What Could Go Wrong? A Practical Guide to Single-Particle Cryo-EM: From Biochemistry to Atomic Models

Abstract

Cryo-electron microscopy (cryo-EM) has enjoyed explosive recent growth due to revolutionary advances in hardware and software, resulting in a steady stream of long-awaited, high-resolution structures with unprecedented atomic detail. With this comes an increased number of microscopes, cryo-EM facilities, and scientists eager to leverage the ability to determine protein structures without crystallization. However, numerous pitfalls and considerations beset the path toward high-resolution structures and are not necessarily obvious from literature surveys. Here, we detail the most common misconceptions when initiating a cryo-EM project and common technical hurdles, as well as their solutions, and we conclude with a vision for the future of this exciting field.

Keywords: 3D reconstruction; Atomic modeling; Microscope instrumentation; Preferred orientation; Sample heterogeneity; Sample preparation; Sample quality assessment; Single-particle cryo-EM.

Figure 1 -. Molecular weight vs. difficulty for cryo-EM structural targets.

Cryo-EM reconstructions for streptavidin (3.2Å), P-Element transposase (3.8Å), P-Rex1:Gβγ (3.2Å), and TcdA1 (2.8Å) shown alongside difficulty.

Figure 2 -. Particle number vs. resolution.

Shown is a 2D histogram for all asymmetric (C1) reconstructions from 3,302 entries from the EMDB. Most high-resolution (i.e. <4Å) reconstructions contain fewer than 150k particles in the final reconstruction.

Figure 3 -. Negative stain sample screening.

A.Example of a good negative stain image with well-folded protein. B. Positive staining & aggregation. C. Chromatin contamination D. Buffer artifacts E. Detergent micelles F. UF crystals. All examples of stain shown here are for the same particle. Scale bar in all panels are 100 nm.

Figure 4 -. Commonly observed issues in cryo-EM sample preparation.

(A) Ideal sample distribution and contrast for ~180 kDa (left) and ~110 kDa particles (right). In these conditions, we expect that the particles are within a single monolayer within the ice film. (B) Sample denaturation. (C) Crowded sample due to thick ice or sample aggregation. (D) Sample is too dilute. Scale bar is 100 nm.

Figure 5 -. Effect of ice thickness on CTF resolution estimation.

(A) Optimal ice thickness for ~100 kDa particles shows clear particle contrast at 1.18 μm defocus and a CTF fit resolution of 3.0Å. (B) Thicker ice on the same sample shows particles. However, the CTF fit resolution is only 6.0Å. Scale bar is 100 nm.

Figure 6 -. Stage tilt does not overcome single-view, preferred orientation samples.

(A) 3D reconstruction of CPF (cleavage and polyadenylation factor) at 3.5Å using EMPIAR 10299. Note density quality as well as narrow histogram of 3D-FSC. (B-D) Using CPF structure from (A), we simulated a single preferred orientation (D) in addition to tilting the preferred orientation by 30° (C) or 40° (D). Note stretched density and wide-distribution of 3D-FSC angles.

Figure 7.. Effect of map resolution on model building.

Map resolution worse than 4.0 Å results in density errors that appear similar to the protein backbone, increasing the difficulty of model building within this resolution regime. (A) The final model in the 3.6 Å density map, and for comparison, the 4.0 Å (B), 4.2 Å (C), and 4.4 Å (D) density maps. The black and red arrows indicate areas that could be misinterpreted as backbone density at lower (> 4 Å) resolution.