Working together. Combining NMR data with computer modeling allows researchers to solve NMR structures of larger proteins.

S. Raman et al., Science

Teamwork Tackles Larger Protein Structures

A protein's shape determines its function, so this molecular geometry is of extreme interest to scientists and drug companies. But it's no easy task to deduce a protein's atomic structure. Now a team has found a way to double the size of a protein that a common technique, called nuclear magnetic resonance (NMR), can decipher.

For researchers looking to solve protein structures, the most popular technique remains x-ray crystallography, which bombards crystals of a protein with high-intensity x-rays to reveal their atomic arrangement. Nevertheless, NMR still has two big advantages: It works on proteins that don't crystallize, and it can reveal a protein's dynamic nature rather than give a mere static picture.

In a nutshell, NMR works as follows: A solution of a molecule is exposed to a magnetic field. The atomic nuclei (especially hydrogen nuclei) behave like gyroscopes. The precise rate at which a nucleus turns depends on the exact magnetic field at its position, which in turn depends on the relative positions and chemical identities of its neighbors. So from the myriad wobblings, researchers can in principle work out the relative locations of all the nuclei. Given the complexity of analysis, however, NMR has worked best for smaller proteins.

NMR researchers typically collect three kinds of experimental data: an easily acquired set, known as chemical shifts, which depends on what kind of atom the nearest neighbor is; a second set known as residual dipolar coupling (RDC), which is influenced by the orientation of various chemical bonds, and a third method called the Nuclear Overhauser Effect (NOE), which helps to interpret some of the chemical shift data. Interpreting all these data is quite involved.

In 2008, David Baker, a biochemist and protein modeling expert at the University of Washington, Seattle, made a breakthrough in simplifying the task. He and his colleagues did away with the need for the NOE and RDC data by combing chemical shift data into their protein-modeling software called ROSETTA. That allowed them to consistently determine the correct shape of moderate-sized proteins--containing up to 100 amino acids--a feat that had been accomplished only in a hit-or-miss way with early protein-modeling software.

For their current study, Baker and colleagues wanted to see if they could extend their hybrid approach to even larger proteins. To do so, they added back a small amount of the NOE and RDC data, but just in the parts of the analysis that deduce the shape of the general backbone of the protein rather then the position of all the individual atoms in each amino acid side chain. They then relied on their ROSETTA software to sort out the position of the atoms in the protein's side chains. The team reports online in Science Express that their latest hybrid succeeded in working out the structure of proteins up to 200 amino acids long.

“It's a very important development and shows where we are likely to be going as a field,” says Ad Bax, an NMR expert at the National Institute of Diabetes and Digestive and Kidney Diseases in Bethesda, Maryland. Proteins of that size have been solved by NMR before, but only with an amount of time and effort beyond most NMR teams. It's also a size of many biologically important proteins, such as proteins wedged into cell membranes, which are difficult to crystallize. So Bax notes that as the new hybrid approach matures, it may prove useful in mapping some of these vital proteins for the first time.

Related Site: