Cryo-EM sample preparation for high-resolution structure studies

Practical advice for preparing samples for use in high-resolution structure determination using cryo-electron microscopy is given.

Abstract

High-resolution structures of biomolecules can be obtained using single-particle cryo-electron microscopy (SPA cryo-EM), and the rapidly growing number of structures solved by this method is encouraging more researchers to utilize this technique. As with other structural biology methods, sample preparation for an SPA cryo-EM data collection requires some expertise and an understanding of the strengths and limitations of the technique in order to make sensible decisions in the sample-preparation process. In this article, common strategies and pitfalls are described and practical advice is given to increase the chances of success when starting an SPA cryo-EM project.

1. Introduction

Cryo-electron microscopy (cryo-EM) is a powerful imaging technique used to visualize biological specimens. Depending on the nature of the sample, two main branches exist. (i) If there are many copies of the same macromolecule, single-particle analysis (SPA) cryo-EM can be employed [each region of interest (ROI) is imaged once and the many copies are computationally averaged]. (ii) If each ROI of the specimen is unique, cryo-electron tomography (cryo-ET) can be employed (each ROI is tilted and imaged multiple times). Here, we consider sample preparation for SPA cryo-EM, which is employed to visualize proteins, nucleic acids and viruses at atomic resolution. SPA cryo-EM has experienced dramatic growth in the past decade, marking a ‘resolution revolution’ (Kühlbrandt, 2014 ▸; Assaiya et al., 2021 ▸; Nogales & Scheres, 2015 ▸). Currently, there are more than 31 000 entries in the Electron Microscopy Data Bank (EMDB; https://www.ebi.ac.uk/emdb), with the highest achieved resolution reaching 1.2 Å (Yip et al., 2020 ▸; Nakane et al., 2020 ▸). At this level, not only C, O and N atoms but also H atoms can clearly be resolved. Cryo-EM has revolutionized structural biology and has contributed significantly to our understanding of the molecular architecture of living organisms.

A modern cryo-EM specimen consists of an aqueous film that spans micrometre-sized holes in a thin carbon/gold film that spans across a cryo-EM ‘grid’. In order to remain stable for long periods of time under vacuum, the thin film is frozen after blotting the liquid sample to ∼100 nm thickness or thinner on the grid and rapidly plunging the grid into liquid ethane, which is cooled by liquid nitrogen and maintained at its melting point (Dubochet & McDowall, 1981 ▸; Dubochet et al., 1988 ▸). The specimen is then imaged in a cryo-electron microscope with the temperature maintained at −170°C or lower. The resulting images are processed based on the foundation set by DeRosier & Klug (1968 ▸) using the projection theorem. If the resulting EM density map achieves high resolution (better than 4 Å), which is routine nowadays, atomic models can be built (Fig. 1 ▸).

Figure 1.

A standard workflow for an SPA cryo-EM project is shown. Potential hurdles that require a return to sample preparation are marked with orange arrows, with the diagnosable issue noted to the right of the arrow.

2. Optimizing the sample before vitrification

Before vitrifying a sample onto a cryo-EM grid, the sample should be fully characterized and optimized using alternative biophysical and biochemical techniques, such as those discussed in the review by Byrnes & Mazzorana (2024 ▸). The main considerations for optimizing a cryo-EM sample before vitrification are homogeneity, concentration and buffer composition (including any additives, ligands or detergents used to stabilize a purified biomolecule). It is important to consider that generic sample-preparation advice is given for ideal specimens, but in reality most samples are not ideal and it is important to work under conditions where your sample is well behaved and retains function. It is critical to consider the experimental goals when optimizing the sample, keeping in mind that not all projects aim for atomic resolution reconstructions. There are exceptions to every rule and understanding the reason for sample requirements will help you to achieve the correct balance of ideal versus practical. The major limitation for reconstructing EM density maps using SPA stems from the low signal-to-noise ratio of the images. Individual particles are difficult to visualize because of low signal to noise in the raw data. The ability to increase the signal to noise using computational averaging of many thousands of particles imposes practical limits on the molecular weight and heterogeneity of samples (Henderson, 1995 ▸).

