Press "Enter" to skip to content

‘We need a people’s cryo-EM.’ Scientists hope to bring revolutionary microscope to the masses

By Eric Hand

The Laboratory for Molecular Biology (LMB), clad in glass the color of sea ice, rises like a futuristic factory above the rapeseed fields of Cambridge, U.K. It is the crown jewel of the U.K. Medical Research Council, a storied government lab that has garnered more than a dozen Nobel Prizes. One of the first came in 1962 after LMB researchers, having pioneered x-ray crystallography, used the technique to decipher the first atomic structures for proteins—those of myoglobin and hemoglobin, which carry oxygen in muscle tissue and blood. X-ray crystallography has dominated structural biology ever since, but it has an Achilles’ heel: Some proteins just can’t be coaxed to form crystals, which scatter x-rays to reveal structure.

One of LMB’s most recent Nobel laureates is Richard Henderson, an unassuming Scotsman who slouches in his office chair in socks and sandals, surrounded by mountains of paperwork. In 2017, Henderson won a share of the chemistry prize for his work in developing detectors for cryo–electron microscopy (cryo-EM). The technique, also pioneered at LMB, is the brash upstart in structural biology, challenging crystallography in resolution and surpassing it in purview: It opens up far more proteins to inspection and captures many more of their natural configurations.

Cryo-EM dodges the problem of crystallization with a flash-freezing process that fixes proteins in thin films of glassy ice. Then, an electron microscope takes thousands of 2D snapshots of the proteins caught in random orientations. A computer stitches them together to reveal the 3D structure, so important in understanding how a protein works—and how a drugmaker might target it.

But cryo-EM has a big problem of its own: long waits to use extraordinarily expensive microscopes. Moreover, tricky sample preparation means that even when researchers get access, much of their time ends up wasted. “There’s this dirty little secret of the field,” says LMB physicist Chris Russo. “It’s almost as hard as making a crystal. There’s a lot of trial and error in it.”

Some researchers are aiming to fix those problems with automation: robots that can make icy protein samples more reliably and with less waste. Other scientists are developing materials that protect protein molecules during freezing. Still others are working on software that could gather data more efficiently.

But Henderson sees those efforts as mere Band-Aids. For him, what holds back the technique is the forbidding cost of a microscope. Henderson, Russo, and a small group of confederates are trying to make cryo-EM affordable.

A top machine costs about $7 million. Preparing a room and installing a microscope can cost just as much. Then come the operational costs—a torrent of electricity, dedicated troubleshooting staff—that can rise to $10,000 per day.

Roughly 130 Krios machines—the microscopes widely considered the best—have been sold by Thermo Fisher Scientific and installed around the world. LMB has the luxury of three for a relatively small staff, and yet even its researchers must wait a month or more to get time.

Most structural biologists have no access at all. “The wait can be from 3 months to infinity,” says Bridget Carragher, codirector of the Simons Electron Microscopy Center in New York City, a mecca for cryo-EM. “It’s becoming the haves and the have-nots.”

No one complains about the quality of the Krios machines, which take months to assemble by hand from thousands of parts in a Dutch factory. They are Cadillacs. But science would benefit from something less posh, Henderson says. “We need a people’s cryo-EM for maybe 10 times less: a Volkswagen Beetle.”

<?xml version=”1.0″ encoding=”utf-8″?????>1–3 micrometersAltered proteinsNucleusIncident electronIncident electronIonized electronIceAirChemical tetherGrapheneA costly viewCryo–electron microscopy (cryo-EM) reveals the structure of proteins by probing a flash-frozen solution with a beam of electrons, and then combining 2D images of individual molecules into a 3D picture. The cost of high-energy microscopes has limited the method’s adoption, but researchers are pushing for cheaper, smaller, lower-energy machines (right).Electron sourceNOWA field emission gun generates a beam of electrons that is typically accelerated to energies of 300 kiloelectronvolts (keV).FUTUREAt lower energies of 100 keV, a microscope wouldn’t need an expensive spark-snuffing system.Bending the beamNOWHigh-energy machines useheavy magnets to focus the electron beam. Thick lead shielding soaks up stray radiation.FUTUREA low-energy machine wouldbe lighter and wouldn’t require as much shielding.Imaging systemNOWElectron detectors, typically made of silicon, are costly.FUTUREA low-energy machine would use a cheap detector customized for 100-keV electrons.High-energy microscopeLow-energy microscope*Rubber ducks as stand-in for proteinsProtein sample dockSample holder, or grid (3-mm diameter)Deep freeze Protein solutions are flash-frozen in baths of liquid ethane to create a clear thin film of ice. But the process is a crapshoot: Proteins are often drawn to, and altered by, the air-water inter- faces. Layers of graphene, studded with sticky chemical tethers, could keep proteins from straying too close.Close encountersElectrons that are deflected by a protein’s nuclei (below left)instead of colliding with its electrons (below right) are key to building a cryo-EM image.Elastic scatteringA low-energy electron beamyields more of thesedeflected electrons, whichproduce cryo-EM images.Inelastic scatteringElectrons that ricochet offan atom’s own electronscan create ions thatdestroy the proteins.Building a 3D pictureIdeally, individual proteins freeze in random orientations. The microscope generates 2D images of eachorientation. A computer identifies the 2D projections and uses them to calculate the 3D structure.


