Locating macromolecular assemblies in cells by 2D template matching with cisTEM
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
Lucas, Himes et al. present multiple practical improvements to the 2D high-resolution template-matching (2DTM) routine for cryo-EM images originally described by Rickgauer et al., eLife 2017. GPU-acceleration and integration into cisTEM make the approach substantially faster and easier to use than the previous CPU-based Matlab implementation. The strengths and weaknesses of the 2DTM are clearly presented and the comparison with 3DTM is thorough. At present the 2DTM approach is likely only suitable for analysis of large assemblies (e.g., ribosomes, proteasomes,etc.) in situ, but future improvements in microscope hardware and the 2DTM routine itself will likely allow application of this approach to smaller complexes.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their names with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Cell Biology (eLife)
Abstract
For a more complete understanding of molecular mechanisms, it is important to study macromolecules and their assemblies in the broader context of the cell. This context can be visualized at nanometer resolution in three dimensions (3D) using electron cryo-tomography, which requires tilt series to be recorded and computationally aligned, currently limiting throughput. Additionally, the high-resolution signal preserved in the raw tomograms is currently limited by a number of technical difficulties, leading to an increased false-positive detection rate when using 3D template matching to find molecular complexes in tomograms. We have recently described a 2D template matching approach that addresses these issues by including high-resolution signal preserved in single-tilt images. A current limitation of this approach is the high computational cost that limits throughput. We describe here a GPU-accelerated implementation of 2D template matching in the image processing software cis TEM that allows for easy scaling and improves the accessibility of this approach. We apply 2D template matching to identify ribosomes in images of frozen-hydrated Mycoplasma pneumoniae cells with high precision and sensitivity, demonstrating that this is a versatile tool for in situ visual proteomics and in situ structure determination. We benchmark the results with 3D template matching of tomograms acquired on identical sample locations and identify strengths and weaknesses of both techniques, which offer complementary information about target localization and identity.
Article activity feed
-
-
Evaluation Summary:
Lucas, Himes et al. present multiple practical improvements to the 2D high-resolution template-matching (2DTM) routine for cryo-EM images originally described by Rickgauer et al., eLife 2017. GPU-acceleration and integration into cisTEM make the approach substantially faster and easier to use than the previous CPU-based Matlab implementation. The strengths and weaknesses of the 2DTM are clearly presented and the comparison with 3DTM is thorough. At present the 2DTM approach is likely only suitable for analysis of large assemblies (e.g., ribosomes, proteasomes,etc.) in situ, but future improvements in microscope hardware and the 2DTM routine itself will likely allow application of this approach to smaller complexes.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the …
Evaluation Summary:
Lucas, Himes et al. present multiple practical improvements to the 2D high-resolution template-matching (2DTM) routine for cryo-EM images originally described by Rickgauer et al., eLife 2017. GPU-acceleration and integration into cisTEM make the approach substantially faster and easier to use than the previous CPU-based Matlab implementation. The strengths and weaknesses of the 2DTM are clearly presented and the comparison with 3DTM is thorough. At present the 2DTM approach is likely only suitable for analysis of large assemblies (e.g., ribosomes, proteasomes,etc.) in situ, but future improvements in microscope hardware and the 2DTM routine itself will likely allow application of this approach to smaller complexes.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1 and Reviewer #2 agreed to share their names with the authors.)
-
Reviewer #1 (Public Reviews):
This paper describes a 2D approach to a problem of identifying macromolecular complexes in cryo-ET. Surprisingly, the authors argue that the approach is more sensitive than 3D approaches, and is computationally much faster. While it is not possible to prove that the method will be superior to all possible 3D approaches, the current implementation will be a useful tool for many people.
