Main

In single-particle cryogenic TEM (cryo-TEM) a three-dimensional (3D) density map is reconstructed from low-dose but high-resolution TEM images of thousands of copies of a macromolecule in vitreous (amorphous) ice1,2. The number of images for a given area of the sample is an important consideration as microscope time is costly, samples are difficult to prepare, sometimes only with sparse regions appropriate for acquisition, and more quality images of the particle typically translates to higher resolution3. Here, we demonstrate new aperture plate designs that produce square and rectangular beams in TEM that increase the number of images that can be recorded within the confines of a 1.2-µm foil hole of a typical cryo-TEM grid by 70%, on average.

In single-particle cryo-TEM, a grid containing a regular array of holes in a support foil4 (usually carbon or gold) is generally used to assist automated acquisition of cryo-TEM images5. Several acquisitions per foil hole, colloquially referred to as ‘multishot’ acquisition, improves the number of particles that can be imaged per unit area of grid6. Figure 1a plots the fraction of the hole that can be acquired using a round beam combined with a 4,096 × 4,096 pixel camera or a rectangular 5,760 × 4,096 pixel camera (the Gatan K3 camera) or a square or rectangular beam matched to the camera for different pixel sizes and thus different fields of view. The number of acquisitions possible in a multishot pattern is annotated for specific examples. Multishot configurations were taken from ref. 7 for round beams and from ref. 8 for square beams. For the rectangular case, we developed our own configurations, all of which are plotted in Supplementary Fig. 1. For a pixel size of less than 1.0 Å, commonly used in cryoelectron microscopy (cryo-EM), an average fraction of 0.32 and 0.41 of a foil hole is covered by the square and rectangle detectors with a round beam. Square and rectangular beams improve this coverage to 0.58 and 0.64, respectively. Taken over the plotted range—a pixel size varying from 1.5 Å to 0.4 Å—the average improvement was calculated to be approximately 70% for both square and rectangular detectors. An alternative approach is to allow the acquisition of some foil; in this case, we explore the most efficient packing of acquisitions in Supplementary Fig. 2 and show that we can achieve similar improvements (up to 72%) for efficiency over round beams.

Fig. 1: Single-particle cryo-TEM images are acquired in an ice-covered foil hole.
figure 1

a, Total fraction of the foil hole that can be both illuminated and acquired using square and round beams with a square (for example, ThermoFisher Falcon IV) or rectangular (for example, Gatan K3) camera as a function of pixel size. The numbers of acquisitions in the multishot configuration are annotated next to the colored plotted lines with multishot configurations at a pixel size of 0.65 Å inset. A square or rectangular beam increases the total fraction of the hole acquired by 70% on average. b, Aperture plates fabricated for a ThermoFisher TEM rod. c,d, Design schematics for Arctica (c) and Krios (d) microscopes. e,f, Low magnification SEM images of the fabricated plates for the Arctica (e) and Krios (f) microscopes. g,h, Higher magnification images of a square aperture from the Arctica plate (g) and a rectangular aperture from the Krios plate (h). i,j, Square (i) and rectangular (j) TEM beams in the Krios.

Full size image

To enable square and rectangular beams in a cryo-TEM, we fabricated our own aperture plates for our ThermoFisher Arctica G2 and ThermoFisher Krios G4 microscopes using reactive ion etching of 200-µm-thick silicon (Si) wafers followed by gold coating to prevent electron beam charging. A ThermoFisher aperture rod is shown in Fig. 1b. To anticipate rotation of the square beam as the condenser lens strength is adjusted to form different sized illumination, we offered several copies of each desired aperture rotated in 15° increments on a single plate—the smallest increment possible given available space. This ‘several apertures on a single plate’ approach is modeled on the aperture designs of Zeltmann et al.9 and means that at least one aperture will always be aligned with the camera within 15°.

The Arctica is a two-condenser instrument, which means it can form a single size of parallel beam only for a given condenser aperture, so the design for this aperture plate (Fig. 1c) incorporated four sets of different sized square apertures to match the beam size to each of the four most used acquisition magnifications as listed in Supplementary Table 1. This tailoring of beam size to magnification improved the flexibility of the instrument for multishot acquisition. The Krios is a three-condenser system, which means it can form a wide range of parallel beam sizes with a single aperture. Therefore, the aperture plate design (Fig. 1d) required only a single size of square and rectangular apertures.

