Democratizing cryo-EM: Broadening access to an expanding field

Science in March 2020

LINK: https://www.sciencemag.org/features/2020/03/democratizing-cryo-em-broadening-access-expanding-field

By Chris Tachibana

Cryo-electron microscopy (cryo-EM) yields atomic-level structures of megacomplexes and tiny compounds. How can your lab get access to this versatile method?

For decades, says Bridget Carragher, cryo-EM was a “niche, hole-in-the-wall” field. But in 2017, cryo-EM passed nuclear magnetic resonance (NMR) spectroscopy for number of annual entries in the Protein Data Bank, the world’s sole repository for 3D structural data on proteins, nucleic acids, and large biological molecules. And now it’s gaining on the granddaddy of structural methods, X-ray crystallography.  

Carragher leads a cryo-EM facility at the New York Structural Biology Center, which is supported by the U.S. National Institutes of Health (NIH) and the Simons Foundation. Two other NIH-funded centers are at Stanford University and Oregon Health & Science University (OHSU). “The trend everywhere is for national cryo-EM facilities,” says Poul Nissen, structural biologist at Aarhus University, which is the Danish national facility, together with the University of Copenhagen.

National centers serve a cryo-EM community that is rapidly expanding as software and hardware breakthroughs, especially in electron detectors, demonstrate how cryo-EM can advance basic research, drug development, and even solar-cell technology (1,2).

Crystallographic resolution—without crystals, but at a cost

Unlike X-ray crystallography, cryo-EM does not require crystallized samples. This eliminates a time-consuming step and allows atomic-level reconstructions of lumpy complexes and integral membrane proteins that have resisted crystallization. It can show conformational changes, such as ribosomes flexing their structure as they go through protein synthesis (3).

Cryo-EM works with unstained, aqueous samples. For single-particle analysis (SPA), its most common application, researchers drop samples onto a grid that is flashcooled by being plunged into liquid ethane. This freezing—or rather vitrifying—is so rapid that sample molecules are immobilized with their structure preserved and without ice crystals that interfere with transmission electron microscopy (TEM). Researchers then take thousands of TEM images by beaming electrons through the sample. Molecules caught in random orientations scatter the electrons, creating patterns used to generate 3D models.

Craig Yoshioka, codirector of the NIH cryo-EM center at OHSU, points out a promising development: Crystallographers who had truncated or mutated proteins to coax them into crystals can now study full-length wildtype proteins using cryo-EM. “This should better represent targets in their native states,” he says, “including with posttranslational modifications like glycosylation.”   

Currently, SPA works best with large samples around 200 kDa, so researchers with smaller proteins might turn to microcrystal electron diffraction (microED), a cryo-EM method with a larger size range. Another issue with SPA is that it uses cell extracts; but inside cells, says University of California, San Diego biophysicist Elizabeth Villa, “proteins aren’t floating in water. They’re packed with other components, interacting with them, or forming networks that break up during extraction.” Villa uses cryo-electron tomography (cryo-ET), which images sections of cells or even tissues, to visualize components in situ.

And cryo-EM has an overarching drawback: cost. Top-of-the-line, 300-kiloelectron volt (keV) cryo-EM machines are around USD 5–7 million, with added costs for space, service contracts, and experienced staff. Pharmaceutical companies may have in-house facilities or use a company like NanoImaging Services. Most cryo-EM clients are from pharma or biotech, says Carragher, a cofounder. Example projects include analyzing vaccines, antibodies, and drug targets. The company is rare among cryo-EM contractors in owning its own equipment, with others often using instruments at partner institutes.

Major research institutions also invest in cryo-EM facilities, but smaller universities can’t afford them. However, scientists including Gabriel Lander’s group at Scripps Research have revealed single-angstrom (Å) structures of proteins using less-powerful 100-keV or 200-keV microscopes (4) that cost USD 1–2 million. These results encourage scientists who call for democratizing cryo-EM with more affordable, workhorse instruments (5).

Community service

Cryo-EM access should increase thanks to the new NIH centers, which have cutting-edge equipment and a focus on service and training—center personnel are not allowed to be coauthors on users’ publications. At full capacity, the OHSU site will be collecting data 24/7 on 200–300 active projects at a time and training 50-plus visiting scientists a year, Yoshika estimates. He expects up to hundreds of reconstructions per year per center.

And services are free. “You write a proposal,” Carragher says, “and if it’s accepted based on criteria, such as scientific merit, feasibility, and need, you get cryo-EM time.” This model is similar to national synchrotron facilities, and many, such as the United Kingdom’s Electron Bio-Imaging Centre at the Diamond Light Source (Oxfordshire) and the Brazilian Nanotechnology National Laboratory (LNNano) of the Brazilian Center for Research in Energy and Materials (Campinas), are located close to synchrotrons.

