Hostname: page-component-cd9895bd7-fscjk Total loading time: 0 Render date: 2024-12-25T23:53:24.725Z Has data issue: false hasContentIssue false
  • English
  • Français

Automated Tools to Advance High-Resolution Imaging in Liquid

Published online by Cambridge University Press:  20 January 2022

G. M. Jonaid ,
Michael A. Casasanta ,
William J. Dearnaley ,
Samantha Berry ,
Liam Kaylor ,
Madeline J. Dressel-Dukes ,
Michael S. Spilman ,
Jennifer L. Gray  and
Deborah F. Kelly Open the ORCID record for Deborah F. Kelly [Opens in a new window]
G. M. Jonaid
Affiliation:
Bioinformatics and Genomics Graduate Program, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA 16802, USA Department of Biomedical Engineering, Pennsylvania State University, University Park, PA 16802, USA Materials Research Institute, Pennsylvania State University, University Park, PA 16802, USA Center for Structural Oncology, Pennsylvania State University, University Park, PA 16802, USA
Michael A. Casasanta
Affiliation:
Department of Biomedical Engineering, Pennsylvania State University, University Park, PA 16802, USA Materials Research Institute, Pennsylvania State University, University Park, PA 16802, USA Center for Structural Oncology, Pennsylvania State University, University Park, PA 16802, USA
William J. Dearnaley
Affiliation:
Department of Biomedical Engineering, Pennsylvania State University, University Park, PA 16802, USA Materials Research Institute, Pennsylvania State University, University Park, PA 16802, USA Center for Structural Oncology, Pennsylvania State University, University Park, PA 16802, USA
Samantha Berry
Affiliation:
Department of Biomedical Engineering, Pennsylvania State University, University Park, PA 16802, USA Materials Research Institute, Pennsylvania State University, University Park, PA 16802, USA Center for Structural Oncology, Pennsylvania State University, University Park, PA 16802, USA
Liam Kaylor
Affiliation:
Department of Biomedical Engineering, Pennsylvania State University, University Park, PA 16802, USA Materials Research Institute, Pennsylvania State University, University Park, PA 16802, USA Center for Structural Oncology, Pennsylvania State University, University Park, PA 16802, USA Molecular, Cellular, and Integrative Biosciences Graduate Program, Huck Institutes of the Life Sciences, Pennsylvania State University, University Park, PA 16802, USA
Madeline J. Dressel-Dukes
Affiliation:
Protochips Inc., Morrisville, NC 27560, USA
Michael S. Spilman
Affiliation:
Direct Electron, LP, San Diego, CA 92128, USA
Jennifer L. Gray
Affiliation:
Materials Research Institute, Pennsylvania State University, University Park, PA 16802, USA
Deborah F. Kelly*
Affiliation:
Department of Biomedical Engineering, Pennsylvania State University, University Park, PA 16802, USA Materials Research Institute, Pennsylvania State University, University Park, PA 16802, USA Center for Structural Oncology, Pennsylvania State University, University Park, PA 16802, USA
*
*Corresponding author: Deborah F. Kelly, E-mail: Debkelly@psu.edu

Abstract

Liquid-electron microscopy (EM), the room-temperature correlate to cryo-EM, is a rapidly growing field providing high-resolution insights of macromolecules in solution. Here, we describe how liquid-EM experiments can incorporate automated tools to propel the field to new heights. We demonstrate fresh workflows for specimen preparation, data collection, and computing processes to assess biological structures in liquid. Adeno-associated virus (AAV) and the SARS-CoV-2 nucleocapsid (N) were used as model systems to highlight the technical advances. These complexes were selected based on their major differences in size and natural symmetry. AAV is a highly symmetric, icosahedral assembly with a particle diameter of ~25 nm. At the other end of the spectrum, N protein is an asymmetric monomer or dimer with dimensions of approximately 5–7 nm, depending upon its oligomerization state. Equally important, both AAV and N protein are popular subjects in biomedical research due to their high value in vaccine development and therapeutic efforts against COVID-19. Overall, we demonstrate how automated practices in liquid-EM can be used to decode molecules of interest for human health and disease.

Keywords

adeno-associated virusCOVID-19liquid-electron microscopynucleocapsid proteinSARS-CoV-2
Type
Software and Instrumentation
Information
Microscopy and Microanalysis , Volume 28 , Issue 2 , April 2022 , pp. 361 - 370
Check for updates
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyright
Copyright © The Author(s), 2022. Published by Cambridge University Press on behalf of the Microscopy Society of America

Introduction

Understanding life's processes in exquisite detail has been a work-in-progress over the last century. Due to recent advances in molecular microscopy, in particular cryo-electron microscopy (EM), scientists can now visualize minute features of whole cells, cellular organelles, and biological assemblies (Li et al., Reference Li, Baker, Jiang, Estes and Prasad2009; Hu et al., Reference Hu, Morado, Margolin, Rohde, Arizmendi, Picking, Picking and Liu2015; Oikonomou & Jensen, Reference Oikonomou and Jensen2016; Deng et al., Reference Deng, Marshall, Rowland, McCoy, Worrall, Santos, Strynadka and Finlay2017; Murata & Wolf, Reference Murata and Wolf2018; Varano et al., Reference Varano, Alden, Dearnaley, Casasanta, Damiano, McConnell, Dukes and Kelly2019). Improvements in electron optics, direct detectors, computing algorithms, and molecular modeling software contributed to this revolutionary capability (Bammes et al., Reference Bammes, Rochat, Jakana, Chen and Chiu2012; Migunov et al., Reference Migunov, Ryll, Zhuge, Simson, Strüder, Batenburg, Houben and Dunin-Borkowski2015). To uncover nanoscale information among flexible molecules, cryo-EM samples are pristinely preserved in thin layers of vitreous ice. This preservation step yields individual snapshots of moving parts, instantly frozen in time. Given enough data, one can begin to extrapolate nanoscale mechanics in a time-resolved manner (Penczek et al., Reference Penczek, Grassucci and Frank1994; Frank et al., Reference Frank, Zhu, Penczek, Li, Srivastava, Verschoor, Radermacher, Grassucci, Lata and Agrawal1995; Frank, Reference Frank2006; Shi et al., Reference Shi, Di Giammartino, Taylor, Sarkeshik, Rice, Yates, Frank and Manley2009).

Viewing biological samples in a fluid environment, however, can significantly expand our knowledge of their natural state of being. It is highly desirable to pair high-resolution structural information with real-time dynamics, as many events occur in the order of milliseconds or faster. Such rapid movements may include changes in pliable protein loops that serve as immune epitopes on human viruses (Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). Improved knowledge of flexible protein structures can inspire the design of new antibody therapies or anti-viral agents to combat emerging variants of influenza, ebola, or SARS-CoV-2 (Kiss et al., Reference Kiss, Chen, Brindley, Campbell, Afonso, Ke, Holl, Guerrero-Ferreira, Byrd-Leotis, Steel, Steinhauer, Plemper, Kelly, Spearman and Wright2014; Pallesen et al., Reference Pallesen, Murin, De Val, Cottrell, Hastie, Turner, Fusco, Flyak, Zeitlin and Crowe2016; Casasanta et al., Reference Casasanta, Jonaid, Kaylor, Luqiu, Solares, Schroen, Dearnaley, Wilson, Dukes and Kelly2021; Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021; Watanabe et al., Reference Watanabe, Castillon, Avalos, Ellisman and Saphire2021). Importantly, viewing materials in a fluid, stochastic setting provides unique insight of their performance in the body.