In a single fabrication run, 32 aperture plates were made at modest cost, demonstrating a process that is easily scalable. Two example aperture plates were then inspected in a ThermoFisher Teneo Volumescope scanning electron microsope (SEM). Low-magnification images of both aperture plates are shown in Fig. 1e,f. All holes milled successfully for the Krios aperture plate but only the squares with side length >20 µm etched all the way through for the Arctica plate. Shown in Fig. 1g is a square aperture from the Arctica plate, with the bottom (opposite to beam, unplated) Si side up. Figure 1h is a rectangular aperture from the Krios plate, with the beam-facing, gold-plated side up. In both apertures, there is some rounding of the vertices of the aperture; this is not problematic because fringes at the edge the beam mean the vertices are not used. An image of the TEM beam with the rectangular and square apertures inserted is shown in Fig. 1i,j, with larger scale images in Supplementary Fig. 3. The effects of the aperture being not perfectly rectangular due to incomplete etching of some Si, as seen in the SEM images in Fig. 1f, are visible in the final beam. The next generation of aperture plates being developed by Norcada show sharper edges. The associated TEM beams are shown in Supplementary Fig. 3 along with stitched montage images produced by these latest apertures and the earlier generation of the aperture plates in Supplementary Fig. 4.

To benchmark cryo-EM acquisition with the new aperture design, we recorded datasets of Equine Apo-Ferritin with the ThermoFisher Falcon IV camera and square aperture, Gatan K3 and rectangular aperture combinations, with both cameras and the standard 50-µm round aperture as a comparison dataset. A ChimeraX10 rendering of a section apo-ferritin model reconstructed from the square aperture dataset is shown in Fig. 2a. Key metrics of the acquisitions are displayed in Fig. 2b and the resolutions of each map are given by the 0.143 criteria by the Fourier shell correlation (FSC) curves in Fig. 2c. The complete set of reconstructed apo-ferritin models are shown in Supplementary Fig. 5. The small but noticeable improvement in resolution between square and round apertures is probably explained by ice thickness, as revealed by consistently higher resolution contrast transfer function (CTF) fits plotted in a histogram in Supplementary Fig. 6. Separate CTF refinement of astigmatism, coma and three-fold aberration11 (Supplementary Fig. 7), including spherical aberration (Cs) (Supplementary Table 2) in Cryosparc did not indicate that the aperture plates were inducing extra aberrations in the objective lens system of the microscope. The rectangular aperture was also applied to the human amylin receptor 1 (AMYR1) bound to salmon calcitonin (sCT) receptor, previously solved by Cao et al.12, which is smaller (189 kDa) than apo-ferritin and not symmetrical. Using rectangular apertures, we were able to solve the structure to 2.80 Å with 138,000 particles, which compares favorably with the 3.0-Å structure in ref. 12 that used 227,000 particles on a 200-kV ThermoFisher Glacios instrument. A Rosenthal–Henderson B-factor plot3 in Supplementary Fig. 8 shows that our resolution is comparable on a per-particle basis with the other structures presented in ref. 12.

Fig. 2: Biological macromolecules reconstructed from single-particle cryo-EM data recorded with square and rectangular condensor apertures.
figure 2

a, Cross-section of the apo-ferritin map, reconstructed from data recorded using a square beam, visualized in ChimeraX10 and with PDB entry 6rjh overlaid13. b, Key parameters for each data collection (for the energy filter (eV) parameter, N/A indicates that no energy filtering was used). c, FSC plots to gauge resolution of each reconstructed model at the 0.143 criteria3. We also benchmarked the rectangular aperture-K3 combination on the Amylin 1 receptor bound to sCT (Amy1R:sCT) studied by Cao et al.12. d,e, Literature structure at 3-Å resolution alongside our reconstruction, which achieved a resolution of 2.8 Å (d) as shown by the FSC curve in e.

Full size image

In summary, we have fabricated a new type of condenser aperture plate for ThermoFisher cryo-TEMs that generate square or rectangular beams. This improves the usable area on the sample by 70% in single-particle cryo-EM with no detectable penalty to the resolution. Refinements in the manufacturing of the apertures and postmanufacturing milling with a focused ion beam instrument will yield further improvements on the sharpness of the beam edges.