For decades, microscopists have opted for machines that run at high energies and require expensive parts and precautions. After revisiting the basic physics, Henderson, Russo, and colleagues showed that a cheaper, lower-energy microscope can take pictures that are just as good, if not better.

Now, Henderson hopes to persuade manufacturers to make his cut-rate machine. At less than $1 million, it should put cryo-EM within reach of thousands of labs. Doing so would democratize the field, he says, and accelerate the discovery of protein structures. At the very least, Carragher says, researchers could use a cheap machine to screen out bad samples, preventing wasted time on a Krios. “People need to learn the trade and try things out,” she says. “They shouldn’t be doing it on the glamour machine.”

Cryo-EM’s rise seems unstoppable. At first, researchers worried that the technique could not go small enough: Before 2010, it could attain resolutions below 4 angstroms—four times the diameter of a hydrogen atom—only for a few symmetric, easy-to-solve structures. But machines like the Krios, along with new detectors that record the path of pertinent electrons before the electron beam fries the sample, have changed the game. In the best cases, researchers can now make maps with resolutions below 2 angstroms, putting cryo-EM on par with crystallography.

Cryo-EM also has distinct advantages over the older technique. Consider the study of the cell’s gatekeepers, membrane proteins, which are drugmakers’ most popular targets. The proteins are tough to crystallize because they flex to let things in and out of the cell. Crystallographers must put them in chemical straitjackets to stabilize them and get them to crystallize. But doing that means losing valuable information about how a floppy protein functions, says Melanie Ohi, a structural biologist at the University of Michigan, Ann Arbor.

By contrast, cryo-EM can freeze proteins in any of their shapes, showing how they act. This year, for example, Sriram Subramaniam, a structural biologist at the University of British Columbia, Vancouver, and colleagues used cryo-EM to obtain snapshots of Cas9, the enzymatic scissors of CRISPR gene editing, as it snipped and cut DNA. A protein “is not a single, static thing,” he says. “Cryo-EM tells you the story of how things work.”

Icy tool gets hot

X-ray crystallography still leads cryo–electron microscopy in solving protein structures. But a turning point came in 2012, when the first high-resolution cryo-EM cameras appeared. Cryo-EM could surpass crystallography in several years.
<?xml version=”1.0″ encoding=”utf-8″?????>200020052010201520202025First direct electron detectorsYearRatio of proteins solved by crystallography to proteins solved by cryo-EM0.0010.010.11


X-ray crystallography isn’t going away anytime soon: It is buttressed by the sunk costs of billion-dollar synchrotrons, government-owned facilities that supply bright x-ray beams for crystallographers and other users. But trends at the Protein Data Bank, a repository of protein structures, reflect cryo-EM’s growing popularity (see chart). The number of proteins solved by x-rays peaked in 2017, whereas the number of cryo-EM solutions is nearly doubling every year. Cryo-EM will soon surpass x-rays, predicts Jim Naismith, director of the Rosalind Franklin Institute, a biological imaging center that will open in 2021 in Didcot, U.K., and will task many of its 200 researchers with improving cryo-EM methods. “In 5 years or less, EM will be the dominant method,” he says. At LMB, it already is: The lab crossed that threshold in 2016.

To alleviate some of the pent-up demand for access, the U.S. National Institutes of Health (NIH) in 2018 announced $130 million in grants to establish cryo-EM centers in New York, California, and Oregon, each with a few Krios machines. As with synchrotrons, time on the machines at those centers is freely available to researchers. Users simply arrive with their frozen samples—or ship them—and cross their fingers for good data.

Yet many leave empty-handed, says Claudia López, co-director of the cryo-EM center at Oregon Health & Science University (OHSU). “You can have the best microscope in the world, but if your sample is no good, there’s nothing you can do about it,” she says.

It’s not just the chanciness of sample preparation; it’s also the lack of access that frustrates Claudio Grosman, a biophysicist at the University of Illinois, Urbana-Champaign, who studies a membrane protein in muscle cells and neurons that binds to nicotine and other drugs. He is waiting to see whether he will get time at the Simons Center, which has added four NIH-funded Krios machines to three existing ones. “I am used to being limited by my own skills in my lab, by my own capacity to read and understand the literature,” he says. “It’s hard to compete with these labs that have a cryo-EM machine in their basement.”

