Computational methods for in situ structure determination with cryo-electron tomography

Session

LST.3 - Advances in Image Processing in Biological Electron Microscopy

Authors

Dr. Muyuan Chen (1), Dr. Zhao Wang (1), Dr. Steven Ludtke (1)

Affiliations

1. Baylor College of Medicine

Keywords

CryoET, cellular tomography, subtomogram averaging

Abstract text

With electron cryotomography (CryoET), biologists can visualize cells in 3D and localize macromolecules within cells at nanometer resolutions [1]. While modern instruments produce massive amounts of tomography data containing extremely rich structural information, it takes tremendous effort to process those data and the results are often limited by the skill of the personnel. Here we present an integrated workflow that covers the entire CryoET data processing pipeline, which greatly reduces human effort and increases the throughput of the process [2]. 

The protocol starts from automated tilt series alignment using an iterative 3D landmark-based approach, and a Fourier reconstruction routine to generate tomograms from aligned tilt series. To identify the target protein from the crowded cellular environment, a convolution neural network based model is developed that significantly reduces human intervention at the particle selection step. After particles are located, CTF correction is performed at the per-particle-per-tilt level, taking the geometry of the tilt and depth of particle within the ice into consideration. Initial models can be generated directly from the particles without external references using a stochastic gradient descent algorithm, which is followed by a fast subtomogram refinement protocol that uses hierarchical orientation search and missing wedge compensation. To further boost the resolution of the averaged structures, a subtilt refinement routine is developed to correct for particle motion during tilt series imaging, and focused refinement methods are implemented to resolve local structural flexibility of the target protein.

To demonstrate the performance of the workflow, we present structures of a drug efflux pump determined on native bacteria membranes at subnanometer resolution. The structures show conformational differences compared to in vitro structures, as well as domains that were previously invisible in purified samples [3]. 

References

[1] Asano, S., Engel, B. D. & Baumeister, W. In Situ Cryo-Electron Tomography: A Post-Reductionist Approach to Structural Biology. J. Mol. Biol. 428, 332–343 (2016).

[2] Chen, Muyuan, et al. "A complete data processing workflow for cryo-ET and subtomogram averaging." Nature methods (2019): 1-8.

[3] This work was partially supported by NIH grants R01GM080139, P01GM121203, and Welch Foundation (Q-1967-20180324). We also would like to thank early users for testing the workflow and providing valuable feedback.