Liquid-EM (liquid-EM) is a premiere technique to visualize the dynamic behavior of macromolecules. High-resolution studies of hard polymers have produced exciting results, and similar details are starting to transpire for soft polymers and biological entities (Zheng et al., Reference Zheng, Smith, Jun, Kisielowski, Dahmen and Alivisatos2009; Yuk et al., Reference Yuk, Park, Ercius, Kim, Hellebusch, Crommie, Lee, Zettl and Alivisatos2012; Chen et al., Reference Chen, Smith, Park, Kim, Ho, Rasool, Zettl and Alivisatos2013; Liao & Zheng, Reference Liao and Zheng2013; Alloyeau et al., Reference Alloyeau, Dachraoui, Javed, Belkahla, Wang, Lecoq, Ammar, Ersen, Wisnet, Gazeau and Ricolleau2015; Park et al., Reference Park, Elmlund, Ercius, Yuk, Limmer, Chen, Kim, Han, Weitz, Zettl and Paul Alivisatos2015a; Ahmad et al., Reference Ahmad, Wang, Nelayah, Ricolleau and Alloyeau2017; Piffoux et al., Reference Piffoux, Ahmad, Nelayah, Wilhelm, Silva, Gazeau and Alloyeau2018; Reboul et al., Reference Reboul, Heo, Machello, Kiesewetter, Kim, Kim, Elmlund, Ercius, Park and Elmlund2021). The lag in liquid-EM applications in life sciences may be accelerated by automated routines regularly used in the cryo-EM field. Opportunities for improvement include specimen preparation, high-throughput data collection, and computational processing of movie data (Fig. 1).

Fig. 1. Steps in the liquid-EM workflow that can be enhanced by automated practices. Liquid-EM sample preparation typically involves specialized side-entry specimen holders (left panel). Using this convention, liquid layers are sandwiched between two materials or microchips. Adapting cryo-EM autoloaders for this step permits researchers to view multiple samples in parallel. EM data collection (center panel) can be enhanced by automated data collection programs to output real-time movies of the imaged samples. Computing procedures (right panel) are accelerated using software procedures performed as parallel jobs on multiple processors or graphics cards (GPUs).

Here, we describe how automated tools can advance high-resolution imaging in liquid. We examine two recent case studies, ranging from large virus assemblies, adeno-associated virus subtype 3 (AAV) to small individual proteins, the SARS-CoV-2 nucleocapsid (N) (Casasanta et al., Reference Casasanta, Jonaid, Kaylor, Luqiu, Solares, Schroen, Dearnaley, Wilson, Dukes and Kelly2021; Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). These macromolecules were specifically chosen, as they have very different dimensions and symmetries. AAV has an icosahedral framework (60-fold symmetry) and is ~25 nm in diameter. The SARS-CoV-2 N protein is asymmetric and a mere 5–7 nm in diameter, depending upon its oligomeric state. For both cases, we describe the manner in which automated routines can enhance specimen handling, imaging workflow, and molecular interpretations of biological structures in solution.

Materials and Methods

Biochemical Preparation of AAV and N Protein

The AAV purification protocol was recently described (Rutledge et al., Reference Rutledge, Halbert and Russell1998; Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). Briefly, HEK293 cells were transfected using at a 2:1:1 ratio of AAV helper (260 μg), AAV cis (130 μg), and AAV trans (130 μg) plasmids and incubated at 37°C with 5% atmospheric CO2 for 24 h. Viruses were purified using affinity chromatography methods that employed AVB sepharose HP (GE Healthcare) and AAV preparations were evaluated via sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis. Fractions were diluted in buffer solution (50 mM HEPES, pH, 7.5; 150 mM NaCl, 10 mM MgCl2, and 10 mM CaCl2) before imaging experiments. The SARS-CoV-2 N protein was expressed and purified from bacteria and is available from RayBiotech, Inc. (Casasanta et al., Reference Casasanta, Jonaid, Kaylor, Luqiu, Solares, Schroen, Dearnaley, Wilson, Dukes and Kelly2021). Samples were diluted to 1 mg/mL in Tris buffer solution (20 mM Tris, pH, 7.5; 150 mM NaCl, 10 mM MgCl2, 10 mM CaCl2).

Sample Preparation and EM Data Collection

For AAV samples, aliquots (~0.2 μL each of 0.1 mg/mL) were added to glow-discharged microchips placed in the tip of a Poseidon Select specimen holder (Protochips, Inc.). A second glow-discharged top chip was placed on top of the wet chip that was sealed with a metal faceplate and three brass screws. The holder was inserted into a ThermoFisher F200C TEM operating at 200 kV under low-dose conditions (<5 e−/Å2). Bright-field TEM was used to acquire movies at 92,000× on-microscope magnification with a DE-12 (Direct Electron, LP) direct electron detector having a pixel size of 6 μm. Movies were recorded at 40 fps and binned 2-fold for a final sampling of 1.01 Å/pixel at the specimen level. For N protein samples, ~0.2 μL each of 1.0 mg/mL were added to glow-discharged SiN or SiO microchips (Simpore, Inc.) placed on a gel-pack substrate. Samples were incubated for 2 min. Glow-discharged gold grids with continuous carbon film (~5 nm thick) were placed on top of the wet microchips. The microchip sandwiches were loaded in a room temperature side-entry specimen holder or clipped using grid clips for an autoloader device. Movies of N protein samples were recorded at 40 fps using 0.5 s exposures at a nominal magnification of 45,000× and 200 kV under low-dose conditions. The same F200C TEM and DE-12 detector operating in linear mode were used for both imaging studies.

Image Processing Procedures

Movies of AAV in solution were processed in RELION-3.08 and cryoSPARC 2.14.2 (Scheres, Reference Scheres2012; Punjani et al., Reference Punjani, Rubinstein, Fleet and Brubaker2017). The drift correction of beam-induced movement was performed using MotionCor2 v1.2.3 (Zheng et al., Reference Zheng, Palovcak, Armache, Verba, Cheng and Agard2017). Auto-picking was performed using RELION, and particles were extracted using a box size of 330 pixels and default parameters as previously described (Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). Extracted particles were used to generate an initial 3D map while imposing no symmetry (C1). Using RELION, 25 refinement iterations were executed implementing a regularization parameter of T = 4, a pixel size of 1.01 Å, and a mask value of 300 Å. For comparison, an ab initio model was reconstructed in cryoSPARC with C1 symmetry using default parameters. This model was used to generate a 3D map employing cryoSPARC's homogenous refinement with I1 symmetry. For validation measures, particle data were divided into two halves and the resolution of each half converged to a common value. The gold-standard fourier shell correlation (FSC) criteria estimated a 3.22 Å resolution, which agreed with the Cref(0.5) evaluation and was consistent with output from the RMEASURE program (Sousa & Grigorieff, Reference Sousa and Grigorieff2007). Using the icosahedral-symmetry operator, the total particle equivalency was 16,800. The map was masked at ~250 Å and visualized using the Chimera software package (Pettersen et al., Reference Pettersen, Goddard, Huang, Couch, Greenblatt, Meng and Ferrin2004; Goddard et al., Reference Goddard, Huang and Ferrin2007; Sousa & Grigorieff, Reference Sousa and Grigorieff2007). Slices through the EM map were incremented at ~5 nm. The same particle extraction and drift correction routines were implemented for N protein to generate 2D class averages. In total, ~10,000 asymmetric particles of N protein were used for averaging routines. Data collection is ongoing to achieve high-resolution 3D structures of the N protein samples in solution.