Although cryo-EM structures of a number of sub-100 kDa proteins have been reported, this is still challenging and requires expertise (Lander & Glaeser, 2021 ▸). For targets smaller than 100 kDa, several strategies to increase the size of a particle using tight-binding protein partners have been employed (Wentinck et al., 2022 ▸). If the experimental aim is to achieve high-resolution reconstructions, typically hundreds of thousands to millions of particles are imaged and then sorted in silico, with final reconstructed volumes typically generated from a subset of tens of thousands of the initially picked particles. The resulting resolution will depend on the homogeneity (both conformational and compositional) of the particles averaged in the final set. If sample homogeneity cannot be achieved through biochemical strategies, it is possible to sort heterogeneity in silico during data processing, but the sample should be well characterized by complementary techniques that inform on the type of heterogeneity that exists in the data. There are many examples of heterogeneous sample analysis in the literature, and this is a unique strength of SPA cryo-EM, but these analyses require advanced data processing (Serna, 2019 ▸). While a more homogeneous sample will yield a higher resolution reconstruction, it is important to consider the research question and balance the effort of ‘purification’ between specimen preparation and data collection and analysis. Heterogeneity can reduce the resolution of a 3D reconstruction but may provide exciting biological insight into your molecule of interest.

Once a sample is as homogeneous as possible, the next consideration is concentration. This determines the distribution of particles on a cryo-EM grid. An ideal specimen has a particle density that is as high as possible without overlap of individual particles (Fig. 2 ▸). As a rule of thumb, 1–5 mg ml⁻¹ of the target biomolecule is a good place to start when making standard plunge-frozen cryo-EM grids. The optimal concentration must be determined empirically for each sample. For samples that cannot be concentrated, the use of support films (thin carbon, graphene, graphene oxide or streptavidin monolayers) on cryo-EM grids can facilitate the achievement of a better particle density, although some of these supports introduce additional background noise or processing challenges (Han et al., 2023 ▸).

Figure 2.

Example high-resolution micrographs of various proteins are shown with various particle distributions. The white bars are 200 nm in length. (a) Particles are too dense or crowded. Particles overlap with one another and will not yield good reconstructions. (b) Particles are too dense in the bottom right corner, but the top area with non-overlapping particles has an ideal density. There is also non-severe surface ice contamination (dark black dots). (c) Particles are too sparse, hindering efficient data collection. (d) Non-uniform particle distribution is the result of uneven ice thickness (too thin ice at the top of the image to contain particles). At the transition point of ice thickness, a preferred orientation (the same circular feature) is clearly visible that is not present in the thickest ice (nearest to the bottom of the image). (e) The addition of detergent below the critical micelle concentration (CMC) transforms a crowded sample into a sparse sample.

A final consideration for sample preparation before moving to vitrification is buffer composition. Additives such as glycerol or detergents will increase the background noise and should generally be avoided. However, if additives are needed to keep your sample stable in solution, high-resolution reconstructions can still be achieved with special attention to data analysis (Basanta et al., 2022 ▸).

3. Common strategies for grid optimization

In modern cryo-EM, automated/semi-automated data collection is routine. Typically, a 24 h Titan Krios data collection will generate 4000–6000 movie-frame stacks, which are sufficient to obtain a high-resolution EM structure for a good sample. To achieve this goal, attention should be given to every step in the preparation of cryo-EM samples. To ensure efficient data collection, there should be enough squares with the right ice thickness, uniform ice thickness in each square and appropriate particle density in each hole with high contrast (Fig. 3 ▸). Because time is always limited on high-end microscopes, there is significant value in screening grids to find the best candidates for efficient data collection (Tan et al., 2016 ▸). Strategies and tricks are discussed in this section.

Figure 3.

An example of an optimal cryo-EM sample for automated data collection using the EPU software on a Titan Krios EM. (a) Images of a cryo-sample on holy carbon TEM grids viewed at different magnifications. The goals for an ideal grid are indicated in white text. (b) A view of a Quantifoil (QF) grid square with multiple 2 µm wide holes, colored blue or green for clarity. Holes are grouped for fast data collection using aberration-free image shift (AFIS). The stage is moved only once for each group of holes marked by red and purple polygons and beam-image shift is employed to take images of each hole in the group.

3.1. Grid type

A TEM grid is a 3 mm circular metal (for example copper, gold, nickel or molybdenum) disk with an open mesh design (Figs. 4 ▸ a and 4 ▸ b). The size of these open areas, often represented by squares, is defined by the mesh size, which denotes the number of grid bars per unit length in both the horizontal and vertical directions. For instance, a 300 mesh TEM grid has approximately 300 open areas per inch. The distance from one side of a grid bar to the same side of an adjacent grid bar is about 85 µm. Given that the grid bar width is approximately 20–30 µm, the resulting open square measures about 60 × 60 µm.

Figure 4.

