Vitrobot-based protocol makes graphene oxide cryo-EM grids reproducible and accessible

Cryo-electron microscopy has transformed structural biology, but a stubborn bottleneck has remained at the very first step of every experiment: the grid. Researchers report a new method for preparing graphene oxide-supported cryo-EM grids using the Vitrobot Mark IV, a plunge-freezing instrument already present in most cryo-EM laboratories, that delivers an average graphene oxide coverage of about 55 percent, with monolayer and bilayer regions accounting for 40 percent. The work, published in BMC Methods, describes a protocol that requires no specialized deposition equipment and no liquid nitrogen or ethane during the coating step itself.

The motivation behind the method lies in the well-known pathologies of cryo-EM sample preparation. When protein solutions are applied to unsupported EM grids, particles often adsorb onto the foil material, such as gold or carbon, resulting in sparse distribution within open holes or aggregation near hole edges. More seriously, particles interact with the air-water interface, a hydrophobic surface that can denature proteins, disrupt macromolecular complexes, or force particles into preferential orientations, particularly in thin ice layers. Beam-induced motion adds a further obstacle: stress accumulated during vitrification can be released under electron radiation, causing ice doming, while the insulating nature of vitreous ice allows charge buildup that distorts image trajectories. Graphene oxide addresses these problems because it reduces the air-water interface area, provides an alternative surface for particles to adsorb to, and, thanks to graphene’s electrical conductivity and mechanical strength, creates more stable imaging conditions.

Graphene oxide is transparent to electrons and introduces no extra background noise, making it an ideal support material. Cryo-electron tomography and single-particle reconstructions have previously demonstrated that particles preferentially adsorb onto the graphene oxide surface rather than the air-water interface in holey grids, preserving macromolecular integrity and minimizing preferential orientation effects. The thickness of the vitrified ice layer is also critical: ice that is too thin may exclude or denature particles, while excessively thick layers obscure structural detail. Because blotting paper wicks buffer unevenly due to its irregular fibers, ice thickness can vary unpredictably. Graphene oxide-supported grids provide a flat, consistent surface that promotes uniformly thin vitreous ice, improving sample preservation and the signal-to-noise ratio.

Although graphene and graphene oxide grids are commercially available from several vendors, the range of materials and hole sizes on offer is limited, and surface assembly deposition methods typically require specialized equipment not found in conventional biochemistry laboratories. The drop-cast method is simpler, but existing protocols suffer from inconsistent quality due to variations in laboratory conditions and user interpretation. Building on the approach pioneered by Avinash Patel and colleagues, the new protocol applies a layer of polyethyleneimine, a polycationic polymer, after glow-discharging the grids, so that the positively charged surface attracts the negatively charged graphene oxide flakes far more effectively than van der Waals forces alone. The authors found that manual application of this method produced grids with low reproducibility, which they attributed to variation in incubation times, manual blotting, inconsistent drying times, different grid batches, and fluctuations in temperature and humidity with local weather.

The solution was to hand the entire deposition process to the Vitrobot Mark IV, which offers a stable atmospheric environment and programmable blotting. In the protocol, Quantifoil or UltrAufoil grids are first cleaned with chloroform and glow-discharged at 15 mA and 0.1 mBar for 30 seconds. The Vitrobot chamber is set to 20 °C and 100 percent humidity. Four sequential blotting steps follow: 4 µl of 1 mg/ml polyethyleneimine in 25 mM HEPES is applied with a 120-second wait before blotting; a double-distilled water wash is blotted immediately, followed by a 120-second drain; 4 µl of 0.2 mg/ml graphene oxide dispersion is applied with another 120-second wait for optimal adherence; and a final water wash is blotted before the grid is recovered. The grid then dries on filter paper for at least five minutes and is used the same day for optimal performance.

To evaluate coverage, the team recorded roughly 600 micrographs per grid across eight Quantifoil and eight UltrAufoil grids with 1.2/1.3 µm hole sizes. The presence of graphene was confirmed by the distinct hexagonal pattern of diffraction spots in the power spectrum, corresponding to a spacing of about 2.13 Å from the crystalline lattice of graphene: monolayers show six diffraction spots, bilayers twelve, and additional layers produce increasingly complex patterns. The grids achieved an average coverage of approximately 55 percent, with 40 percent being mono- and bilayer, and the coverage was distributed across the grid rather than focalized in one region, which increases the chance that a graphene layer will be present regardless of the ice gradient during protein application. Notably, neither grid material showed a marked advantage; Quantifoil and UltrAufoil grids yielded similar coverage levels and layer distributions, and the performance was comparable to some commercially available graphene oxide grids.

