The basic concept of Preassis is to pull a portion of sample suspension through an EM grid and simultaneously remove excess liquid from the backside of the grid with the assistance of suction/pressure. In its simplest setup (Fig. 1a), an EM grid is placed on a filter paper which is rested on the mouth of a Buchner flask and pumped with a certain pumping speed. A droplet of the sample suspension is then deposited onto the grid. Due to the suction underneath the filter paper, the excess liquid is “pulled” through the grid. Then the grid is manually picked up using a tweezer and plunged into liquid ethane. Details of this setup and specimen preparation procedures are available in the Supplementary Protocol. The overall vitrified ice thickness on the EM grid can be tuned by changing the pressure, hole size of the EM grid, and the time over which the pressure is applied. While the pressure can be changed continuously, the change of the hole size is done by choosing the type of EM grids. In the setup described above, the pressure is adjusted by changing the pumping speed, which is linearly proportional to the pumping speed in the range of 20% to 80% (Supplementary Fig. 1). It is worth noting that Preassis can be applied on pre-clipped EM grids used for auto-loading as well, which makes this method very promising for future automation.
The most important advantage of Preassis is its ability to handle protein crystals grown in highly viscous buffers. We applied it to crystals of Sulfolobus acidocaldarius R2-like ligand-binding oxidase (R2lox) (Fig. 1b and c) grown with 44% PEG 400. It was difficult to remove enough amount of liquid by Vitrobot even with extreme blotting conditions (2 layers of filter paper on each side, strong blotting force (16), and long blotting time (10 s)), as shown in Fig. 1c. Only a few grid squares were electron beam transparent and the ice layers are very thick (as represented in Supplementary Fig. 2b and c), making it extremely difficult to obtain sufficient MicroED data with good diffraction quality. The diffraction quality of MicroED data of R2lox couldn’t be improved and the R2lox project got stuck for more than one year until the Preassis method was proposed. Using Preassis with 30.7 mbar pressure and Quantifoil grid R3.5/1, the viscous liquid was efficiently removed obtaining a lot of grid squares suitable for searching crystals for MicroED data collection, as shown in Fig. 1b. Owing to the reduced vitrified ice thickness, the resolution of electron diffraction (ED) data was significantly improved from 9.0 Å to 3.0 Å (as shown by ED patterns in Fig. 1b and c). Consequently, sufficient and high quality MicroED data were collected from one single TEM grid prepared by Preassis, and a typical MicroED dataset of R2lox is shown in Supplementary Video. Preassis is crucial for the successful structure determination of R2lox, the first novel protein structure solved by MicroED3. Furthermore, Preassis can preserve 10 times more crystals on the EM grid than that of Vitrobot, as shown in Fig. 1d and e. This is mainly because that Preassis takes away the excess liquid from the back-side (copper-side) of an EM grid and the holey carbon layer of the grid can act as a sieve to keep more crystals, while the Vitrobot uses two-side blotting and the excess liquid and precious crystals can be taken away from both sides.
It is worth noting that unlike single particle cryo-EM experiments where 10 000 to 100 000 of particles are required for a complete structure determination, MicroED experiments require only a limited number (up to 50) of crystals. Therefore, a successful MicroED specimen preparation is to ensure there are sufficient microcrystals covered by thin ice on several grid squares, so that high-quality MicroED data can be collected. The influences of pressure and hole sizes were investigated by combining TEM images and SAED patterns. Each specimen preparation condition was repeated 2 to 5 times and the reproducibility is relatively good, as illustrated in Supplementary Fig. 3. It is worth mentioning that the ice layer thickness can vary throughout an EM grid prepared by the current Preassis setting, as shown in Supplementary Fig. 4. This may be due to non-uniform contact of the EM grid to the filter paper. Such a gradient of ice thickness may not be a disadvantage, and instead increase the chance of finding suitable crystals for MicroED data collection.
Figure 2 illustrates how the pressure and hole size affect the overall ice thickness on the EM grid and the resulting quality of electron diffraction patterns using orthorhombic lysozyme crystals as an example23. By applying a low pressure of 17.2 mbar in a combination of EM grids with a small hole size of 1.2 µm (R 1.2/1.3 Quantifoil grids), usable specimens could be obtained (Fig. 2b). However, the vitrified ice was relatively thick as seen by the reduced transparent area in each grid square and strong ice rings in corresponding ED patterns (Fig. 2b). A better EM grid with thinner ice layers could be obtained by slightly increasing the pressure to 27.7 mbar, where the sharp edges of grid squares are visible in the TEM images (Fig. 2a). With thinner ice, it was easier to find suitable crystals for MicroED data collection, and ice rings were eliminated in the ED patterns. Consequently, the quality of MicroED data was improved in terms of resolution and signal to noise ratio.