LNNano is the only cryo-EM facility in Latin America, and is supported by government and State of São Paulo funding. Industrial clients are charged for services, but service and training are free for academic researchers after project evaluation, says LNNano researcher Rodrigo Portugal.

LNNano Senior Scientist Marin van Heel says cryo-EM is a powerful tool for structure-based drug and vaccine design, so it is essential in the region because of “big needs, like in neglected diseases such as Zika.” SPA research is underway at LNNano with collaborators in Brazil, Peru, Uruguay, and Argentina.

Besides cost, the major burden at LNNano facilities is brain drain. Despite holding multiple workshops and the annual Brazil School for Single Particle Cryo-EM, “people get headhunted away to a center or pharma company in another country,” van Heel says.

“It’s THE issue”

Cryo-EM software and hardware have “advanced amazingly,” Yoshika says, “but it can still be difficult to reliably take any protein from a gene to a structure.” Cryo-EM doesn’t require large crystals, but sample purity, heterogeneity, and concentration are still important.

“Sample prep isn’t an issue,” says Carragher, “it’s THE issue.” During vitrification, “particles glue themselves together, stick to the air–water interface, adopt sulky conformations, or fall apart.” Commercial automated systems make sample preparation more reliable. However, a downstream challenge is caused by terabytes of data that require dedicated workstations.

Digital developments

Cryo-EM users uniformly praise software groups for advancing data analysis and structure resolution. Open-source software, such as RELION from Sjors Scheres at the UK’s Medical Research Council Laboratory of Molecular Biology, and work by others including Niko Grigorieff at Janelia Research Campus and the University of Massachusetts, have been instrumental to the field, says Yoshioka. An up-and-coming computational advance, he notes, is real-time processing and reconstruction as data are collected.

That’s what cryoSPARC Live does. Currently in beta testing, the software comes from University of Toronto spinout Structura Biotechnology, run by brother-and-sister team Ali Punjani and Saara Virani. CryoSPARC Live adds to the cryoSPARC package of SPA tools, including 2D image curation and 3D reconstruction without prior structural knowledge.

CryoSPARC Live, Virani says, shows initial images after a few minutes,6-Å to 8-Å 3D structures in about an hour, and refined high-resolution structures a few hours later. Researchers can make real-time adjustments, such as moving the sample to focus on the best areas and deciding how much data to collect, saving time and money, she says. With demand for cryo-EM growing rapidly, the field is wrestling with commercialization issues. Punjani explains that cryoSPARC is free for academic users, while commercial clients such as pharma companies must buy a license.

 A computational angle on cryo-EM democratization, Punjani says, is to modify algorithms to get better images from lower-end microscopes. Also, cloud-hosted computation would let labs rent processing time as needed instead of investing in dedicated hardware.

Full tilt on innovations

“Single-particle will be the bread-and-butter method for high-resolution cryo-EM for a while,” Yoshioka says. Advances in other areas extend the size range for resolving structures and allow views of the cell interior.

Getting high-resolution images of proteins smaller than 100 kDa pushes the limit of current SPA. MicroED, developed by Tamir Gonen’s group, achieves atomic resolution for size ranges of complexes larger than 200 kDa to organic molecules under 10 carbon atoms. MicroED uses crystals one-billionth the size needed for X-ray crystallography, explains Gonen, now at the University of California, Los Angeles (UCLA). In microED, vitrification protects samples so that diffraction patterns are generated by rotating a single microcrystal through an electron beam, capturing all angles for 3D reconstruction of its molecules.

Gonen used microED to visualize structural changes in a channel as a sodium ion passed through. “Because we used crystals containing only about 1,000 units,” he says, “we could tease out smaller differences and capture a transition state” (6).

Medicinal chemists, forensic scientists, and drug developers are excited about the “powder-to-structure” application of microED. Gonen’s group and others published methods for 30-min identification of small molecules such as ibuprofen or biotin by structure, including in mixtures (7).

Gonen has worked with Thermo Fisher Scientific to develop relatively easy-to-use microED hardware and software. “You don’t need to know much now to get a sample into a ‘scope and collect data. It could make microED more available to the community,” he says. Steve Reyntjens, Thermo Fisher’s director of product marketing, says the microED package is easy to add as an optional item on new microscopes or as a retrofit to existing instruments.

The David Geffen School of Medicine at UCLA has a microED center that works with academic and industry scientists and offers microED training, including at an annual summit coming up in October 2020 (8).

 Cryo-ET reveals cellular contents not as they appear in textbooks or videos “with empty space, a particle, then empty space,” says Villa. It shows cells jam-packed with molecules. Instead of analyzing images of extracted proteins or complexes as SPA does, cryo-ET collects data on one cell, getting multiple views by tilting it on a grid. Averaging the images in a process similar to SPA results in 3D reconstructions.