The structural results demonstrate that the method supports high-resolution work. For apoferritin, 10,000 movies were collected at the ESRF CM01 facility on a Titan Krios operating at 300 keV with a K3 detector in counting mode. After processing in cryoSPARC, including Topaz denoising and particle picking that yielded 2.1 million retained particles from 2.2 million picked, a final homogeneous refinement with O symmetry produced a 1.89 Å map. For β-galactosidase, 5,486 movies were processed, with approximately 3.2 million particles picked and 2.58 million retained; after 2D classification, ab-initio reconstruction, and 3D classification, 96,406 particles were refined with D2 symmetry to a 2.33 Å map. Particles contributing to the final reconstructions even included those picked from micrographs with five or more graphene oxide layers, confirming that multilayer coverage does not compromise image quality, and particles near graphene oxide flake edges can be excluded during classification due to edge artefacts.

Two further findings strengthen the practical case for the protocol. First, when 3 µl of 0.5 mg/ml protein solutions were applied, particles were well distributed across grid holes, and the apoferritin concentration was fourfold lower than that used for Quantifoil holey grids, suggesting that particle loss to the air-water interface is minimized; tomogram analysis showed the majority of apoferritin particles adsorbed to the graphene oxide layer with only a small fraction at the air-water interface. Second, the grids have a usable shelf life: grids stored at room temperature without humidity control for 21 days performed comparably to freshly prepared ones after a brief 5-second glow discharge, with stored and fresh grids reaching apoferritin resolutions of 1.77 Å and 1.76 Å respectively. A stored grid without glow discharge still yielded a 1.89 Å map, but the droplet contact angle was less acute, suggesting reduced hydrophilicity, so the authors do not recommend omitting the glow discharge step. They also caution against preparing large batches for long-term storage, since stability beyond the tested 21 days remains uncertain, and previous studies suggest graphene oxide is more stable in liquid suspension than in dried form.

The method’s broader value lies in accessibility and flexibility. Because it uses a commercially available plunge freezer that most cryo-EM users already operate, it lowers the barrier to experimenting with supported grids at low cost, and the polyethyleneimine priming means it can be applied to grid materials and hole sizes not commonly available commercially. The protocol was also validated beyond test proteins: grids prepared this way were used by François-Xavier Gillet and colleagues to reconstruct the plastid-encoded RNA polymerase from Sinapis alba at 3.0 Å, a 21-subunit complex of approximately 1.15 MDa whose low purification yield made standard holey grids infeasible. The authors note limitations as well: only one grid can be processed at a time, making the method unsuitable for large-scale production, and blotting force can vary significantly between individual Vitrobot units, so calibration should ideally be performed per machine. Adaptation to other controlled-environment plunge freezers, such as the Leica EM GP, should be possible but would require optimization of blotting and drying parameters. During optimization, the team found that increasing graphene oxide concentration improved overall coverage but reduced the proportion of mono- and bilayers and increased powder diffraction, and that raising glow discharge strength to 30 mA did not improve graphene oxide distribution, suggesting gentler conditions are more favorable.

For laboratories wrestling with fragile complexes, orientation-sensitive proteins, or samples that denature at the air-water interface, the protocol offers a reproducible entry point into supported-grid cryo-EM using equipment they already own. The authors anticipate that grids prepared with this method will be compatible with reported functionalization strategies, including single-stranded DNA, covalent modifications, and antibody immobilization, potentially enabling purification on the grid and enhanced particle adsorption. As supported grids continue to gain popularity in cryo-EM studies, easy-to-adopt methods like this one may help broaden their use across the structural biology community.

Subject of Research: A reproducible Vitrobot-based drop-cast method for preparing graphene oxide-supported cryo-EM grids

Article Title: Ready to GO: a reliable and reproducible method for preparing graphene oxide grids using the Vitrobot

Article References: Kwong, H. S., Grinzato, A., Freiherr von Scholley, G. L., Hograindleur, M.-A., & Kandiah, E. (2026). Ready to GO: a reliable and reproducible method for preparing graphene oxide grids using the Vitrobot. BMC Methods, 3(1), Article 26. https://doi.org/10.1186/s44330-026-00079-2

Image Credits: AI Generated

DOI: 10.1186/s44330-026-00079-2

Keywords: cryo-EM, graphene oxide, Vitrobot, grid preparation, polyethyleneimine, apoferritin, beta-galactosidase, air-water interface, structural biology, sample preparation, single-particle analysis, BMC Methods