Results and Discussion

Liquid-EM Sample Enclosures

There are a growing number of protocols to prepare samples in liquid enclosures (Pettersen et al., Reference Pettersen, Goddard, Huang, Couch, Greenblatt, Meng and Ferrin2004; Goddard et al., Reference Goddard, Huang and Ferrin2007; Sousa & Grigorieff, Reference Sousa and Grigorieff2007; Yuk et al., Reference Yuk, Park, Ercius, Kim, Hellebusch, Crommie, Lee, Zettl and Alivisatos2012; Dukes et al., Reference Dukes, Thomas, Damiano, Klein, Balasubramaniam, Kayandan, Riffle, Davis, McDonald and Kelly2014; Dearnaley et al., Reference Dearnaley, Schleupner, Varano, Alden, Gonzalez, Casasanta, Scharf, Dukes and Kelly2019; Keskin et al., Reference Keskin, Kunnas and de Jonge2019; Dunn et al., Reference Dunn, Adiga, Pham, Bryant, Horton-Bailey, Belling, LaFrance, Jackson, Barzegar, Yuk, Aloni, Crommie and Zettl2020; Serra-Maia et al., Reference Serra-Maia, Kumar, Meng, Foucher, Kang, Karki, Jariwala and Stach2021). To determine the proper method, researchers must consider the size and properties of the material, the sample's inherent contrast, and the project goals. For instance, if quantitative assessments of long-range movements are desired, thicker liquid layers are more ideal than thinner layers that may limit diffusion. If smaller macromolecules are being evaluated for single-particle structural studies, similar to cryo-EM practices, thinner liquid layers offer a better signal-to-noise ratio (SNR) in TEM images. Ultra-thin liquid layers can be achieved through the use of graphene liquid cells (GLCs) along with other carbon sandwiching techniques (de Jonge & Ross, Reference de Jonge and Ross2011; Yuk et al., Reference Yuk, Park, Ercius, Kim, Hellebusch, Crommie, Lee, Zettl and Alivisatos2012; Chen et al., Reference Chen, Smith, Park, Kim, Ho, Rasool, Zettl and Alivisatos2013; Park et al., Reference Park, Elmlund, Ercius, Yuk, Limmer, Chen, Kim, Han, Weitz, Zettl and Paul Alivisatos2015a, Reference Park, Park, Ercius, Pegoraro, Xu, Kim, Han and Weitz2015b; Keskin & de Jonge, Reference Keskin and de Jonge2018; Keskin et al., Reference Keskin, Kunnas and de Jonge2019; Reboul et al., Reference Reboul, Heo, Machello, Kiesewetter, Kim, Kim, Elmlund, Ercius, Park and Elmlund2021). One complication with the GLC method is the background subtraction step needed to remove the graphene lattice that convolutes each image.

Alternatively, two different methods can be used to produce a range of liquid thicknesses. One method includes silicon nitride (SiN) microchips and commercially available specimen holders (Dukes et al., Reference Dukes, Thomas, Damiano, Klein, Balasubramaniam, Kayandan, Riffle, Davis, McDonald and Kelly2014; Gilmore et al., Reference Gilmore, Winton, Demmert, Tanner, Bowman, Karageorge, Patel, Sheng and Kelly2015; Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). Another technique that is gaining popularity involves sandwiched materials that employ one microchip and one amorphous carbon layer. This technique is optimal to enclose low-contrast biological materials and can be used with a variety of side-entry specimen holders or autoloader devices (Dearnaley et al., Reference Dearnaley, Schleupner, Varano, Alden, Gonzalez, Casasanta, Scharf, Dukes and Kelly2019).

First, the Poseidon Select (Protochips, Inc.) system uses commercially available microchips to contain samples in liquid droplets (Fig. 2a). The base chip contains imaging windows with dimensions of 500 μm × 100 μm in x- and y-dimensions and ~30 nm in the z-dimension. Optional microwells of 10 μm in x- and y-dimensions can be arranged across the microchips. Each microwell accommodates liquid layers specified by integrated spacers (Fig. 2b).

Fig. 2. Poseidon-select specimen technique for liquid-EM. (a) The Poseidon Select holder is assembled using SiN microchips placed upon an O-ring fitting. The two-chip configuration is hermetically sealed with a faceplate and brass screws. The liquid sample is applied to the base chip (steps 1 and 2), followed by a top chip that creates a nanoscale biosphere (step 3). The sealed assembly is placed inside the TEM for EM imaging (step 4). (b) The imaging window of the base chip contains an array (500 μm × 100 μm) of integrated microwells (10 μm × 10 μm) that are ~30 nm thick. The sealed assembly can accommodate variable volumes that are randomly distributed across the chip.

Prior to use, SiN microchips are cleaned by soaking them for 2 min in acetone followed by 2 min in methanol. The cleaned chips are air-dried then glow-discharged for two cycles using a Pelco EasiGlow (Ted Pella, Inc.). Small aliquots of the liquid sample (~0.2 μL) are applied to the base chip. A glow-discharged top chip is added to the liquid layer, and the assembly is hermetically sealed by a faceplate and three brass screws. The specimen holder is placed in a TEM for imaging and analysis (Fig. 2a, steps 1–4). There is natural variability in solution flow across the chips or within the microwells, producing samples that range from dry to full capacity (Fig. 2b).

A second specimen loading option is the microchip sandwich technique. A major benefit of this system is that samples can be examined in thin liquid layers independent of specialized holders. Its use was first demonstrated by Dearnaley et al. (Reference Dearnaley, Schleupner, Varano, Alden, Gonzalez, Casasanta, Scharf, Dukes and Kelly2019) to evaluate phage interactions with whole bacteria (Dearnaley et al., Reference Dearnaley, Schleupner, Varano, Alden, Gonzalez, Casasanta, Scharf, Dukes and Kelly2019). The technique yields thin liquid samples amenable to electron tomography applications. Microchip sandwiches are produced by first placing a glow-discharged microchip upon an adhesive gel pack. Liquid samples are then added to the glow-discharged microchips (Fig. 3a). Glow-discharged EM grids with attached amorphous continuous carbon film (~5 nm thick) are placed on top of the microchip containing the liquid sample. The sandwiches are gently lifted out of the gel pack with forceps and loaded into the tip of a side-entry specimen holder or an autoloader device (Fig. 3b). Microchips composed of SiN or silicon dioxide (SiO) can contain single imaging windows as large as 1,000 μm × 1,000 μm in x- and y-dimensions and 75–100 nm in the z-dimension. Arrays of imaging windows are also useful options with dimensions of 250 μm × 250 μm in the x- and y-axes and 5–10 nm thick (Simpore, Inc.) (Fig. 3c). In an alternate configuration, EM grids may be placed on the gel pack with the liquid sample being added directly to the grids. Microchips are then placed on the wet grids. This configuration is desired as the sample material does not adhere well to the microchip surface.

Fig. 3. The microchip sandwich technique. (a) A stepwise schematic of a microchip sandwich technique. Microchips are placed on a gel pack and the liquid sample is added to the substrate. The sample randomly spreads across the microchip surface. A carbon support film deposited on a gold EM grid is added to the wet microchip. (b) The microchip sandwich is placed at the tip of a room temperature specimen holder and sealed by the clipping mechanism (left panel). In comparison, a microchip sandwich is shown in the clipping device used with an autoloader mechanism (right panel). (c) Side view drawing of a microchip assembly. Carbon layers are ~5 nm thick, while SiN microchips contain imaging windows that are 5–10 nm thick within an array of imaging windows. SiO microchips can contain windows that are ~1,000 μm × 1,000 μm in x- and y-dimensions.

By employing autoloader devices, multiple microchip sandwiches can be produced and viewed in parallel. The clipping device along with a clipped microchip assembly is shown in Figure 3b prior to loading samples into a Titan Krios (ThermoFisher Scientific). The assembly is mechanically sealed using standard grid clips, and the autoloader can accommodate up to 12 samples per imaging session. A clear advantage of automated specimen production is that many samples can be screened for optimum liquid thickness and dose conditions in a rapid time frame. One disadvantage is the inability to flow analytes or other liquids during imaging sessions. For single-particle analysis, however, the automated microchip sandwich technique can easily compete with high-throughput results obtained with cryo-EM.

