In 1995, Richard Henderson made a bold prediction: using electron cryomicroscopy (cryo-EM), it should be possible under ideal conditions to calculate a 3 Å structure of a protein from just 10,000 molecular images (1).
At the time, cryo-EM couldn’t achieve such high resolution; it would take another decade to get anywhere near 3 Å resolution, and even then orders of magnitude more molecules would be needed.
But Henderson's vision may soon come true. Two papers published this year have raised the bar on cryo-EM by employing a new class of detection device. The first study, directed by Sjors Scheres of the MRC Laboratory of Molecular Biology in Cambridge, UK, determined the structure of the Saccharomyces cerevisiae 80S ribosome to 4.5 Å (2); the second, led by Yifan Cheng and David Agard at the University of California, San Francisco, solved the Thermoplasma acidophilium 20S proteasome to 3.3 Å (3).
In single-particle cryo-EM, particles are photographed as they lie, meaning each has a different orientation and conformation. To compute a structure, a computer must consider up to hundreds of thousands of particles, overlaying them one by one to produce a composite three-dimensional image.
In theory, it should be possible to get near atomic resolution images with cryo-EM. But for much of its history, the technique has essentially been an exercise in “blob-ology,” said Fred Sigworth, Professor of Cellular and Molecular Physiology at the Yale University School of Medicine. “You get a structure to 10–15 Å, and your protein is a blob.”
In part, that’s due to two effects. Irradiating a sample with a high-energy electron beam causes proteins to move, thereby blurring the image, while at the same time incinerating the proteins and degrading the sample. But according to Sigworth, “keeping the [energy] dose low means the images are always horribly grainy, low-signal-to-noise-ratio pictures.”
The other problem lies in the detector used to capture images. Traditionally, the best detector for cryo-EM has been silver-halide film. Film is impractical for collecting thousands of images, so researchers most recently have used CCD digital cameras. However, cameras image light, not electrons, and the conversion of images to electron scale comes at the expense of resolution.
The two new studies address these problems by using a new class of detector that detects electron impacts directly, providing sharper images and higher signal-to-noise ratios. The authors couple those detectors to image processing algorithms that correct the beam-induced motion in the protein, much as astronomers do when imaging stars over long periods.
Cheng and his team imaged their proteasome sample using a direct detection camera called the Gatan K2 Summit, which can directly count electrons and also break a single exposure into multiple short “frames.” The team collected 24 frames for every image, each representing 0.2 seconds.
The composite images from the camera were incredibly sharp compared to traditional CCD images, thanks to improved signal-to-noise ratios and high resolution afforded by direct counting. But not surprisingly, individual objects moved from one frame to the next. So the team applied some motion correction tricks, tracking features in the specimen from frame to frame and applying appropriate shifts to the images. They also removed the first two frames, in which objects tended to move the most, and the final few frames, when radiation damage to the sample was greatest.
From about 120,000 particles, each with 14-fold symmetry, Cheng’s team resolved the 700 kDa, 20S proteasome to 3.3 Å, on par with the previously solved crystal structure (3.4 Å) and far better than the previous best cryo-EM attempt, 5.6 Å.
Until now, cryo-EM researchers could only achieve that level of resolution when imaging highly symmetric viruses, which are up to 100 times larger than the proteasome. According to Cheng, his team has been able to push the size down even further, to 300 kDa or so, making the technique amenable to, for instance, membrane proteins that are traditionally difficult to study with x-ray crystallography. Furthermore, Cheng and his team showed more recently that they could solve the proteasome structure down to 3.6 Å using just 10,000 particles (4).
Scheres and his team used an FEI Falcon detector to image the yeast ribosome to about 4 Å. As an asymmetrical particle, the ribosome is a tougher nut to crack for cryo-EM. The group performed their analysis using just 30,000 particles, a feat that is relatively close to Henderson’s prediction 2 decades earlier. “Suddenly we’re very close to the theoretical limit,” said Sigworth.
That said, don’t expect direct detection-based cryo-EM to completely replace x-ray crystallography. Crystallography produces sharper images, and works well for relatively small monomeric proteins that currently are not suitable for cryo-EM.
Plus, because proteins don’t form crystals in cryo-EM, they can assume multiple conformational shapes, complicating image analysis. “We don’t yet know if this is going to work with any random protein,” Sigworth said.
 R. Henderson, “The potential and limitations of neutrons, electrons and X-rays for atomic resolution microscopy of unstained biological molecules,” Quart Rev Biophys, 28:171–93, 1995.
 X.-C. Bai et al., “Ribosome structures to near-atomic resolution from thirty thousand cryo-EM particles,” eLife, 2:e00461, Feb. 19, 2013.
 X. Li et al., “Electron counting and beam-induced motion correction enable near-atomic-resolution single-particle cryo-EM,” Nat Methods, 10:584–90, June 2013. [ePub May 5, 2013]
 X. Li et al., “Influence of electron dose rate on electron counting images recorded with the K2 camera,” J Struct Biol, 2013. DOI: 10.1016/j.jsb.2013.08.005