“One View that Rules them All” and Other Vitrification Hurdles in Cryo-EM

How often are proteins designated for single particle analysis by cryo-EM “well behaved”?

We hear this question a lot from clients seeking to understand their chances of getting a map quickly and without complication. And we thought we knew the answer…after all, NanoImaging Services has multiple microscopes, and experts who have tackled a wide range of structural biology projects over many years.

The team weighed in on the percentage of “good samples”: 15%? 20%? 

We were wrong.

When we compiled the data, based on nearly 2 years worth of projects, across a large number of sample types, the answer was less than 8% of samples were “good samples” - meaning they could go quickly (in 2 weeks or less) from initial vitrification trials to a full 3D reconstruction at 3.5 Å resolution or better. The remaining 92% had one or more issues that required a more extensive vitrification condition search and optimization of the data collection and processing parameters in order to be solved.

What did not surprise us was the most common challenge among these samples: preferred orientation. This issue plagued 29% of all samples. (Sample instability upon vitrification and sample heterogeneity, at 22% and 17%, respectively, were also fairly common.) We thought it would be interesting to revisit preferred orientation (an earlier blog post is here) and what can be done to overcome this challenge that hinders almost ⅓ of cryo-EM projects. 

What is preferred orientation, and why is it a problem?

In cryo-EM, the 3D structure of a molecule is reconstructed from 2D projection images collected using a TEM microscope; ideally, the set used for 3D reconstruction contains a large number of different 2D views, reflecting the different orientations of the particle in the ice. These views are sorted and computationally combined to create the 3D reconstruction. (For a 3 minute overview of the process, see this video by Prof. Gabe Lander.) Preferred orientation (PO) refers to the phenomenon observed when, upon vitrification, particles are frozen in only a limited number of unique views. This lack of variation hinders, and sometimes precludes, 3D reconstruction. 

In the large majority of cases, preferred orientation is caused by air-water interface (AWI) effects. The AWI is the area where a solution meets the air, and it can be a hostile environment for proteins and macromolecular complexes. In a typical cryo-EM grid preparation, blotting and plunging into cryogens takes seconds, during which the sample can interact with the AWI hundreds to thousands of times, resulting in PO or in denaturation. 

Cryo-EM teams have developed many ways to combat PO, which can be grouped into three strategies: reducing (or eliminating) the interactions of the protein/macromolecule with the AWI, physically altering the AWI itself, and adopting specific data collection procedures to observe the less common views. We recently undertook a series of experiments to compare different approaches using beta-amylase, a protein that had been used in a prior study of PO. 

• Detergents: detergents can modify the AWI or prevent the protein from interacting with the AWI. Both nonionic and zwitterionic detergents can keep protein particles away from the AWI, though the results vary from sample to sample, with no clear winner. Detergents can also cause other issues, as they may interact with or destabilize the protein itself, or interfere with the vitrification process, resulting in ice that is too thick or too thin. We examined the effects on beta-amylase of several of the most commonly used detergents: Tween-20, Fluorinated OM, OG, Cymal-5, Fluorinated Foscholine 8 (FFC8) and CHAPSO on beta-amylase. FFC8 produced the best results (evaluated as view distribution and map quality); on the other hand, CHAPSO resulted in the worst map, which had a marked decrease in quality and resolution, especially in areas corresponding to surface exposed loops.

• Different grid substrates: the use of continuous support grids (either continuous carbon (CC) or graphene/graphene oxide grids (GO)) in which a continuous layer of supporting material is used to capture the protein (rather than having it suspended a thin layer of ice in holes) has been shown to be useful in addressing the PO. Using these grids, the macromolecule is kept away from the AWI, thus minimizing chances of PO or denaturation. We tested both CC and GO grids with beta-amylase, and in both cases noted an improvement in view distribution. Unfortunately it was accompanied by a decrease in overall resolution of the maps, as a result of the high background introduced  by the continuous layer substantially decreasing the sample’s signal-to-noise ratio, and with it a reduced possibility of correctly picking particles and aligning them in 2D. The utility of continuous support grids is thus directly proportional to the particle size: for larger (>200 kDa) molecules, they can provide a solution for PO, for smaller molecules, the loss in signal-to-noise ratio can be completely debilitating.

• ‍Vitrification methods: increasing the speed at which the sample is vitrified (i.e., reducing the time between sample deposition on the grid and the moment the grid hits the liquid ethane) has been proposed as a possible solution for the AWI effects, since this may greatly reduce the chances of any given particle hitting the AWI. We tested side-by-side grids prepared with a Vitrobot (standard “blot and plunge” method) and grids prepared with an SPT Labtech chameleon (which sprays the sample onto a grid, while the grid is moving toward the liquid ethane). For the Vitrobot, the time between sample deposition and vitrification is in the order of seconds; for the chameleon it can be much faster, in the order of a few 100s of milliseconds. In the beta-amylase set of experiments we conducted, the chameleon grids performed better, with an improved view distribution and a higher quality map.

‍

• ‍Long or tilted data collection(s): of all methods, this approach is the least sample-dependent, and is also the most widely used. Here at NIS, we’ve found over 56% of PO cases can be overcome with this strategy. Collecting a large number of images or tilting the grid at 30-40^o during data collection allows for the identification of rare or unseen views. These approaches don’t come without issues however. First, long data collection may be costly due to limited access to microscope time. Second, with tilted data collection, ice thickness increases, with a reduction of the signal-to-noise ratio and possible loss of resolution, making this approach almost unusable for very small particles. Finally, use of gold grids is also essential when contemplating a tilted data collection, to minimize beam-induced particle motion, as well as having access to more sophisticated (albeit readily available) processing tools such as per particle CTF.

Conclusions

Based on the beta-amylase experiments we performed, as well as on our accumulated experience across a large number of sample classes, we can say that, for now, there is unfortunately no silver bullet to address PO. In fact, in many cases, a combination of approaches (for example, detergents, gold grids, and tilted data collection, or merging data from a tilted data collection and a chameleon grid) is necessary to overcome PO and obtain a set of images that allow for a high resolution 3D map reconstruction. For now, PO (and other vitrification hurdles) must be addressed on a case-by-case basis, and having access to both experienced microscopists as well as to the top of the line vitrification tools remains the best approach to address this complex and common problem.

Schedule your cryo-EM consultation to learn more about how we can help solve your most challenging structures.