EM Data Collection for Single-Particle Analysis in Liquid

Once adequate samples have been produced and screened, there are different modes of data recordings beneficial to liquid specimens. One method involves recording long-framed movies through in situ applications using a direct electron detector (Dearnaley et al., Reference Dearnaley, Schleupner, Varano, Alden, Gonzalez, Casasanta, Scharf, Dukes and Kelly2019; Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). The second method involves collecting short-framed movies of individual images for single-particle analysis (Casasanta et al., Reference Casasanta, Jonaid, Kaylor, Luqiu, Solares, Schroen, Dearnaley, Wilson, Dukes and Kelly2021). Movies can be recorded in linear mode or in counting mode using different direct detectors. Counting mode is preferred if longer exposures are needed for beam-sensitive samples and may provide improved resolution at higher voltages. Linear mode has proven adequate for structural observations in liquid at the level of ~3 Å resolution for stable specimens such as human viruses. Here, we describe these imaging practices in greater detail in a manner analogous to single-particle cryo-EM practices.

Long-framed movies can be acquired using direct detectors recording at 40–100 frames per second (fps), and protocols are adjusted depending on the inherent contrast and stability of the sample. Overall, time frames for these recordings may range from 10 s to 2 min or longer. As biological materials are beam-sensitive and produce weak-phase contrast in comparison to hard materials, frame averaging routines are essential. The electron dose is fractionated across each movie and low-dose conditions are used (1–10 electrons/Å2/s) at 200–300 kV. Magnifications vary between 15,000× and 100,000×, depending on the features and resolution goals of the project. Long-framed movies are segmented into individual time points to provide recordings that are suitable for structural analysis. For example, movies that are 10 s in length can be divided into 10 movies having 1 s increments. Segmentation procedures are performed using the SPLIT function in the EMAN2 software package (Tang et al., Reference Tang, Peng, Baldwin, Mann, Jiang, Rees and Ludtke2007). Movies and frames are subjected to motion correction procedures using the MotionCor 2 software package and contrast transfer function (CTF) correction using CTFind (Tang et al., Reference Tang, Peng, Baldwin, Mann, Jiang, Rees and Ludtke2007; Rohou & Grigorieff, Reference Rohou and Grigorieff2015). Both packages are integrated into structure analysis programs such as cryoSPARC or RELION (Scheres, Reference Scheres2012; Punjani et al., Reference Punjani, Rubinstein, Fleet and Brubaker2017). After these corrections are performed, 2D classification procedures are used to assess sample heterogeneity as a quality control measure (Fig. 4).

Fig. 4. Workflow of EM data collection and single-particle procedures using automated routines. Long-framed movies are segmented and short-framed movies are used without segmentation. Frames are corrected for drift and the CTF of the instrument using automated routines in cryoSPARC or RELION. Reference-free classification is performed to assess heterogeneity in the images (left panel). For 3D structure determination, ab initio models are calculated from subsets of data and used for automated particle selection procedures. The output from 3D classification provides information for particle grouping to calculate a higher-resolution map. The final map is masked, sharpened. Structural resolution is determined by comparing half-maps (FSC-0.143 cutoff) and by comparing the experimental map to a perfect theoretical model (Cref(0.5) criteria).

Short-framed movies are recorded under low-dose conditions using automated data collection programs, such as SerialEM (Mastronarde, Reference Mastronarde2018). The graphical interface for the program allows the user to specify the detector, frame rate, dose, and selected areas to image. Spacing between imaging areas is provided by the user to avoid beam damage due to overlap between exposures. Imaging parameters for the DE-12 direct detector typically include 1–2 s exposures at 40 fps and a total dose of ~40–60 electrons/Å2/pixel. These factors may vary depending upon the contrast and stability of the sample. Once individual movies are acquired, image frames are corrected for drift, beam-induced movement, and CTF using MotionCor2 and CTFind, similar to long-framed exposures. Classification routines provide insight as to resolution and features in the particle data. Selected high-resolution images are further processed for 3D structure determination.

For automated particle selection procedures, ab initio models calculated in cryoSPARC or RELION employ subsets of images to define sufficient features for template-matching algorithms. Following rigorous particle selection over the entire dataset, 2D and 3D classification routines are deployed (Fig. 4, right panel). Particles with higher-resolution features are combined and used for 3D refinement procedures. The final maps are masked and sharpened using structure determination programs such as Phenix (Adams et al., Reference Adams, Afonine, Bunkóczi, Chen, Davis, Echols, Headd, Hung, Kapral, Grosse-Kunstleve, McCoy, Moriarty, Oeffner, Read, Richardson, Richardson, Terwilliger and Zwart2010). Structural resolution is calculated by assessing FSC values plotted against spatial resolution (1/Å). Consensus resolution values at the 0.143 cutoff value are referred to as “Gold-standard” resolution in half-map comparisons (Henderson et al., Reference Henderson, Sali, Baker, Carragher, Devkota, Downing, Egelman, Feng, Frank, Grigorieff, Jiang, Ludtke, Medalia, Penczek, Rosenthal, Rossmann, Schmid, Schröder, Steven, Stokes, Westbrook, Wriggers, Yang, Young, Berman, Chiu, Kleywegt and Lawson2012). Additional comparisons to perfect reference models at the expected resolution value are an unbiased means to analyze experimental and theoretical resolution values using the Cref(0.5) criteria (Rosenthal & Rubinstein, Reference Rosenthal and Rubinstein2015). Here, we further demonstrate these routines through two test subjects with varying dimensions and inherent symmetries.

Case Studies: AAV and the SARS-CoV-2 N Protein

The structure of a new AAV-based COVID vaccine candidate was recently determined at ~3.22 Å resolution using liquid-EM (Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). To accomplish this feat, we used the Poseidon Select system and SiN microchips containing integrated microwells. The imaging windows were 500 μm × 100 μm with microwells arranged across the bottom chip surface, 10 μm in the x- and y-dimensions and ~30 nm in the z-dimension. Purified AAV (0.1 mg/mL) was prepared in an aqueous buffer solution containing 50 mM HEPES (pH 7.5), 150 mM NaCl, 10 mM MgCl2, and 10 mM CaCl2. The quantities of AAV (~0.2 μL) were applied to glow-discharged microchips containing microwells and incubated for 2 min in the tip of the specimen holder. A second glow-discharged chip was added to the base chips, and the assembly was hermetically sealed in the holder and loaded into a ThermoFisher F200C TEM. Long-framed movies of diffusing viruses were acquired using in situ mode on a DE-12 direct detector. Time spans of the recordings ranged from 5 s to 2 min. Imaging parameters included a frame rate of 40 fps at a magnification of 92,000×. Movies were exported with 2-fold image binning for a sampling of 1.01 Å/pixel. The beam intensity was 1 e−/Å2/s, and binned movie frames were used for downstream image processing.

Resulting movies were analyzed for migration patterns, and longer-framed movies are divided into 2-s increments to better manage diffusion that can limit spatial resolution (Fig. 5a). The optimal movies with limited dose, drift, and beam damage occurred in a ~20 s time span using a total dose of 20 e−/Å2. Individual frames were extracted and split into subsequent movies with each new movie containing 20 frames. Frames were drift-corrected using MotionCor2, and all frames were imported into the cryoSPARC and RELION software packages. Each AAV particle has 60 copies of its asymmetric protomer. By imposing icosahedral symmetry operators, we incorporated 16,800 individual particle views into the 3D structure (Fig. 5b). Extracted structures of the VP1 protomer subunit showed features consistent with a model of the VP1 crystal structure (PDB code, 3KIC, A chain; Lerch et al., Reference Lerch, Xie and Chapman2010). The resulting EM map was determined at 3.22 Å resolution, according to the 0.143-FSC criteria and the Cref(0.5) validation analysis. Slices through the EM map show features expected at the appropriate level of spatial resolution (Fig. 5c). The recently reported AAV structure represents the first biological assembly determined at high resolution while contained in a liquid environment (Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). In addition to the high-resolution 3D structure, the liquid-EM data also contained dynamic reconstructions that were not present in cryo-EM data determined for the same AAV preparation.