TEM grids and sample preparation for SPA cryo-EM. (a) Cartoon depiction of a TEM grid. (b) A square on a TEM grid. (c) A cross-section of the square shown along the dashed line in (b). (d) 4 µl of an aqueous solution applied onto a TEM grid held by tweezers. (e) A piece of filter paper is applied to blot the excess solution from the TEM grid. The semicircular region seen surrounding the grid is the solution that is wicked from the grid surface. (f) The blotted grid is plunged into liquid ethane (represented by a bright green box) cooled by a bath of liquid nitrogen (LN2; represented by a teal box). (g) The coexistence of solid ethane (red arrow) and liquid ethane is required for good cryo-sample preparation.

When collecting SPA data, beam-image shift is usually employed to increase the number of targets acquired for each slow mechanical sample-stage movement (Cheng et al., 2018 ▸). An example is shown in Fig. 3 ▸(b): holes are grouped and only one stage movement is carried out for each group of holes. To maximize the number of holes per group (per stage movement), 300 mesh TEM grids are recommended and 400 mesh TEM grids are to be avoided. 200 mesh grids can be used, but the foils are more fragile and likely to break during grid handing.

To hold aqueous specimens on TEM grids, a thin carbon film with patterned holes is usually used for SPA cryo-EM (Fig. 4 ▸ c). The size of the hole and the spacing between holes are defined by two numbers: Rx/y. For example, R1/2 means a 1 µm sized hole and a 2 µm spacing from the edge of one hole to the edge of an adjacent hole. The most commonly used grids are R1.2/1.3, R2/2 and R0.6/1 (other sizes can be found on vendor’s websites). These films hold aqueous solution and allow images to be recorded without the carbon film in the background. The hole size and spacing can influence the resulting ice thickness. To choose the best hole size for each sample, the magnification (and thus the field of view) used for imaging and the ability to take multiple images per hole should be considered, along with the hole sizing that gives optimal ice thickness. In some special cases, an ultrathin carbon film or a graphene film is coated over a holey carbon film to help reduce preferred orientation or other distribution problems for some samples. If the sample needs to be tilted inside an EM, ultrathin gold grids (gold foil is used instead of carbon film) are preferred to reduce beam-induced motion (Russo & Passmore, 2014 ▸).

3.2. Glow discharge

Usually, carbon film is hydrophobic, repelling water and preventing the aqueous solution from spreading evenly across the surface. To render carbon film hydrophilic, glow discharge is commonly used. This involves subjecting the grids to a low-pressure plasma discharge in the presence of a suitable gas, typically air, in a glow-discharge unit. A voltage (a few 100 V to a few kV) is applied between two electrodes. This voltage ionizes the gas, causing it to glow, hence the name ‘glow discharger’ (Fig. 5 ▸). The ionized gas creates a plasma within the chamber. The ions and electrons in the plasma bombard the carbon surface, rendering the carbon film hydrophilic.

Figure 5.

Glow discharging renders cryo-EM grids hydrophilic. (a) The commonly used commercially available Pelco Easiglow Glow Discharge Cleaning System. (b) The characteristic glow in the vacuum chamber of the glow discharger confirms the unit’s function of rendering hydrophobic films hydrophilic. (c) Cartoon depiction of the effect of glow discharging. A ball of liquid (shown in blue) will not spread on a hydrophobic carbon surface. It will spread evenly over the support surface after it has been made hydrophilic by glow discharging. (d) Pictures of a 3 mm grid with carbon support film before (top) and after (bottom) glow discharging with 4 µl aqueous solution applied.

With the Pelco Easiglow Glow Discharge Cleaning System (Fig. 5 ▸), the parameters are 20–30 mA and 20 s for carbon grids, 20–30 mA and 120 s for gold grids and 10 mA and 5 s for ultrathin carbon-coated grids. To obtain the best results, the glow-discharged grids should be used within 30 min. If a molecule of interest interacts poorly with the grid foil, there are methods to coat the grids with a positively charged polyelectrolyte such as poly(l-lysine) (Hrebík et al., 2022 ▸), or the use of a different foil material entirely (carbon versus gold) may help.