Even with seven machines, Carragher says queues at the Simons Center are long and fretful. “Everyone wants more time,” she says, “everybody thinks they’re not getting enough.”

It’s easy to see why. At LMB in July 2019, postdoc Byung-Gil Lee finally had time on a Krios after weeks of waiting—and weeks of purifying a protein until hundreds of trillions of copies were concentrated in a single drop. The object of his desire: cohesin, a protein involved in cell division that is defective in many cancers. Only bits of its structure have been solved with crystallography, and Lee had just 24 hours to try for a better picture using cryo-EM.

With a pipette, he beaded the solution onto 10 sample holders, each the size of a flea. Then he froze them in a flash, as a robot plunged them into a bath of liquid ethane. He held a hair dryer up to a dentist’s palette of tiny tools and blasted away residual water, which could contaminate the samples. With tweezers, he slotted the sample holders into a rack and doused it in hissing liquid nitrogen to keep the samples frozen. He loaded the rack into the humming Krios towering behind him. An unseen mechanism slipped it into a bright beam of electrons accelerated to three-quarters the speed of light.

Swiveling to a computer command station, Lee fiddled with knobs and a joystick to zoom in. Splotches and blank spots on the sample holder, or grid, indicated that the thin films of ice didn’t completely form. “I don’t think this grid has good ice,” he says. “This is common.”

Lee ultimately found that just three of his 10 sample holders were worth inspecting. He didn’t get enough snapshots for the computer to build up a 3D picture of the protein. He had to get back in the queue and wait for more microscope time.

Byung-Gil Lee hopes cryo–electron microscopy will give him a close-up view of cohesin, a protein involved in cell division.


How best to tackle such problems? One remedy is to be thriftier with samples of proteins, which are time-consuming and expensive to make and then casually wasted. Typically, researchers use pipettes to apply microliters of the protein solution to sample holders and blot away the excess with filter paper. In 2016, Carragher and colleagues reported developing a dispenser that sprays the solution like an inkjet printer, releasing picoliters instead of microliters—only one-millionth as much.

Naismith wants to look further upstream and reduce the amount of protein solution made in the first place. Cryo-EM needs fewer copies of a protein molecule than x-ray crystallography, and yet purifying techniques haven’t changed significantly in 20 years, he says. Once the Rosalind Franklin Institute opens its doors, he says, its researchers will work on techniques that could supplant chromatographic columns—the tall tubes that separate and purify proteins.

Sample preparation has a more fundamental problem: The air-water interfaces at the top and bottom of the thin films are perilous for proteins. Drifting proteins that happen to encounter an edge in the moments before flash-freezing—which is likely given the films’ thinness—tend to stick to the surface. “This peels them open and really destroys them,” Russo says.

Russo and colleagues found a way to protect them, by undergirding the thin film with graphene, a one-atom-thick layer of pure carbon that is transparent to the electron beam. The researchers patterned the graphene with “functional groups”—chemical studs such as carboxyl or amine groups. Those groups jut into the solution and stick to passing proteins, preventing them from wandering into the air-water interface above. “Instead of having two dangerous surfaces, we’re replacing one of them with one we can control,” Russo says.

Other researchers are trying to wring efficiencies at the back end of the process. New K3 detectors released in 2017 by the company Gatan are not only accurate, but fast. They can vacuum up 6000 pictures in 1 day, several times the rate of a few years ago, Carragher says, allowing more proteins to be tested.

But Henderson says the major bottleneck is the cost of the machines, which keeps them scarce.

A lack of competition helps explain the cost. Hitachi and JEOL make cryo-EM microscopes, but they have not put a dent in Thermo Fisher’s commanding market share. Gatan has a similar hold on the market for detectors, which are sometimes sold separately from the microscopes at roughly $1 million a pop.

The monopolistic pressures nearly got worse. In June 2018, Thermo Fisher made a $925 million bid to buy Gatan and form a cryo-EM superpower. But in April 2019, the U.K. Competition and Markets Authority found that such a merger would reduce competition and lead to even higher prices. By June 2019, Thermo Fisher had scrapped the deal. “We need competition,” Carragher says. “Unless somebody’s chasing you, why would you innovate?”

Steve Reyntjens, Thermo Fisher’s director of product marketing for cryo-EM, says the company is not resting on its laurels. Scientists’ productivity per Krios has risen as the company has improved the microscope’s detectors and data collection software, he says. “We have a track record of innovating.”