Fig. 5. Structural features of AAV in solution. (a) Steps for data acquisition using the DE-12 direct detector included 40 fps at 200 kV and a cumulative dose of 20 e−/Å−2. Regions of interest at 5, 10, and 20 s and fourier transforms demonstrate sub-nanometer information in the data. The edge of each panel is ~1/3 Å. The left side shows CTF estimates, and the right side shows experimental data. (b) The density of the VP1 capsid protein extracted from the solution structure. The density was compared with a model of the VP1 crystal structure (PDB code, 3KIC, A chain). The highlighted loop conveys specificity for the virus to evade the immune system. The model structure of VP1 calculated at 3.2 Å resolution shows good agreement with the EM structure. Scale bar is 10 Å. (c) The high-resolution structure of AAV determined from particles in solution. Slices through the AAV assembly with the atomic model (blue; PDB code, 3KIC, all chains) placed in the EM map (gray). A magnified region near the 5-fold axis shows some side chains present and distinct within the density. Scale bar is 5 nm. Adapted from Jonaid et al. (Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021).

For additional proof-of-concept experiments, we applied the same liquid-EM methodologies to a more challenging subject, the SARS-CoV-2 N protein. We recently determined the first cryo-EM structure of the asymmetric N protein monomer (~48 kDa), and this system presented a unique opportunity to test the limits of our structural assessments. While the project is a work in progress, we describe our initial results to inspire new experimental design opportunities.

The SARS-CoV-2 nucleocapsid (N) protein is highly antigenic in humans making it a key analyte to detect antibodies for virus exposure. N protein is also the target in rapid antigen tests along with new pandemic-preparedness tools, such as therapeutics to fight long-COVID. The SARS-CoV-2 N protein was expressed and purified from bacteria and is commercially available from RayBiotech, Inc.. Previous biochemical analysis suggests that N protein monomers, dimers, and oligomers may be present in their active form in the body (Hsieh et al., Reference Hsieh, Chang, Huang, Lee, Hsiao, Kou, Chen, Chang, Huang and Chang2005; Chang et al., Reference Chang, Hou, Chang, Hsiao and Huang2014; Peng et al., Reference Peng, Du, Lei, Dorje, Qi, Luo, Gao and Song2020).

Cryo-EM samples of the N protein were prepared in Tris buffer solution [0.1mg/mL in 20 mM Tris (pH 7.5), 150 mM NaCl, 10 mM MgCl2, and 10 mM CaCl2] and tethered to SiN microchips coated with nickel-nitrilotriacetic acid (Ni-NTA) layers (Casasanta et al., Reference Casasanta, Jonaid, Kaylor, Luqiu, Solares, Schroen, Dearnaley, Wilson, Dukes and Kelly2021). For liquid-EM samples, we used SiN microchips with an array of imaging windows (10 μm × 10 μm) that were etched down to 10 nm thick. The microchip sandwich technique was employed and carbon-coated gold grids were applied to wet microchips containing 1.0 mg/mL of N protein sample in Tris buffer solution. EM data were collected under low-dose conditions using a F200C TEM (ThermoFisher Scientific) operating at 200 kV (Fig. 6a). Automated routines were implemented in SerialEM, and movies were acquired using a DE-12 direct detector in linear mode at 0.5 s exposures at 40 fps. Motion correction was applied during image acquisition at a magnification of ~45,000×. Final sampling at the specimen level was 1.05 Å/pixel, and imaging parameters were equivalent to those used for cryo-EM. Particles were selected from the images using automated routines in the cryoSPARC software package. The particle data were subjected to 2D classification, and the resulting averages were compared to our 4.3 Å cryo-EM structure of the N protein monomer (Fig. 6b).

Fig. 6. Structural assessment of the SARS-CoV-2 N protein in liquid. (a) Data collection and image processing steps for individual movies in regions containing N protein particles. Low-dose conditions were used to acquire movies on a DE-12 detector at 40 fps for 0.5 s exposures. Drift and CTF corrections were performed on movie frames and fourier transforms demonstrate sub-nanometer information is present. (b) 2D class averages from liquid-EM data (top row) are in good agreement with the cryo-EM structure displayed in different orientations (bottom row). Box size is 10 nm, and particle length is ~5 nm. The N-terminal domain (NTD), RNA-binding site, dimerization domain, and C-terminal domain (CTD) of the N protein are indicated for the cryo-EM structure. (c) Class averages (top row) and colored contour maps (middle row) show dynamic and compact views of N protein dimers in liquid. Box size is 15 nm, and particle length is ~7 nm. Colors represent significant levels: blue (0–1σ); magenta (1–2σ); yellow (2–3σ); and green (>3 s). Corresponding models of N protein dimers (bottom row) were placed in similar orientations, as the class averages for comparison.

Liquid-EM data analysis was performed using ~10,000 asymmetric particles selected from the images. While the presence of N protein monomers was expected, based on cryo-EM findings, we also noted the presence of N protein dimers for the first time in our analyses. Monomers tended to naturally associate in the liquid enclosure during the imaging process. The molecular architecture of the N protein dimers appeared to show a spectrum of features from dynamic to compact in nature. This interpretation was based on the degree of negative space and relative flexibility present in the 2D class averages. Colored contour maps that displayed a larger degree of open space and rotational freedom in the assemblies were considered to be more dynamic (Fig. 6c, left panel). Classes with less negative space and more concentrated intensities in colored contour maps were considered to be more compact (Fig. 6c, right panel). Differences in the dynamic and compact states are likely influenced by conformational flexibility, Brownian motion, or transient protein associations in solution. Beam-induced movements and drift were corrected using MotionCor2 during image processing procedures, minimizing their role in interpreting these effects.

To better interpret the dimer structures, we used N protein structures determined from prior cryo-EM results (Casasanta et al., Reference Casasanta, Jonaid, Kaylor, Luqiu, Solares, Schroen, Dearnaley, Wilson, Dukes and Kelly2021). We positioned N protein dimerization domains in close proximity to each other to form a theoretical model of the dimer showing features consistent with the class averages (Fig. 6c, bottom row). The models were in good agreement with the experimentally determined data and contour maps. Expanding these promising findings with more data and 3D structural analysis will undoubtedly shed light on the precise manner in which dimers and oligomers form in solution, along with insights for RNA-binding activities.

Conclusions

While future challenges occupy the road ahead for the liquid-EM field, there are also exciting opportunities to elevate current workflows by incorporating automated routines. Recent work showed that virus particles in liquid were remarkably stable under low-dose conditions, producing structures with spatial resolutions on par with cryo-EM (Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). High frame rate direct detectors and automated data collection procedures can accelerate imaging procedures yielding up to ~200 movies/h. These new methods permitted us to record real-time movies of viral assemblies and small proteins for extended periods of time in liquid. While some side chains were resolved in EM maps of AAV, some disordered regions persisted presenting a challenge for determining atomic details in solution. Unresolved regions in EM maps were due to natural dynamics or Brownian motion of the biological assemblies. As beam-induced movements and specimen drift can now be corrected computationally, the field is moving toward observing live protein dynamics decoupled from beam effects.