TEM requires samples thinner than most cells, so researchers like Villa pair cryo-ET with focused ion-beam (FIB) micromachining. Villa credits Mike Marko at the New York State Department of Health for showing that this typical materials science method is feasible for vitrified samples. As a postdoc in Wolfgang Baumeister’s lab at the Max Planck Institute of Biochemistry in Munich, Germany, Villa helped advance the technique.

 Cryo-FIB uses a special cryo-EM instrument that aims a beam of large ions, such as gallium, across a vitrified sample to plane it down. “You blast it on the top and bottom until you have a window sustained by the sides of the cell that is thin enough for cryo-ET,” Villa says. If a cell’s tomography shows multiple copies of a protein or complex, subtomogram averaging of individual images of that molecule can reconstruct its 3D structure.

Villa’s lab added correlative light and electron microscopy (CLEM) to cryo-ET to determine the 14-Å structure of LRRK2. The researchers tagged the protein, which is involved in Parkinson’s disease, to locate it in cells using CLEM. Then they used FIB to mill down the cells for cryo-ET to obtain in situ 3D structures of LRRK2, including its close association with microtubules (9). LRRK2 can’t be crystallized, “but with cryo-ET and subtomogram averaging, we solved the structure while it was still in the cell,” Villa says.

According to Villa, improvements ahead for cryo-ET include developing specialized TEM grids for growing cells before inducing a change, such as stimulating neurons or exposing human cells to medicines. Researchers using light microscopy could choose when to vitrify the sample for cryo-ET, Villa says, “to see at high resolution what happened at the point you did something to the cell.”

Sample preparation for cryo-ET is low throughput, but a variation of CLEM with multiplexed fluorescent markers, developed by John Briggs’ group at the European Molecular Biology Laboratory in Germany, could more quickly identify cells for cryo-ET. Cryo-ET will allow observation of molecules in their native environment and in whole tissues, such as molecular views of “the connectome” of neuron-to-neuron interactions, Nissen says. The Thermo Fisher second-generation FIB instrument, notes Reyntjens, has cryo-liftout capability for manipulating miniscule samples cut from vitrified tissue thinned to 100 nm–150 nm and transferred to a cryo-EM instrument for tomography.

NIH will soon launch national cryo-ET centers. The current cryo-EM centers, Carragher explains, collect tomograms if users have samples ready for cryo-ET. The national centers will provide access to equipment, plus assistance with tricky cryo-ET specimen preparation.

In addition to developments like CLEM, structural analysis that combines cryo-EM with data from multiple sources is on the rise. “Increasingly, people use crystallography, NMR, CLEM, mass spec—everything out there—to get an answer,” Carragher says. “But if we want these tools in everyone’s toolkit, they need to be more accessible.”

New ambitions

Along with solving the access problem, Nissen observes, the field should shift its perspective from focusing only on structures to “asking what the structure is doing in the cell in its native state. Getting label-free, time-resolved structures in natural contexts is the ultimate goal and also a new level of ambition to instill in students and postdocs.”

Nissen and others predict increasing industry use of cryo-EM for developing antibody therapeutics, small molecule drugs, and diagnostics. “We should also work with the medical community on unmet diagnostic needs,” he says, “where histology doesn’t show good differences between disease and healthy tissue. We might find molecular differences in tissues [by also] using cryo-ET.”

van Heel, who helped develop cryo-EM and has watched its use grow, says about working in the field, “It’s challenging at the moment, but it’s a great time to be alive. There’s no time for vacation.”

References

1. Y. Cheng, Science 361, 876–880 (2018), doi: 10.1126/science.aat4346.
2. A. A. Herzing, D. M. DeLongchamp, Matter 1, 1106–1107 (2019), https://doi.org/10.1016/j.matt.2019.10.015.
3. J. Frank, Curr. Opin. Struct. Biol. 22, 778–785 (2012), https://doi.org/10.1016/
   j.sbi.2012.08.001.
4. M. A. Herzik, M. Wu, G. C. Lander, Nat. Comm. 10, 1032 (2019), https://doi.org/10.1038/s41467-019-08991-8.
5. E. Hand, Science 367, 354–358 (2020), https://doi.org/10.1126/science.367.6476.354.
6. S. Liu, T. Gonen, Commun. Biol. 1, 38 (2018), https://doi.org/10.1038/s42003-018-0040-8.
7. C. G. Jones et al., ACS Cent. Sci. 4, 1587–1592 (2018), https://doi.org/10.1021/acscentsci.8b00760.
8. UCLA First Annual MicroED Summit, https://cryoem.ucla.edu/summit.
9. R. Watanabe et al., bioRxiv 1–22 (2019), https://doi.org/10.1101/837203.