Richard Henderson wants to enable more researchers to use cryo–electron microscopy, a technique he helped develop and shared a Nobel Prize for.


And a Krios is expensive for legitimate reasons, including the gear that accelerates electrons to energies of 300 kiloelectronvolts (keV). Operating at voltages of 300 kilovolts, more than 2500 times stronger than electricity from a U.S. wall socket, the machine requires a bulky transformer and thick, heavily insulated cabling. Massive, costly magnets are needed to focus the high-energy electrons into beams. “Everything needs to be tightly controlled and minutely aligned,” Reyntjens says.

The high energies also require expensive safety measures. As the powerful electric fields accelerate the electrons, arcs—little lightning bolts—can form. To prevent arcing, the field emission guns that produce the electrons are suffused in sulfur hexafluoride (SF6) gas, which snuffs out sparks. But the gas destroys ozone, makes toxic byproducts, and must be recycled. X-rays generated when stray 300-keV electrons hit metal are another hazard, requiring several centimeters of lead shielding around the microscope.

Krios machines are too tall for the average ceiling heights in most countries, and labs have to build special footings to handle the load, between 1000 and 2000 kilograms. López says OHSU spent $20 million to renovate space for its new microscopes.

Yet researchers widely believe the high energies are indispensable. The goal is to send electrons close enough to atomic nuclei in a protein to feel an attractive electrostatic force. Their paths are bent, or scattered elastically—meaning they lose no energy. The pattern of those elastically scattered electrons can be used to build up a picture.

Low-energy electrons are prone to getting too close. They collide either with the nucleus or, more often, with one of the protein’s many electrons, scattering inelastically. That causes two problems: The electrons deposit energy into the specimen, eventually destroying it, and they ricochet at lower speeds and odd angles that don’t contribute to a picture. At high energies, electrons zip through faster, with less time to “feel” the electrostatic forces, resulting in fewer of the good elastic scattering events.

In a trade-off between gathering information and avoiding specimen damage, 300 keV is seen as cryo-EM’s sweet spot. But no one had actually tested that assumption. “Amazing, huh?” Russo muses.

When Henderson, Russo, and a colleague did so, they found a surprise: At 100 keV, they did indeed get more bad events, but also many more good ones—enough to make the trade-off worthwhile. For thin specimens, Henderson says, 100-keV microscopes should actually be better.

To prove the point, the LMB team cobbled together a 100-keV machine from spare parts and, in October 2019, published results in the crystallography journal IUCrJ showing they could resolve atoms on five well-known proteins. Such a microscope doesn’t need thick lead shielding or an expensive SF6 system, and the scope can be much smaller and lighter. “It’s actually a huge difference,” Naismith says.

The team hopes now to entice a company to build a 100-keV machine. At an August 2019 microscopy meeting in Portland, Oregon, Henderson and Russo met privately with Thermo Fisher’s CEO and officials from other manufacturers to coax them into making something new and selling it for less. “They listen politely,” Henderson says, but so far “we’re on our own.”

Reyntjens points to Glacios, a 200-keV cryo-EM microscope that Thermo Fisher sells for half as much as a Krios, and says, “Our long-term strategy is to support the democratization of cryo-EM.”

Henderson set about finding smaller companies to take up the 100-keV mantle. He persuaded a U.K. company, York Probe Sources, to build a cheap field emission gun. “I started from a clean slate and designed it for 100 kilovolts,” says founder and electronics engineer Mohamed El Gomati.

Henderson also enlisted the help of Swiss company Dectris, which says it can build detectors for less than £150,000. Sacha De Carlo, Dectris’s EM business development manager, expects the campaign by Henderson and colleagues will create demand for 100-keV detectors. “He’s extremely influential,” De Carlo says.

And next month, JEOL will deliver to LMB the first of three cheap 100-keV microscopes—a basic model to which Henderson plans to add the field emission gun and detector.

If he can wow fellow scientists with good results—and show that the price tag is well under $1 million—he’s confident that Thermo Fisher will snap to attention. “Either through shame or blackmail or bribery, we’ll get them to do it in the end,” Henderson jokes.

Henderson already knows where he will put 100-keV machines. At LMB, he has been given space in a room formerly devoted to x-ray crystallography—a metaphor not lost on him. The small cryo-EM machines will fit easily in the room, with no need for renovations. One of the in-house x-ray sources has already been cleared away. A few technicians work on a declining library of crystals while Henderson squints to visualize the tools he’s trying to muscle into existence. “X-rays are on their way out, really,” he says. Cryo-EM—cheap cryo-EM—is on its way in.

*Correction, 23 January, 12:45 p.m.: An earlier version of this story incorrectly stated that Richard Henderson was LMB’s most recent Nobel laureate.

Source: Science Mag