Interestingly, we found that images of liquid specimens displayed enhanced contrast values with greater SNRs in the range of 2–3 (Jonaid et al., Reference Jonaid, Dearnaley, Casasanta, Kaylor, Dukes, Spillman and Kelly2021). By comparison, using the same electron dose rate adjusted for exposure time, images of cryo-EM specimens have much lower contrast with an SNR of ~0.1. We believe the enhanced SNR in liquid specimens is due to thinner liquid layers that encompass the specimen, relative to thicker ice layers. Virus particles in liquid layers exhibited ~88% of the maximum achievable contrast. Extrapolating this information, one may expect a ~2.8 Å resolution if 100% SNR is achieved under similar imaging conditions. Resolution-limiting factors to consider include solution thickness, beam effects, and natural protein movements. Recent dose analysis studies suggested that the biological samples may be more stable in a liquid environment than in ice (Keskin & de Jonge, Reference Keskin and de Jonge2018; Keskin et al., Reference Keskin, Kunnas and de Jonge2019). Higher beam tolerance is attributed, in part, to charge equilibration following the formation of solvated electrons.

Typically, the number of particles extracted from one microchip sample can range from 10,000 to 100,000, depending upon the number of areas containing optimum liquid thickness values, similar to cryo-EM methods. Here, we provide compelling evidence that very small, asymmetric proteins can also be analyzed using the same techniques that work for large symmetric particles, while it is difficult to fully control liquid thicknesses over long-ranged areas. A better perspective, adopted from the cryo-EM field, is to simply acquire movies of specimens present in thin layers randomly distributed through the imaging space. Programs such as SerialEM use atlas imaging that permits the user to select only regions of optimal thickness while providing thousands of good images per specimen (Mastronarde, Reference Mastronarde2018). Coupled with autoloader devices and automated instruments, such as the Titan Krios, one can expand data collection routines to include thousands of images across multiple samples, without the need to control for solution thicknesses. Overcoming resolution limits in highly symmetric structures may involve sub-volume averaging procedures currently implemented by the single-particle imaging field.

Overall, there remains a great need to understand the properties of biological molecules in liquid, as they would naturally occur in tissues and cells. Liquid-EM technology can yield information for both structure and function to help address this need. Incorporating automated routines into current workflows will undoubtedly elevate the field to provide live insights of vaccine candidates, antibody-based therapies, or drug-based nanoparticles to benefit human health and better manage disease progression.

Acknowledgments

The authors thank Dr. Carol Bator (PSU, Huck Institutes of the Life Sciences) for her expertise and assistance with EM data collection. The authors acknowledge Dr. Luk H. Vandenberghe (Harvard Medical School, Department of Ophthalmology) for providing purified AAV-3.

Author contributions statement

M.A.C., L.K., M.J.D., M.S.S., J.L.G., and D.F.K. assisted with specimen preparation procedures and data collection. G.M.J., W.J.D., M.A.C., L.K., S.B., and D.F.K. assisted with experimental design and data processing. All authors assisted with writing and reviewing the manuscript.

Financial support

This work was supported by the National Institutes of Health and the National Cancer Institute (Nos. R01CA193578, R01CA227261, and R01CA219700 to D.F.K.).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Footnotes

†

These authors have contributed equally to this work.