Methods

Arctica G2 aperture plate

The aperture needed to be separated by an amount greater than the size of the electron beam at the C2 aperture plane in the TEM. This was measured by translating the 20 µm C2 aperture from the optic center until the beam disappeared. This point was found to vary with spot size but, with the larger beam, radius at the C2 plane was 300 µm for the ThermoFisher Arctica. Two designs were created, one where all the aperture holes were located inside a 1.2-mm disc (Fig. 1c) and another where all were located inside a 2-mm disc. Four of each 15° aperture rotations were created for a total of 4 × 4 = 16 apertures. Fringes at the edge of the illumination and beam-position drift over the course of an acquisition were measured independently to both be approximately 75 nm, so 300 nm was added to the sensor size (that is, an extra 150-nm margin to each side of the beam) to calculate the desired size of the square beam. The multiplicative factor that related aperture to beam size at the parallel condition was measured to be 29.4 using the 20-µm aperture already installed on the Arctica. This factor was used to calculate the aperture hole size that would result in a desired beam size. A schematic of the final design is shown in Fig. 1c; the optimal layout of the 16 aperture holes to ensure maximum separation between different holes while still fitting inside a 1.2-mm circle is taken from ref. 7, which ensured that apertures were spaced just over 300-µm apart.

Krios G4 aperture plate

The Krios at the Ian Holmes Imaging Centre is equipped with both a K3 (4,096 × 5,960 pixel) and Falcon 4 (4,096 × 4,096 pixel) cameras so both square and rectangular apertures needed to be fabricated into the same plate. The initial design (Fig. 1c) featured square and rectangular apertures that measured 50 µm in the smallest dimension, with four 15° rotations covering a 60° range for the square apertures and twelve 15° rotations for the rectangular apertures. This size of aperture is insufficient for forming a small enough parallel beam to match the camera at magnifications higher than ×130,000. Attempting to reduce the beam size further meant that microscope beam mode changed from ‘parallel’ to ‘condensing,’ resulting in a nonuniform beam brightness that was not usable for acquiring high-quality data. We have revised the design for future aperture plates to feature apertures that measure 30 µm in the smallest dimension so that sufficiently small beam sizes to cover the full range of magnifications are achievable.

Fabrication of aperture plates by reactive ion etching

Aperture designs, specifically the positioning of the vertices of each aperture on the 3.04-mm plate were generated using a Python script included in the Supplementary Information. The ezdxf Python library14 was used to write these vertices as Polygon objects in the AutoCAD drawing exchange format (dxf) and these are also included in the supplementary material. The dxf designs were transferred to a chrome/glass photomask that is used in ultraviolet (UV) photolithography as a 1:1 transfer mechanism. The substrate of choice for the apertures was 200-µm-thick Si due to the relative strength of Si and its ability to be etched into shapes with high aspect ratios that were necessary for the apertures. To transfer the pattern from the photomask to the Si wafer, AZ40XT photoresist was spin-coated and soft-baked onto the Si substrate at about 25-µm thickness. The pattern transfer was then achieved by 365-nm UV exposure of the patterned photomask onto the photoresist. Subsequent development of the pattern removed the UV-exposed areas of the photoresist, creating areas on the Si that could then be removed by reactive ion etching (RIE) using an Oxford PlasmaLab RIE machine. Areas not targeted for etching were protected by the remaining photoresist. After the features were formed into the Si substrate, the remaining sacrificial protective resist was removed using RIE with O2 plasma. To minimize charging of the apertures, which would produce probe aberrations, 1 µm of conducting Au film was then sputter deposited onto the Si wafer covering the aperture plates and making them conductive on the surface facing the beam. Finally, the wafer was cut into individual pieces correctly sized to fit into the TEM aperture holding mechanism. Before loading into the condenser aperture holder of the microscope, a gaseous N2 jet was used to remove dust from the apertures and the plates were plasma cleaned for 5 min in an O/Ar Fischione 1020 plasma cleaner. The ThermoFisher Krios G4 microscope column was vented and the existing 70-µm aperture plate was replaced with the design shown in Fig. 1d; normal vacuum was achieved after an overnight pumping cycle.

