fitting parameters at different tilt increments.
(A) SNR measured from projection images of the sample at effective zero-tilt position. SNR was defined as the ratio of the squared mean to the squared standard deviation of each image. (B) SNR measured from the sample volume that represents the lamellae. As the tomogram SNR distribution of particularly the 2° and 3° dataset showed similar SNR distributions, we performed formal significance testing for the data in this panel (see B). Kruskal–Wallis test confirmed significant differences across conditions (H=270.97, p<0.001); pairwise Mann–Whitney U tests with Bonferroni correction revealed all pairs were significantly different except 2° vs. 3° (p=0.093), indicating comparable SNR for these two tilt increments. (C) Residual alignment error after tilt-series alignment in Etomo. (D, E) CTF estimation parameters. (D) Figure of merit, which indicates the confidence level to which the CTF was fitted, and (E) the final resolution to which the CTF was fitted; both estimated with Gctf on projection images of the sample at effective zero-tilt position. (F) Scatter plot showing the final resolution as shown in (E) against the local lamella thickness for all conditions, with coefficients of determination of R2=0.30, 0.38, 0.66, 0.61, and 0.60 for the respective datasets.
Figure 2—figure supplement 1
Local lamella thickness per condition.
(A) Local lamella thickness for all tomograms used in this study. (B) Local lamella thickness for all tomograms used for the TM and STA analyses. Owing to the poor tomogram reconstruction quality observed for the 10° tilt increment data, lamella thickness could not be determined for each tomogram for this condition.
Figure 2—figure supplement 2
CTF estimation parameters from CTFFIND4.
(A) CTFFIND4 CC score, which indicates the confidence level to which the CTF was fitted. (B) Maximum fitted resolution of the CTF using CTFFIND4. For clarity, only the images of the untilted specimen were used in the analysis.
Figure 3 with 4 supplements
The effect of tilt increment on template matching.
(A) Classes after the first round of 3D classification. (B) Maximum F1-score for tomograms used for TM and STA. (C) Area under the precision–recall curves. See corresponding Figure 3—figure supplement 4.
Figure 3—figure supplement 1
Template matching scores.
(A–E) Volumes with template matching scores were converted to z-scores. For visualisation purposes, here only a subvolume of 500 × 500 × 400 voxels were isolated from representative volumes. Panels (A–E) represent results for 1°, 2°, 3°, 5°, and 10° increments, respectively. A maximum intensity projection was performed along the last dimension (z) and plotted (left). An overlay the scores and the tomogram of the same region (middle) and a histogram of the score volume (right). Scalebar shown in panel (A) is 100 nm and applies to all relevant panels.
Figure 3—figure supplement 2
3D classification results after high-confidence TM.
Resulting classes after 3D classification with the intention to remove junk particles, performed on subtomograms with binning factor 6, corresponding to a voxel size of 11.8 Å, for tilt increments of 1° (A), 2° (B), and 3° (C). Z-score thresholds of 6, 6, and 5.5 were used for the 1°, 2°, and 3° data, respectively. The total number of particles is indicated above the classes, as well as the number of particles per class below the class. No significant number of junk particles could be detected for these datasets.
Figure 3—figure supplement 3
3D classification results after TM to remove junk particles.
Resulting classes after 3D classification to remove junk particles, performed on subtomograms with binning factor 6, corresponding to a voxel size of 11.8 Å/px, for tilt increments of 1° (A), 2° (B), 3° (C), 5° (D), and 10° (E). Z-score thresholds of 5 were used for all data, apart from the 10° data, where a threshold value of 3.75 was used. For the 10° dataset, a binning factor of 2 is used, corresponding to a voxel size of 3.9 Å. The total number of starting particles is indicated at the top. The number of particles per class is indicated below each class. Successive classification rounds are visible as rows. Particle classes that were discarded have been marked with a red arrow, particle classes that were taken to the next round of classification are indicated with a downward orange arrow, and particle classes that were accepted as true positives are indicated with a green check mark. The number of accepted particles after 3D classification is indicated below, as well as the percentage with respect to the number of starting particles.
Figure 3—figure supplement 4
F1- and PR-curves for all tomograms.