References

Adams, PD, Afonine, PV, Bunkóczi, G, Chen, VB, Davis, IW, Echols, N, Headd, JJ, Hung, L-W, Kapral, GJ, Grosse-Kunstleve, RW, McCoy, AJ, Moriarty, NW, Oeffner, R, Read, RJ, Richardson, DC, Richardson, JS, Terwilliger, TC & Zwart, PH (2010). PHENIX: A comprehensive python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr 66, 213–221.CrossRefGoogle ScholarPubMed
Ahmad, N, Wang, G, Nelayah, J, Ricolleau, C & Alloyeau, D (2017). Exploring the formation of symmetric gold nanostars by liquid-cell transmission electron microscopy. Nano Lett 17, 4194–4201.CrossRefGoogle ScholarPubMed
Alloyeau, D, Dachraoui, W, Javed, Y, Belkahla, H, Wang, G, Lecoq, H, Ammar, S, Ersen, O, Wisnet, A, Gazeau, F & Ricolleau, C (2015). Unravelling kinetic and thermodynamic effects on the growth of gold nanoplates by liquid transmission electron microscopy. Nano Lett 15, 2574–2581.CrossRefGoogle ScholarPubMed
Bammes, BE, Rochat, RH, Jakana, J, Chen, D-H & Chiu, W (2012). Direct electron detection yields cryo-EM reconstructions at resolutions beyond 3/4 nyquist frequency. J Struct Biol 177, 589–601.CrossRefGoogle ScholarPubMed
Casasanta, MA, Jonaid, GM, Kaylor, L, Luqiu, WY, Solares, MJ, Schroen, ML, Dearnaley, WJ, Wilson, J, Dukes, MJ & Kelly, DF (2021). Microchip-based structure determination of low-molecular weight proteins using cryo-electron microscopy. Nanoscale 13, 7285–7293.CrossRefGoogle ScholarPubMed
Chang, C-K, Hou, M-H, Chang, C-F, Hsiao, C-D & Huang, T-H (2014). The SARS coronavirus nucleocapsid protein–forms and functions. Antiviral Res 103, 39–50.CrossRefGoogle ScholarPubMed
Chen, Q, Smith, JM, Park, J, Kim, K, Ho, D, Rasool, HI, Zettl, A & Alivisatos, AP (2013). 3D motion of DNA-Au nanoconjugates in graphene liquid cell electron microscopy. Nano Lett 13, 4556–4561.CrossRefGoogle ScholarPubMed
Dearnaley, WJ, Schleupner, B, Varano, AC, Alden, NA, Gonzalez, F, Casasanta, MA, Scharf, BE, Dukes, MJ & Kelly, DF (2019). Liquid-cell electron tomography of biological systems. Nano Lett 19, 6734–6741.CrossRefGoogle ScholarPubMed
de Jonge, N & Ross, FM (2011). Electron microscopy of specimens in liquid. Nat Nanotechnol 6, 695–704.CrossRefGoogle ScholarPubMed
Deng, W, Marshall, NC, Rowland, JL, McCoy, JM, Worrall, LJ, Santos, AS, Strynadka, NCJ & Finlay, BB (2017). Assembly, structure, function and regulation of type III secretion systems. Nat Rev Microbiol 15, 323–337.CrossRefGoogle ScholarPubMed
Dukes, MJ, Thomas, R, Damiano, J, Klein, KL, Balasubramaniam, S, Kayandan, S, Riffle, JS, Davis, RM, McDonald, SM & Kelly, DF (2014). Improved microchip design and application for in situ transmission electron microscopy of macromolecules. Microsc Microanal 20, 338–345.CrossRefGoogle ScholarPubMed
Dunn, G, Adiga, VP, Pham, T, Bryant, C, Horton-Bailey, DJ, Belling, JN, LaFrance, B, Jackson, JA, Barzegar, HR, Yuk, JM, Aloni, S, Crommie, MF & Zettl, A (2020). Graphene-sealed flow cells for in situ transmission electron microscopy of liquid samples. ACS Nano 14, 9637–9643.CrossRefGoogle ScholarPubMed
Frank, J (2006). Three-Dimensional Electron Microscopy of Macromolecular Assemblies: Visualization of Biological Molecules in Their Native State. New York: Oxford University Press.CrossRefGoogle Scholar
Frank, J, Zhu, J, Penczek, P, Li, Y, Srivastava, S, Verschoor, A, Radermacher, M, Grassucci, R, Lata, RK & Agrawal, RK (1995). A model of protein synthesis based on cryo-electron microscopy of the E. coli ribosome. Nature 376, 441–444.CrossRefGoogle Scholar
Gilmore, BL, Winton, CE, Demmert, AC, Tanner, JR, Bowman, S, Karageorge, V, Patel, K, Sheng, Z & Kelly, DF (2015). A molecular toolkit to visualize native protein assemblies in the context of human disease. Sci Rep 5, 14440.CrossRefGoogle ScholarPubMed
Goddard, TD, Huang, CC & Ferrin, TE (2007). Visualizing density maps with UCSF Chimera. J Struct Biol 157, 281–287.CrossRefGoogle ScholarPubMed
Henderson, R, Sali, A, Baker, ML, Carragher, B, Devkota, B, Downing, KH, Egelman, EH, Feng, Z, Frank, J, Grigorieff, N, Jiang, W, Ludtke, SJ, Medalia, O, Penczek, PA, Rosenthal, PB, Rossmann, MG, Schmid, MF, Schröder, GF, Steven, AC, Stokes, DL, Westbrook, JD, Wriggers, W, Yang, H, Young, J, Berman, HM, Chiu, W, Kleywegt, GJ & Lawson, CL (2012). Outcome of the first electron microscopy validation task force meeting. Structure 20, 205–214.CrossRefGoogle ScholarPubMed
Hsieh, P-K, Chang, SC, Huang, C-C, Lee, T-T, Hsiao, C-W, Kou, Y-H, Chen, I-Y, Chang, C-K, Huang, T-H & Chang, M-F (2005). Assembly of severe acute respiratory syndrome coronavirus RNA packaging signal into virus-like particles is nucleocapsid dependent. J Virol 79, 13848–13855.CrossRefGoogle ScholarPubMed
Hu, B, Morado, DR, Margolin, W, Rohde, JR, Arizmendi, O, Picking, WL, Picking, WD & Liu, J (2015). Visualization of the type III secretion sorting platform of Shigella flexneri. Proc Natl Acad Sci U S A 112, 1047–1052.CrossRefGoogle ScholarPubMed
Jonaid, GM, Dearnaley, WJ, Casasanta, MA, Kaylor, L, Dukes, MJ, Spillman, MS & Kelly, DF (2021). High-resolution imaging of human viruses in liquid droplets. Adv Mater 33, 1–13.Google ScholarPubMed
Keskin, S & de Jonge, N (2018). Reduced radiation damage in transmission electron microscopy of proteins in graphene liquid cells. Nano Lett 18, 7435–7440.CrossRefGoogle ScholarPubMed
Keskin, S, Kunnas, P & de Jonge, N (2019). Liquid-phase electron microscopy with controllable liquid thickness. Nano Lett 19, 4608–4613.CrossRefGoogle ScholarPubMed
Kiss, G, Chen, X, Brindley, MA, Campbell, P, Afonso, CL, Ke, Z, Holl, JM, Guerrero-Ferreira, RC, Byrd-Leotis, LA, Steel, J, Steinhauer, DA, Plemper, RK, Kelly, DF, Spearman, PW & Wright, ER (2014). Capturing enveloped viruses on affinity grids for downstream cryo-electron microscopy applications. Microsc Microanal 20, 164–174.CrossRefGoogle ScholarPubMed
Lerch, TF, Xie, Q & Chapman, MS (2010). The structure of adeno-associated virus serotype 3B (AAV-3B): Insights into receptor binding and immune evasion. Virology 403, 26–36.CrossRefGoogle ScholarPubMed
Li, Z, Baker, ML, Jiang, W, Estes, MK & Prasad, BVV (2009). Rotavirus architecture at subnanometer resolution. J Virol 83, 1754–1766.CrossRefGoogle ScholarPubMed
Liao, H-G & Zheng, H (2013). Liquid cell transmission electron microscopy study of platinum iron nanocrystal growth and shape evolution. J Am Chem Soc 135, 5038–5043.CrossRefGoogle ScholarPubMed
Mastronarde, DN (2018). Advanced data acquisition from electron microscopes with SerialEM. Microsc Microanal 24, 864–865.CrossRefGoogle Scholar
Migunov, V, Ryll, H, Zhuge, X, Simson, M, Strüder, L, Batenburg, KJ, Houben, L & Dunin-Borkowski, RE (2015). Rapid low dose electron tomography using a direct electron detection camera. Sci Rep 5, 14516.CrossRefGoogle ScholarPubMed
Murata, K & Wolf, M (2018). Cryo-electron microscopy for structural analysis of dynamic biological macromolecules. Biochim Biophys Acta Gen Subj 1862, 324–334.CrossRefGoogle ScholarPubMed
Oikonomou, CM & Jensen, GJ (2016). A new view into prokaryotic cell biology from electron cryotomography. Nat Rev Microbiol 14, 205–220.CrossRefGoogle ScholarPubMed
Pallesen, J, Murin, CD, De Val, N, Cottrell, CA, Hastie, KM, Turner, HL, Fusco, ML, Flyak, AI, Zeitlin, L & Crowe, JE (2016). Structures of ebola virus GP and sGP in complex with therapeutic antibodies. Nat Microbiol 1, 1–9.CrossRefGoogle ScholarPubMed
Park, J, Elmlund, H, Ercius, P, Yuk, JM, Limmer, DT, Chen, Q, Kim, K, Han, SH, Weitz, DA, Zettl, A & Paul Alivisatos, A (2015 a). 3D structure of individual nanocrystals in solution by electron microscopy. Science. https://science.sciencemag.org/content/349/6245/290.abstract (accessed October 1, 2021).CrossRefGoogle ScholarPubMed
Park, J, Park, H, Ercius, P, Pegoraro, AF, Xu, C, Kim, JW, Han, SH & Weitz, DA (2015 b). Direct observation of wet biological samples by graphene liquid cell transmission electron microscopy. Nano Lett 15, 4737–4744.CrossRefGoogle ScholarPubMed
Penczek, PA, Grassucci, RA & Frank, J (1994). The ribosome at improved resolution: New techniques for merging and orientation refinement in 3D cryo-electron microscopy of biological particles. Ultramicroscopy 53, 251–270.CrossRefGoogle ScholarPubMed
Peng, Y, Du, N, Lei, Y, Dorje, S, Qi, J, Luo, T, Gao, GF & Song, H (2020). Structures of the SARS-CoV-2 nucleocapsid and their perspectives for drug design. EMBO J 39, e105938.CrossRefGoogle ScholarPubMed
Pettersen, EF, Goddard, TD, Huang, CC, Couch, GS, Greenblatt, DM, Meng, EC & Ferrin, TE (2004). UCSF chimera—A visualization system for exploratory research and analysis. J Comput Chem 25, 1605–1612.CrossRefGoogle ScholarPubMed
Piffoux, M, Ahmad, N, Nelayah, J, Wilhelm, C, Silva, A, Gazeau, F & Alloyeau, D (2018). Monitoring the dynamics of cell-derived extracellular vesicles at the nanoscale by liquid-cell transmission electron microscopy. Nanoscale 10, 1234–1244.CrossRefGoogle ScholarPubMed
Punjani, A, Rubinstein, JL, Fleet, DJ & Brubaker, MA (2017). cryoSPARC: Algorithms for rapid unsupervised cryo-EM structure determination. Nat Methods 14, 290–296.CrossRefGoogle ScholarPubMed
Reboul, CF, Heo, J, Machello, C, Kiesewetter, S, Kim, BH, Kim, S, Elmlund, D, Ercius, P, Park, J & Elmlund, H (2021). SINGLE: Atomic-resolution structure identification of nanocrystals by graphene liquid cell EM. Sci Adv 7, eabe6679.CrossRefGoogle ScholarPubMed
Rohou, A & Grigorieff, N (2015). CTFFIND4: Fast and accurate defocus estimation from electron micrographs. J Struct Biol 192, 216–221.CrossRefGoogle ScholarPubMed
Rosenthal, PB & Rubinstein, JL (2015). Validating maps from single particle electron cryomicroscopy. Curr Opin Struct Biol 34, 135–144.CrossRefGoogle ScholarPubMed
Rutledge, EA, Halbert, CL & Russell, DW (1998). Infectious clones and vectors derived from adeno-associated virus (AAV) serotypes other than AAV type 2. J Virol 72, 309–319.CrossRefGoogle ScholarPubMed
Scheres, SHW (2012). RELION: Implementation of a Bayesian approach to cryo-EM structure determination. J Struct Biol 180, 519–530.CrossRefGoogle ScholarPubMed
Serra-Maia, R, Kumar, P, Meng, AC, Foucher, AC, Kang, Y, Karki, K, Jariwala, D & Stach, EA (2021). Nanoscale chemical and structural analysis during in situ scanning/transmission electron microscopy in liquids. ACS Nano 15, 10228–10240.CrossRefGoogle ScholarPubMed
Shi, Y, Di Giammartino, DC, Taylor, D, Sarkeshik, A, Rice, WJ, Yates, JR 3rd, Frank, J & Manley, JL (2009). Molecular architecture of the human pre-mRNA 3′ processing complex. Mol Cell 33, 365–376.CrossRefGoogle ScholarPubMed
Sousa, D & Grigorieff, N (2007). Ab initio resolution measurement for single particle structures. J Struct Biol 157, 201–210.CrossRefGoogle ScholarPubMed
Tang, G, Peng, L, Baldwin, PR, Mann, DS, Jiang, W, Rees, I & Ludtke, SJ (2007). EMAN2: An extensible image processing suite for electron microscopy. J Struct Biol 157, 38–46.CrossRefGoogle ScholarPubMed
Varano, A, Alden, N, Dearnaley, W, Casasanta, M, Damiano, J, McConnell, J, Dukes, M & Kelly, DF (2019). Customizable Cryo-EM chips improve 3D analysis of macromolecules. Microsc Microanal 25, 1310–1311.CrossRefGoogle Scholar
Watanabe, R, Castillon, G, Avalos, RD, Ellisman, M & Saphire, E (2021). Visualization of intracellular ebola virus nucleocapsid assembly by cryo-electron tomography. Microsc Microanal 27, 1708–1711.CrossRefGoogle Scholar
Yuk, JM, Park, J, Ercius, P, Kim, K, Hellebusch, DJ, Crommie, MF, Lee, JY, Zettl, A & Alivisatos, AP (2012). High-resolution EM of colloidal nanocrystal growth using graphene liquid cells. Science 336, 61–64.CrossRefGoogle ScholarPubMed
Zheng, H, Smith, RK, Jun, Y-W, Kisielowski, C, Dahmen, U & Alivisatos, AP (2009). Observation of single colloidal platinum nanocrystal growth trajectories. Science 324, 1309–1312.CrossRefGoogle ScholarPubMed
Zheng, SQ, Palovcak, E, Armache, J-P, Verba, KA, Cheng, Y & Agard, DA (2017). Motioncor2: Anisotropic correction of beam-induced motion for improved cryo-electron microscopy. Nat Methods 14, 331–332.CrossRefGoogle ScholarPubMed
Figure 0