3.3. Freezing method

The most common method of preparing samples for SPA cryo-EM remains fundamentally the same as was developed 50 years ago (Dubochet & McDowall, 1981 ▸; Adrian et al., 1984 ▸; Dubochet et al., 1988 ▸). Firstly, 3–6 µl of an aqueous solution is applied onto a glow-discharged TEM grid. Excess solution is then blotted from the grid using a piece of filter paper and the grid is subsequently plunged into liquid ethane cooled by liquid nitrogen. Due to the small volume of the aqueous solution left on the grid (less than 1 nl), the freezing speed exceeds 100 000°C s⁻¹, completing the freezing process within a few milliseconds and resulting in the formation of amorphous ice. Once vitrified, the grid is transferred to liquid nitrogen and all subsequent manipulation must be performed below the temperature of devitrification (∼−160°C).

The form of resulting ice and the transformation from amorphous ice to cubic and hexagonal ice depend on the temperature. To ensure that the sample is frozen at the correct temperature, liquid ethane must be at its melting temperature (−183°C; Table 1 ▸). The most reliable temperature indicator is a white solid in the small liquid-ethane container (Fig. 4 ▸ f). The coexistence of solid and liquid ethane defines the melting temperature. As maintaining the correct liquid-ethane temperature requires practice, an alternative has been proposed: a mixture of 63% propane and 37% ethane (Tivol et al., 2008 ▸). This mixture does not freeze at −196°C (the boiling temperature of liquid nitrogen). Consequently, the container of the mixture can be cooled directly by liquid nitrogen and will not solidify. It is essential to ensure direct contact between the alkane-mixture container and the liquid nitrogen.

Table 1. Properties of commonly used cryogens and water.

	Melting point (°C)	Boiling point (°C)	Heat of vaporization (kJ kg−1)	Heat capacity (kJ kg−1 K−1)	Heat to boil (kJ kg−1)	Heat to evaporate (kJ kg−1)	Liquid density (kg m−3)
Nitrogen	−210	−196	6	0.9–1.6	13–22	19–28	809
Ethane	−183	−89	489	2.3–3.5	216–329	705–818	546
Propane	−188	−42	428	1.63	238	666	580
Water	0	100	2257	4.185	418.5	2675.5	1000

Currently, the most common commercially available semi-automated cryo-sample preparation instruments are the Vitrobot from Thermo Fisher Scientific and EM Grid Plunger from Leica. Due to the nature of the blotting process, cryo-sample preparation is not fully reproducible even with these semi-automated commercial plungers, and freezing parameters such as blotting time must be determined empirically. An ideal grid has ice as thin as possible to contain the particle of interest in multiple orientations. Particles in thinner ice will have better signal to noise and thus better data quality (for practical considerations, see Neselu et al., 2023 ▸), but often the particle distribution and orientation will differ in ice of different thicknesses (Fig. 2d ▸). Because of the brute-force nature of standard plunge-freezing, a grid will rarely have uniform ice thickness; thus, multiple regions of a grid with different ice thicknesses should be evaluated during screening (Fig. 3 ▸). If there are regions of thin ice with good particle distribution, the next quality metric to consider is how many of these regions are available for imaging on one grid. An ideal grid will have tens of grid squares with similarly ‘good ice’ (Fig. 3 ▸). If your grid meets these standards, then you are ready for data collection on a high-end microscope.

Before collecting a large data set, it is helpful to produce duplicates or triplicates of cryo-grids using the same freezing parameters and make two extra cryo-grids with blotting times 1 s longer and shorter than the optimal time. These two extra grids are in case the optimal freezing parameters give ice that is too thin or too thick. The duplicates/triplicates are used for screening and high-resolution data collection. Often, the screening of cryo-grids is carried out using low-end side-entry EMs. Although it is possible to retrieve a grid and use it for high-resolution data collection on another EM, it is inevitable that there will be ice contamination on the grid. Thus, it is strongly recommended to make duplicates/triplicates for SPA cryo-EM studies. A key strategy to mitigate other sources of ice contamination is to reduce grid manipulations and perform all grid manipulations as quickly as possible.

Handling grids quickly is not intuitive for most hands, but develops with practice. It is critical to work with clean liquid nitrogen. Ensure that liquid nitrogen is dispensed from tanks with dry hoses into dry transfer dewars. Working in a de­humidified environment, if available, will reduce ice contamination. Finally, all tools should be kept as dry as possible by not transferring them into and out of liquid nitrogen or using them for prolonged periods. It is good practice to warm and dry any tools that start to collect visible ice with heat blocks or hair dryers before continuing to use them.