-
Reviewer #2 (Public Review):
Lucas, Himes et al. present multiple practical improvements to the 2D high-resolution template-matching (2DTM) routine for cryo-EM images originally described by Rickgauer et al., eLife 2017. Moreover, the authors assess the 2DTM approach for macromolecular identification in situ using the example of the M. pneumoniae ribosome and compare it to the conventional 3D low-resolution template-matching (3DTM) approach followed by subtomogram averaging in tomograms of the same areas. Implementation of GPU-acceleration and integration into cisTEM make the approach substantially faster and easier to use than the previous CPU-based Matlab implementation. The strengths and weaknesses of the 2DTM are clearly presented and the comparison with 3DTM is thorough. At present the 2DTM approach is likely only suitable for …
Reviewer #2 (Public Review):
Lucas, Himes et al. present multiple practical improvements to the 2D high-resolution template-matching (2DTM) routine for cryo-EM images originally described by Rickgauer et al., eLife 2017. Moreover, the authors assess the 2DTM approach for macromolecular identification in situ using the example of the M. pneumoniae ribosome and compare it to the conventional 3D low-resolution template-matching (3DTM) approach followed by subtomogram averaging in tomograms of the same areas. Implementation of GPU-acceleration and integration into cisTEM make the approach substantially faster and easier to use than the previous CPU-based Matlab implementation. The strengths and weaknesses of the 2DTM are clearly presented and the comparison with 3DTM is thorough. At present the 2DTM approach is likely only suitable for analysis of large assemblies (e.g., ribosomes, proteasomes,etc.) in situ, future improvements in microscope hardware and the 2DTM routine itself will likely allow application of this approach to smaller complexes.
A point of concern is the degree of reference-bias in the results of the 2DTM approach. The authors acknowledge this concern and that conventional use of the FSC is not a suitable validation metric for this approach nor for determining an appropriate filtering cutoff for a resulting reconstruction. The proposed validation metric of the emergence of additional known density features in a reconstruction, which are not present in the template, resulting from 2DTM hits is sensible. However, emergence of additional unknown densities in a reconstruction resulting from cellular data will be difficult to segregate from noise, especially since filtering of the reconstruction is determined ad hoc instead of by an objective metric.
Nevertheless, the implementation of 2DTM is a major step forward in molecular identification in a crowded cellular environment. Even for ribosomes, 3DTM coupled with subtomogram averaging can be a time-consuming process and false-positives can persist despite extensive classification. The complementary approach presented by the authors of acquiring a nominally untitled frame-series for 2DTM that is followed by a conventional tilt-series for 3DTM of the same area could be particularly well-suited to answer questions that require an accurate "molecular census" and/or attempting a hybrid subtomogram averaging approach.
-
Reviewer #3 (Public Review):
The authors have implemented GPU-accelerated 2D template matching for localization and identification of macromolecules in projection images and apply it to ribosomes in M.pneumoniae cells. They optimize parameters of the workflow and compare it to 3D template matching in volumetric data. The interesting outcome of the study is that the 2D approach has higher specificity than the more time-consuming 3D strategy.
Strengths: This work provides an experimentally and computationally fast workflow to obtain structures from whole cells. Some efforts are made to assess specificity and sensitivity of the approach, which nevertheless remain somewhat qualitative in the absence of a ground truth. The resolution assessment figures suggest that reconstructions of relatively high resolutions have been obtained. The claim …
Reviewer #3 (Public Review):
The authors have implemented GPU-accelerated 2D template matching for localization and identification of macromolecules in projection images and apply it to ribosomes in M.pneumoniae cells. They optimize parameters of the workflow and compare it to 3D template matching in volumetric data. The interesting outcome of the study is that the 2D approach has higher specificity than the more time-consuming 3D strategy.
Strengths: This work provides an experimentally and computationally fast workflow to obtain structures from whole cells. Some efforts are made to assess specificity and sensitivity of the approach, which nevertheless remain somewhat qualitative in the absence of a ground truth. The resolution assessment figures suggest that reconstructions of relatively high resolutions have been obtained. The claim that detection specificity is higher in 2D than in 3D is surprising and interesting.
Weaknesses: The work remains on an empirical level as surprising advantages of the 2D approach compared to 3D are revealed, but there is little effort to get to the basis of these observations. Moreover, details on the compared 3D approach (and its parameter optimization analogous to the 2D approach), which the rather general conclusions would require, are missing. Lastly, the 3D approach has been applied to the strongly pre-irradiated sample, which may make observations such as a lower specificity in the 3D case almost a self-fulfilling prophecy. Thus, the 2D vs 3D comparison is not convincing in the current form.
In summary, the 2D implementation of in situ structure determination is interesting and of potential interest to a large audience. However, the comparison to the 3D equivalent appears somewhat incomplete and the rather general conclusions require further validation.
-
