Scientists begin cryo-EM with two-dimensional electron micrographs of a subject—in this case an intact rotavirus particle at many different angles. Though the images of individual structures, like the eight or so pictured here, are rather blurry, computer algorithms can average thousands of them and achieve a clear, highly detailed three-dimensional picture.

Credit: James Chen, Ethan Settembre, Scott Aoki, Xing Zhang, Richard Bellamy, Philip Dormitzer, Stephen Harrison, and Nikolaus Grigorieff.

This three-dimensional cryo-EM image of a rotavirus is known as a density map. Each color represents a different protein. Two proteins key to infection can be seen on the temporary, outermost coat of the triple-coated virus particle. The VP4 spike protein, which helps the virus puncture the membranes of target cells, is shown in red. The VP7 protein, which holds VP4 in check until the virus is primed for infection, is gold. In green is a third protein, VP6, found in the middle coat.

Credit: James Chen, Ethan Settembre, Scott Aoki, Xing Zhang, Richard Bellamy, Philip Dormitzer, Stephen Harrison, and Nikolaus Grigorieff.

The density map is detailed enough to allow researchers to zoom in on any given element. This close-up shows a VP7 protein at near-atomic resolution. In the center, VP7 is shown as a yellow backbone enclosed by a blue net representing a surrounding electron cloud. To each side, yellow ribbon-like elements show the locations of nearby VP7 proteins, represented at a lower resolution. VP7 is a trimer, meaning it is composed of three identical molecules. It locks the spike protein VP4 in place until changing conditions compel it to alter its shape, freeing VP4 to puncture a target host cell.

Credit: James Chen, Ethan Settembre, Scott Aoki, Xing Zhang, Richard Bellamy, Philip Dormitzer, Stephen Harrison, and Nikolaus Grigorieff.

It's not easy to prepare a sample in this way, but the bigger challenge for the cryo-EM community, says Grigorieff, has been getting to an atomic-scale resolution of one to five angstroms (the average atom is roughly two angstroms).

“For a long time, people were getting around 20 angstroms resolution, where they might be able to distinguish different proteins but not different amino acids within the proteins,” he says. “Then it was six angstroms or so, which still isn't good enough to show atomic-level detail.”

Grigorieff is intent on pushing that limit. In the rotavirus study, his group achieved a resolution of about four angstroms.

There are two primary ways to get three-dimensional structures from cryo-EM images. Grigorieff uses an approach called single particle reconstruction, taking two-dimensional images of thousands of copies of the virus particle and then averaging them with computer algorithms into a three-dimensional picture. In the other approach, called computed tomography cryo-EM, a single molecular structure is photographed in a series of pictures from many angles to arrive at a similar three-dimensional average, much like the CT-scan technology used by physicians. In both cases, the raw images appear fuzzy, so the averaging step is key.

“Basically, you have a bunch of very snowy images, and then you stack them up on top of each other and out comes a pattern,” says Grigorieff. “When we do enough of these—100,000 or even a million—then we are suddenly able to see much more detail in the underlying pattern.” Or rather, computers see the pattern and interpret it, creating a color-mapped model that shows the orientation of molecules and the ways they interact with each other.

Grigorieff spends much of his time creating and refining those algorithms, which must be powerful enough to analyze all those thousands of pictures—and flexible enough to interpret a variety of structures. It helps, Grigorieff says, to work with a highly symmetrical particle like rotavirus, which has the same symmetry as the 20-faced solid called the icosahedron. “The symmetry makes the task of averaging one million molecules much easier,” he explains.

“The rotavirus work showed that we can get sufficient resolution to build atomic models,” he says. “Because we now have proof of concept... we have more confidence that we will eventually be able to get similar resolution with more difficult samples.” Eventually, Grigorieff would like to tackle cellular behemoths such as the spliceosome, a structure made of 150 or so proteins that edits messenger RNA into a more readable form.

	PAGE 2 of 2	Go Back