Each line represents a single tomogram used in the analysis. (A–E) F1-curves for tomograms used for TM for 1°, 2°, 3°, 5°, and 10° increments, respectively. On the horizontal axes, the z-score threshold used for peak extraction is plotted and on the vertical axes the F1-score for particles extracted using that threshold. (F–J) Precision–recall curves for the same data.
Figure 4 with 1 supplement
The effect of tilt increment on STA.
(A) Rosenthal–Henderson plots of particle subsets when refined and averaged in RELION 3.1. (B) Rosenthal–Henderson plots of tomogram subsets when refined and averaged in M. Inset: ribosome density maps using all particles, coloured according to the local resolution. (C) Resolution progression for each M refinement, considering all available particles. The grid subdivision shown on the horizontal axis is used for both the Image warp and the Volume warp grids. For more details, see the ‘Methods’ section. (D) Resolution gain for each refinement round, also shown in panel (C).
Figure 4—figure supplement 1
Fourier shell correlation curves.
(A) FSC curve for all tilt increments after refinement in RELION. (B) FSC curve for all tilt increments after refinement in M. Thresholds at FSC of 0.5 and 0.143 are indicated with dashed horizontal lines.
Author response image 1
Tables
Table 1
Datasets and acquisition parameters used in this study.
| Tilt increment | Number of tilts | Total dose (e-/Å2) | Dose/tilt (e-/Å2) | Acquisition time (min) | File size | Tilt series acquired | Pixel size (Å) |
|---|---|---|---|---|---|---|---|
| 1° | 121 | 120–130 | 1.1 | 65 | 3.8 Gb | 64 | 1.971 |
| 2° | 61 | 120–130 | 2.0 | 25 | 2.0 Gb | 113 | 1.971 |
| 3° | 41 | 120–130 | 3.2 | 20 | 1.3 Gb | 128 | 1.971 |
| 5° | 25 | 120–130 | 4.8 | 14 | 800 Mb | 122 | 1.971 |
| 10° | 13 | 120–130 | 9.2 | 9 | 417 Mb | 89 | 1.971 |
Key resources table
| Reagent type (species) or resource | Designation | Source or reference | Identifiers | Additional information |
|---|---|---|---|---|
| Strain, strain background (Dictyostelium discoideum) | Ax2-214 | dictyBase, Depositor Guenter Gerisch | DBS0235534 | |
| Chemical compound, drug | HL5 medium | Formedium | HLB0102 | |
| Chemical compound, drug | Ampicillin | Formedium | A9518 | |
| Chemical compound, drug | Geneticin G418 | Sigma-Aldrich | G5013 | |
| Software, algorithm | SerialEM (4.0.6, 4.0.10 & 4.0.20) | Mastronarde, 2005 | RRID:SCR_017293 | https://bio3d.colorado.edu/SerialEM/ |
| Software, algorithm | IMOD (4.11.5) | Kremer et al., 1996 | RRID:SCR_003297 | https://bio3d.colorado.edu/imod/ |
| Software, algorithm | Matlab (2019b) | The MathWorks, Inc | RRID:SCR_001622 | https://www.mathworks.com/products/matlab.html |
| Software, algorithm | Gctf (v1.06 2016-05-22) | Zhang, 2016 | RRID:SCR_016500 | https://github.com/ProteinGod/Gctf |
| Software, algorithm | Ctffind 4.1 | Rohou and Grigorieff, 2015 | RRID:SCR_016732 | https://grigoriefflab.umassmed.edu/ctf_estimation_ctffind_ctftilt |
| Software, algorithm | WARP (1.0.9) | Tegunov and Cramer, 2019 | RRID:SCR_018071 | https://github.com/warpem/warp |
| Software, algorithm | Relion 3.1 | Zivanov et al., 2018 | RRID:SCR_016274 | https://github.com/3dem/relion |
| Software, algorithm | M (1.0.9) | Tegunov et al., 2021 | https://github.com/warpem/warp | |