4. Additional challenges

If particles become aggregated, unfolded or are not visible upon vitrification, and complementary methods have been used to validate that this is occurring after application of the sample to the grid, there are multiple strategies to troubleshoot. Parameters that can be varied include the protein concentration, the addition of additives (for example surfactants, detergents or protein ligands), the addition of binding partners and cross-linking. While it has successfully been used, cross-linking is not straightforward and should be considered as a last resort (Stark, 2010 ▸). Cross-linking efficiency should be characterized using complementary methods (SDS–PAGE, mass photometry etc.) before putting your cross-linked sample onto a grid. If this short review leaves you wanting further advice, other great reviews are available (Cianfrocco & Kellogg, 2020 ▸; Passmore & Russo, 2016 ▸; Glaeser et al., 2021 ▸).

A major issue for cryo-EM grid preparation is that the number of troubleshooting options is greater than the amount of time available on an EM to evaluate all of them. In this case, it is useful to scan the literature for the freezing conditions of similar samples or to reach out to more experienced cryo-EM users and ask for advice. In addition to the standard plunge-freezers, there are a number of new grid-making methods and instruments that can be considered (Weissenberger et al., 2021 ▸)

Finally, after optimizing the overall grid quality, ice thickness and particle distribution, there are still issues that can arise that are only apparent after initial data processing. One common issue is preferred orientation. When particles are only seen from a few orientations, 3D reconstruction suffers from artifacts. For some samples, preferred orientation is clearly visible by eye (Fig. 2d ▸). It is otherwise diagnosable by seeing only a few views in 2D classification. For more symmetric or globular specimens, screening will require the collection of enough micrographs to process through 3D reconstruction to diagnose whether the orientation issues have been resolved. Such issues can sometimes be identified after 1–2 h of data collection if on-the-fly processing is used (Mendez et al., 2023 ▸). If such issues are apparent, data collection can be stopped or altered (for example using tilted data collection) to make the best use of the microscope time. While tilted data collection is an option for overcoming effects of preferred orientation in 3D reconstruction (Tan et al., 2017 ▸), tilting a specimen reduces the signal to noise and results in increased beam-induced motion; thus, alternative grid preparation is often needed to achieve the best possible resolution (Drulyte et al., 2018 ▸). Common strategies, as already discussed above, are also useful for achieving different and, if you are lucky, more diverse orientation distributions. If different grid-preparation methods lead to differing preferred orientation issues, combining data from grids prepared using multiple strategies can be useful (Klebl et al., 2020 ▸; Xu et al., 2023 ▸).

5. Concluding thoughts

Sample preparation is a hurdle in every SPA cryo-EM project. While there are common troubleshooting strategies (Table 2 ▸), it is difficult to devise a single standard workflow. Even with a biochemically ideal sample, issues encountered during vitrification sometimes require a return to the bench to optimize samples (Fig. 1 ▸). Trial and error and a little bit of luck are all part of cryo-EM sample preparation. We also highly recommend asking the advice of more experienced groups or instrumentation facility staff. Time spent doing biochemistry to make a sample as homogeneous as possible can save a significant amount of time during grid preparation and data processing. Time spent at the grid-preparation stage can make data collection on high-demand instruments more efficient. When preparing cryo-grids, meticulous planning is crucial. Ensuring that all necessary tools, supplies and samples are ready, designing an efficient workflow and conducting the freezing process swiftly contribute to obtaining high-quality cryo-EM grids for cryo-EM structure determination.

Table 2. Common troubleshooting steps for cryo-EM grid preparation.

Issue	Potential solution
Non-vitreous ice	Check cryogen temperature, cool down tools before touching cryo-grids, change holder temperature
Thick ice	Change glow-discharge or blotting parameters
Low particle density	Increase concentration or use support films or affinity grids if concentration cannot be increased
Non-uniform particle distribution in ice	Change support films, change glow-discharge and/or blotting parameters, addition of detergents or additives
Ice contamination	Clean liquid nitrogen, shorten steps, perform steps in humidity-controlled room if possible, more careful handling
Preferred orientation	Different samples, different substrate (graphene or ultrathin carbon-coated grids), different grid-making technique, addition of detergents, tilted data collection
Heterogenenous sample	Careful data processing to understand heterogeneity. If experimental goals are still not met: additional purification, use of ligand binding partners or additives to stabilize states.

Funding Statement

The Laboratory for BioMolecular Structure (LBMS) is supported by the DOE Office of Biological and Environmental Research (KP1607011). The National Center for CryoEM Access and Training (NCCAT) is supported by the National Institutes of Health Common Fund Transformative High-Resolution Cryo-Electron Microscopy program (U24 GM129539) and is located at the Simons Electron Microscopy Center at the New York Structural Biology Center, supported by grants from the Simons Foundation (SF349247) and NY State Assembly.