Furthermore, the hole size of the EM grid has a strong influence on the ice thickness. Instead of increasing the pressure, thinner ice layers can also be obtained by using EM grids with larger holes because it is easier for liquids to pass through. When a R 3.5/1 grid with a hole diameter of 3.5 µm was used (Fig. 2d), the ice became thinner compared to that of R 1.2/1.3 grids prepared under the same pressure (17.2 mbar, Fig. 2b). High resolution reflections could be obtained in the corresponding ED pattern and no ice rings are observed. On the other hand, a combination of high pressure (27.7 mbar) with a large hole size (R 3.5/1 grid) led to slight dehydration of the lysozyme crystals, as indicated by reduced resolution in the ED pattern (Fig. 2c). Our results show that optimal ice layer thicknesses could be obtained under several conditions; both combinations with high pressure/small hole size (Fig. 2a) and low pressure/large hole size (Fig. 2d) produced optimal EM grids for MicroED data collection. This is significant because it allows us to select EM grids according to the size and shape of the crystals, and control and fine-tune the ice thickness by adjusting the pressure. Ideally, the hole size of the grid should be slightly smaller than or comparable to the largest crystal dimensions to maximize the hole areas and at the same time prevent the loss of crystals through the holes. This is particularly important when the initial density of crystals is low. We anticipate that the size and shape of microcrystals may also affect the ice thickness. Therefore Preassis was applied also to tetragonal lysozyme crystals with an isotropic shape and the results are shown in Supplementary Figs. 4 and 5a. Compared to the optimal pressure (27.2 mbar) for the preparation of orthorhombic lysozyme crystals with rod-like morphology on the R 1.2/1.3 grids, a higher pressure (37.2 mbar) was needed to achieve similar ice thickness and MicroED data quality for the tetragonal lysozyme crystals. It is observed that despite the absence of vitrified ice in the neighbouring empty holes, most crystals are still embedded in ice, showing microcrystals attract more liquids than the EM grids do presumably caused by the stronger surface tension. Consequently, the applied pressure needs to be adjusted according to the size and shape of the crystals.
For non-viscous samples (e.g. no high molecular weight polymers in the buffer), usable specimens can be prepared by Preassis even without applying pressure if other parameters (e.g hole size and time) are carefully chosen (Supplementary Fig. 5). If the crystal suspension is viscous due to e.g. high concentration or high molecular weight of PEGs in the buffer, high pressures (e.g. \(>\) 180 mbar) are required. To systematically study the specimen preparation under different viscosities, several experiments were conducted with various concentrations of PEG 6000 mixed with microcrystals. Submicron-sized crystals of an inorganic sample zeolite ZSM-5 were used instead since it is difficult to obtain the same type of protein crystals from mother liquids of different viscosities. The results are shown in Supplementary Fig. 6. We found that, for crystal suspension with PEG 6000 lower than 25%, usable grids could be obtained without applying pressure when grids with large holes were used (e.g. Quantifoil R 3.5/1, Supplementary Fig. 6f and g). However, the amorphous ice on grids prepared under those conditions was relatively thick in most grid squares, which reduces the contrast of the crystals and makes it difficult to find suitable crystals for MicroED data collection. By applying 181 mbar pressure, the specimens could be improved (Supplementary Fig. 6b and c). When the crystal suspension was very viscous, such as containing 35% PEG 6000, high pressure is necessary in order to obtain usable grids (Supplementary Fig. 6d and h). In reality, the pressure could be further increased to reduce the vitrified ice thickness for the cases as shown in Supplementary Fig. 6c and d. In conclusion, grids with large hole sizes and high pressure are recommended for viscous crystal suspensions. In cases of crystals grown in extremely viscous buffers such as LCPs, it is difficult to obtain usable EM grids by Preassis alone, a combination with other methods may be necessary, e.g. adding detergents, oils, or lipase24 to decrease the viscosity of the LCP.
Even if a good or usable specimen is obtained, finding ideal grid squares and good crystals for data collection is also important. This is because the thickness distribution of the amorphous ice layer is often not homogeneous across the grid, sometimes is not uniform within the same grid square. For crystals grown in non-viscous mother liquors, ideal grid squares and crystals have the following features: 1) grid squares with clear and sharp edges when viewed in low magnification (Fig. 3a), indicating the vitrified ice is very thin, 2) crystals found in these squares with blurred edges on the carbon film, indicating the crystals are still hydrated, and 3) crystals hanging over the empty holes with relatively clear edges, indicating the ice is very thin. It is desirable to collect MicroED data from these types of crystals, and at the regions where they are hanging over the holes. In these cases, carbon background is avoided, the ice thickness is minimized, and the crystals are usually still hydrated. A representative example including images and diffraction patterns is shown in Fig. 3a. When the buffer is viscous (e.g. 44% PEG 400 and 30% PEG 4000), the grid squares no longer show sharp edges and transparent areas are reduced compared to those prepared under non-viscous conditions, as shown in the low magnification images in Fig. 3b and c. Under such circumstances, the ideal crystals for MicroED data collection are still those hanging over the holes (crystal images and ED patterns in Fig. 3b and c) despite the surrounding vitrified ice. Furthermore, for both viscous and nonviscous cases, most holes in the optimized grid squares are empty without vitrified ice layer except for the holes with crystals nearby or on top of the holes.
In conclusion, Preassis is a simple and promising method for preparing MicroED specimens. Our results demonstrate that the ice thickness can be adjusted by tuning pressure and selecting grids with different hole sizes. More importantly, Preassis allows the preparation of MicroED specimens for crystals grown in high viscous media and keeps a relatively high density of crystals on the EM grid. Here we provide a guideline on how to select parameters for specimen preparation of a new protein crystal sample using Preassis. Common features of good grid squares and crystals for high-quality MicroED data collection are discussed, which will be important for future automation of data collection and high-throughput structure determination by MicroED. Preassis is simple and can be easily implemented in all cryo-EM labs. New implementations of Preassis, such as developing automation and adding an environmental chamber, may improve the throughput and reproducibility of MicroED specimen preparation. We believe that this method will significantly widen the application of MicroED.