X-Ray Crystallography Without Crystals

X-ray crystallography, a technique that can produce images of molecules with exquisite detail, has one drawback: It works best on crystals, in which many copies of a molecule are lined up in a regular array. But in this week's Nature, scientists report creating a diffraction image from a noncrystalline sample. The new technique could lead to images of all kinds of things that cannot be crystallized--from cells to individual protein molecules.

In x-ray crystallography, x-rays are targeted at a crystal; bouncing off the atoms, the rays produce a set of diffraction spots, which researchers can mathematically reconstruct into an image of the molecule. That calculation requires measuring the x-rays' intensity and determining their phase, the relative timing--a harder problem traditionally solved by comparing the diffraction pattern from a pure crystal with one from a similar crystal in which heavy metal atoms substitute for some components of the crystal. The signals from the metal atoms provide reference points from which the phase of the other x-rays can be worked out.

But with noncrystalline samples, x-rays generate splotchy patterns instead of sharp spots. In these splotches, however, the intensity varies smoothly from one pixel to the next in a manner related to the phase. In the early 1980s, researchers suggested that it might be possible to use that information to work out the phase. Now Jianwei Miao, a physicist at the State University of New York, Stony Brook, and his colleagues have created a computer algorithm to do just that. The program combines the intensity data in the splotchy diffraction pattern with random phase information generated by the computer to churn out an approximate image of the target responsible for the diffraction. It then reconverts the image back into the corresponding diffraction intensity data and phase information. Combining the new phase information with the original intensity data, it generates a new picture--and so on. Repeating the cycle about 1000 times produces a final image.

Using their technique, the team imaged an array of tiny gold dots with a resolution of 75 nanometers--not anywhere near the resolution of top-notch crystalline samples, which can be hundreds of times finer, but already better than the best optical microscopes. "It's really a brilliant experimental achievement," says Eaton Lattman, an x-ray crystallographer at The Johns Hopkins University in Baltimore.