Fig. 1. Steps in the liquid-EM workflow that can be enhanced by automated practices. Liquid-EM sample preparation typically involves specialized side-entry specimen holders (left panel). Using this convention, liquid layers are sandwiched between two materials or microchips. Adapting cryo-EM autoloaders for this step permits researchers to view multiple samples in parallel. EM data collection (center panel) can be enhanced by automated data collection programs to output real-time movies of the imaged samples. Computing procedures (right panel) are accelerated using software procedures performed as parallel jobs on multiple processors or graphics cards (GPUs).

Figure 1

Fig. 2. Poseidon-select specimen technique for liquid-EM. (a) The Poseidon Select holder is assembled using SiN microchips placed upon an O-ring fitting. The two-chip configuration is hermetically sealed with a faceplate and brass screws. The liquid sample is applied to the base chip (steps 1 and 2), followed by a top chip that creates a nanoscale biosphere (step 3). The sealed assembly is placed inside the TEM for EM imaging (step 4). (b) The imaging window of the base chip contains an array (500 μm × 100 μm) of integrated microwells (10 μm × 10 μm) that are ~30 nm thick. The sealed assembly can accommodate variable volumes that are randomly distributed across the chip.

Figure 2

Fig. 3. The microchip sandwich technique. (a) A stepwise schematic of a microchip sandwich technique. Microchips are placed on a gel pack and the liquid sample is added to the substrate. The sample randomly spreads across the microchip surface. A carbon support film deposited on a gold EM grid is added to the wet microchip. (b) The microchip sandwich is placed at the tip of a room temperature specimen holder and sealed by the clipping mechanism (left panel). In comparison, a microchip sandwich is shown in the clipping device used with an autoloader mechanism (right panel). (c) Side view drawing of a microchip assembly. Carbon layers are ~5 nm thick, while SiN microchips contain imaging windows that are 5–10 nm thick within an array of imaging windows. SiO microchips can contain windows that are ~1,000 μm × 1,000 μm in x- and y-dimensions.

Figure 3

Fig. 4. Workflow of EM data collection and single-particle procedures using automated routines. Long-framed movies are segmented and short-framed movies are used without segmentation. Frames are corrected for drift and the CTF of the instrument using automated routines in cryoSPARC or RELION. Reference-free classification is performed to assess heterogeneity in the images (left panel). For 3D structure determination, ab initio models are calculated from subsets of data and used for automated particle selection procedures. The output from 3D classification provides information for particle grouping to calculate a higher-resolution map. The final map is masked, sharpened. Structural resolution is determined by comparing half-maps (FSC-0.143 cutoff) and by comparing the experimental map to a perfect theoretical model (Cref(0.5) criteria).

Figure 4

Fig. 5. Structural features of AAV in solution. (a) Steps for data acquisition using the DE-12 direct detector included 40 fps at 200 kV and a cumulative dose of 20 e−/Å−2. Regions of interest at 5, 10, and 20 s and fourier transforms demonstrate sub-nanometer information in the data. The edge of each panel is ~1/3 Å. The left side shows CTF estimates, and the right side shows experimental data. (b) The density of the VP1 capsid protein extracted from the solution structure. The density was compared with a model of the VP1 crystal structure (PDB code, 3KIC, A chain). The highlighted loop conveys specificity for the virus to evade the immune system. The model structure of VP1 calculated at 3.2 Å resolution shows good agreement with the EM structure. Scale bar is 10 Å. (c) The high-resolution structure of AAV determined from particles in solution. Slices through the AAV assembly with the atomic model (blue; PDB code, 3KIC, all chains) placed in the EM map (gray). A magnified region near the 5-fold axis shows some side chains present and distinct within the density. Scale bar is 5 nm. Adapted from Jonaid et al. (2021).

Figure 5

Fig. 6. Structural assessment of the SARS-CoV-2 N protein in liquid. (a) Data collection and image processing steps for individual movies in regions containing N protein particles. Low-dose conditions were used to acquire movies on a DE-12 detector at 40 fps for 0.5 s exposures. Drift and CTF corrections were performed on movie frames and fourier transforms demonstrate sub-nanometer information is present. (b) 2D class averages from liquid-EM data (top row) are in good agreement with the cryo-EM structure displayed in different orientations (bottom row). Box size is 10 nm, and particle length is ~5 nm. The N-terminal domain (NTD), RNA-binding site, dimerization domain, and C-terminal domain (CTD) of the N protein are indicated for the cryo-EM structure. (c) Class averages (top row) and colored contour maps (middle row) show dynamic and compact views of N protein dimers in liquid. Box size is 15 nm, and particle length is ~7 nm. Colors represent significant levels: blue (0–1σ); magenta (1–2σ); yellow (2–3σ); and green (>3 s). Corresponding models of N protein dimers (bottom row) were placed in similar orientations, as the class averages for comparison.