| Software, algorithm | STOPGAP (0.7.1) | Wan, 2023; Wan et al., 2024 | https://github.com/wan-lab-vanderbilt/STOPGAP | |
| Software, algorithm | GAPSTOP (v0.3) | Cruz-León et al., 2024; Wan et al., 2024 | https://gitlab.mpcdf.mpg.de/bturo/gapstop_tm; Turoňová, 2025 | |
| Software, algorithm | ChimeraX (1.6) | Meng et al., 2023; Pettersen et al., 2021 | RRID:SCR_015872 | https://www.cgl.ucsf.edu/chimerax/ |
| Software, algorithm | cryoCAT | Turoňová, 2024b | https://github.com/turonova/cryoCAT | |
| Software, algorithm | Adobe Illustrator 2022 | Adobe | https://www.adobe.com/ | |
| Software, algorithm | Python (3.9.7) | https://www.python.org/ | https://www.python.org/downloads/release/python-397/ | |
| Software, algorithm | SciPy (1.13.1) | https://scipy.org/ | RRID:SCR_008058 | https://github.com/scipy/scipy |
| Software, algorithm | Numpy (1.26.4) | https://numpy.org/ | RRID:SCR_008633 | https://github.com/numpy/numpy |
| Software, algorithm | Matplotlib (3.9.4) | https://matplotlib.org/ | RRID:SCR_008624 | https://github.com/matplotlib/matplotlib |
| Software, algorithm | Seaborn (0.13.2) | Waskom, 2021 | RRID:SCR_018132 | https://seaborn.pydata.org/ |
| Software, algorithm | starparser (v1.38) | Chabaan | https://github.com/sami-chaaban/starparser; Chaaban, 2022a | |
| Other | Pelco easiGlow Glow Discharger Cleaning System | Ted Pella, Inc | easiGlow | Used for hydrophilisation of EM grids prior to vitrification |
| Other | Quantifoil R 1/4, 200 Mesh, Au, SiO2 film | Quantifoil | EM grids and support film | |
| Other | Whatman filter paper #1 | Whatman | WHA1001329 | Blotting paper during vitrification |
| Other | EM GP2 Automatic Plunge Freezer | Leica Microsystems | Leica GP2 | Used for plunge-freezing |
| Other | Aquilos dual beam cryo-FIB/SEM | Thermo Fisher Scientific | Aquilos cryo-FIB | Used for the cryo-FIB milling of lamellae |
| Other | Titan Krios G4 (300 kV cryo-transmission electron microscope) | Thermo Fisher Scientific | Titan Krios | Used for cryo-ET data acquisition, see ‘Methods’ |
| Other | Falcon 4 direct electron detector with Selectris X imaging filter | Thermo Fisher Scientific | F4 with selectris X | Direct electron detector and energy filter use for data acquisition, see ‘Methods’ |
Table 2
Parameters of M refinement rounds.
| Geometry | Tilt series | CTF | ||||
|---|---|---|---|---|---|---|
| M refinement | Image warp grid | Particle poses | Stage angles | Volume warp grid | Defocus | Tilt series acquired |
| Round 1 | 3 × 3 | ✓ | ✓ | 3 × 3 × 2 × 10 | - | - |
| Round 2 | 6 × 6 | ✓ | ✓ | 6 × 6 × 8 ×10 | - | - |
| Round 3 | 6 × 6 | ✓ | ✓ | 6 × 6 × 8 × 10 | ✓ | ✓ |
| Round 4 | 10 × 10 | ✓ | ✓ | 10 × 10 × 10 × 10 | ✓ | ✓ |
Author response table 1
| 1° | mean=0.0612 | std=0.0259 | n=63 |
| 2° | mean=0.0406 | std=0.0091 | n=119 |
| 3° | mean=0.0381 | std=0.0135 | n=119 |
| 5° | mean=0.0322 | std=0.0117 | n=122 |
| 10° | mean=0.0145 | std=0.0055 | n=89 |
Additional files
-
MDAR checklist
- https://cdn.elifesciences.org/articles/111639/elife-111639-mdarchecklist1-v1.docx
-
Source code 1
Jupyter notebook used to pseudo-randomly select particles based on whole tomograms for M refinements.
- https://cdn.elifesciences.org/articles/111639/elife-111639-code1-v1.zip
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)
Optimising the tilt increment for in situ cryo-electron tomography
eLife 15:RP111639.
https://doi.org/10.7554/eLife.111639.3
