MADE TO ORDER Hellinga's computational methods produce huge libraries of mutant proteins, including one that binds well to TNT (top). DUKE UNIVERSITY PHOTOGRAPHY

The feat represents a big step forward for both directed evolution--in which protein functions are altered by creating and selecting numerous mutations--and for computational chemistry's ability to accurately describe the complex interactions and functions of proteins.

The computer removes the limitations of time and search space in a lab, allowing chemists to rapidly examine enormous numbers of possible mutants. The potential applications of these new design techniques range from sensors to chiral separations to enzymes.

Associate biochemistry professor Homme W. Hellinga and colleagues at Duke chose as their starting point the superfamily of medium-sized periplasmic binding proteins in E. coli. The proteins have hinged structures that clamp down onto a ligand--a sugar or amino acid. The receptors in new versions of the proteins are engineered to bind to a series of entirely different ligands, including the explosive trinitrotoluene (TNT), the neurotransmitter serotonin, and L-lactate [Nature, 423, 185 (2003)].

Hellinga's work, says computational chemist William A. Goddard III at California Institute of Technology, represents a "breakthrough in computational design of receptors."

Stephen L. Mayo, associate professor of biology and chemistry at Caltech and a pioneer in computational protein design, says the work illustrates the field's evolution. "This clearly shows now that you can make the transition from doing just structure design to doing function design."

The periplasmic binding protein superfamily, with its many natural variants, was a promising place to start. "Nature is saying this is a really nice motif to evolve," Hellinga says.

Hellinga's group, including grad students Loren L. Looger and Mary A. Dwyer and postdoc James J. Smith, created new and augmented well-known computational strategies, docking different ligands and simultaneously evolving the protein to fit them. The result was vast libraries of 10⁵³ to 10⁷⁶ possible choices. They whittled their selection down to just 17 compounds with promising binding characteristics. Amazingly, the calculations required only modest resources--a computer cluster of 30 processors.

Then, in the lab, the team created those 17 proteins, which performed as they'd hoped. Drawing on previous work, the group inserted a tag on the protein that fluoresces when it binds, as a marker of the mutants' abilities. Additionally, the mutants retained the natural proteins' ability to drive signal transduction pathways.

Ultimately, chemists seek the ability to design enzymatic catalysts, notes William F. DeGrado, professor of biochemistry and biophysics at the University of Pennsylvania School of Medicine, who wrote a commentary accompanying the article. Hellinga's group hopes to do just that.

Chemical & Engineering News
Copyright © 2003 American Chemical Society