Once the new apertures were loaded into the ThermoFisher Krios G4, their positions were found by adjusting the x and y positioning of the aperture rod to bring different apertures into view and then recording the position indicated in the ThermoFisher utility tadBhvApertures.exe. The recorded positions could be used to bring certain apertures back in separate microscope sessions.

We found no noticeable impact on the alignment time or complexity on the instrument, with common alignments such as coma aberration correction using a Zemlin tableau15 converging within a few iterations; see Supplementary Fig. 7 for a typical output of a coma correction job with the square and rectangular apertures inserted.

Montage image data collection

Montages of vitreous-ice-covered quantifoil holes were recorded using the ThermoFisher EPU software (v.3.2), which is intended for automated single-particle cryo-EM data collection. This software allows for nearby areas to be imaged using the beam shift beam-deflection coils in the upper part of the objective lens system. A compensating deflection using the image-shift beam-deflection coils moves the deflected beam back to the same position on the camera. For the Arctica montage (Supplementary Fig. 4a) a square 2 × 2 array of acquisitions was used, with the square beam condensed to be roughly two-thirds of the size of the camera, whereas a 2 × 3 array of acquisitions was used for the Krios and Gatan K3 camera combination. The images comprising the montage were aligned manually using the Inkscape v.1.1.2 (https://inkscape.org/) graphics editor and the resulting image position offsets were used in a python script to additively combine the images.

Data collection using circular, square and rectangular beams in cryo-EM

Equine apo-ferritin (Sigma-Aldrich, cat. no. A3641) solution (4 µl) was pipetted onto gold 1.2/1.3 Quantifoil UltrAuFoil grids, which had been glow discharged for 180 s in a Quorum GloQube with a 15-mA current, and plunge-frozen in liquid ethane using a ThermoFisher Vitrobot Mk IV. The vitrobot chamber was set to room temperature (22°C) with a humidity of 95% and a blot force of −1 and blot time of 4 s was used. Key parameters and metrics of the datasets are shown in Fig. 2b. For each dataset, the processing was as follows: videos were aligned using the patch motion correction algorithm of the CryoSPARC cryo-EM processing software16 with CTF-fitting also performed using the patch motion correction algorithm. Particle picking was performed using a template generated by two-dimensional (2D) classification of 100 manual picks. The final particles were selected through two rounds of 2D classification and went into 3D homogenous refinement. Particles were split into different optics groups using the EPU_group_AFIS.py python script17, which parses metadata.xml files outputted by EPU software (ThermoFisher) to extract beam-tilt values for each acquisition, and clusters together similar beam-tilt values. This grouping was then imported to CryoSPARC using an in-house script (get_exp_group_id.py), which is included in the supplementary materials). Homogenous refinement in CryoSPARC was performed with per-group high aberration refinements. Cs and per-particle defocus were also refined to determine the effect of the square aperture on the Cs.

The Amy1R:sCT grids were the same as those used in ref. 12. Sample (3 μl) was pipetted onto glow discharge 1.2/1.3 Quantifoil UltrAuFoil grids and then subsequently blotted and vitrified in liquid ethane using a ThermoFisher Vitrobot Mk IV with the chamber set to 4°C and 100% humidity. Images were acquired on a ThermoFisher Krios G4 cryo-TEM with rectangular aperture plate at a magnification of ×105,000 (pixel size 0.833 Å) using an energy filtered K3 with a slit width of 10 eV. A total of 3,333 videos were collected, motion-corrected and CTF fitted using Cryosparc Patch motion correction and Patch CTF jobs. Picking using templates from a preliminary 2D classification of blob picks yielded 1,441,500 initial particles from which 138,105 final particles were selected using several rounds of 2D classification and 3D heterogenous refinement jobs, the latter against ab initio models generated in Cryosparc. A final resolution of 2.80 Å was achieved with a Cryosparc Local Refinement job which uses a 3D binary mask to target 3D refinement. The mask was generated by eliminating the detergent micelle.

Included in the Supplementary Information are the aperture designs in the AutoCAD DXF format along with the script to generate these designs and the get_exp_id.py Python script used to extract optics groups from Relion style star files and import them into Cryosparc.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.