To catch a glimpse of evolution, Pim Stemmer doesn't have to set sail for the Galapagos. He needn't trudge through the Costa Rican rain forest or blast into Antarctic sea ice. For a front-row view of evolution in action, Stemmer simply walks over to a lab bench at Maxygen, a biotech startup in Redwood City, Calif., where he oversees research and development. At Maxygen, researchers are tapping evolutionary principles to "breed" new proteins, successively fine-tuning specific traits-say, heat tolerance or the ability to stick to a cancer cell-by tinkering with the underlying genetic code and selecting only the best of the new genes for the next round of improvement. In farm fields and kennels, exploiting evolution is old hat. Farmers favor a corn crop that best survives summer, for instance, and with decades of patience, breeders can produce dogs as specialized as poodles and Pomeranians. But in the lab, evolving better proteins is a recent feat-and a faster one. By focusing their efforts at the genetic level, one molecule at a time, researchers can coax a protein toward perfection in just a matter of weeks. In the process, companies like Maxygen hope to crank out safer medications, more potent cleansers and healthier crops. Just a few years out of the gate, these young firms are already bringing new protein products to market, forging alliances with such giants as Dow Chemical, DuPont and Novartis and launching successful IPOs (see table, "Evolving Industry")-all evidence that "directed evolution" is really happening, not just in the test tube, but also in the marketplace.

Evolutionary Alternative

Directed evolution is an alternative to "rational protein design, " a technique that became popular in the 1980s. In rational protein design, researchers try to craft a new molecule-perhaps an antibody or an enzyme-by first studying an existing protein's structure and then modifying it via targeted mutations to the gene that encodes it. But that sort of painstaking methodology can prove to be difficult. Not only must researchers determine the sequence of amino acids-the 20 building blocks that make up all proteins-but they must also understand the complicated pattern of folding that the chain of amino acids undergoes to become a functional, three-dimensional molecule. Even after bringing sophisticated computer tools to bear on the problem, researchers say, it's hard to unravel the workings of a protein folded up like origami, much less create a new one that behaves the way you want it to. "Rational design has failed miserably at helping us make useful proteins," says Caltech's Frances Arnold, who sits on Maxygen's scientific advisory board.

Frustrated with the limitations of rational design, a growing group of researchers are borrowing from nature's tool kit instead, mimicking the basic processes of evolution-the generation of genetic diversity and the selection of desirable traits-to improve proteins, even without understanding their complicated structures. And turning to nature can pay off quickly. A team led by MIT chemical engineer Dane Wittrup, for example, evolved an antibody fragment to bind 10,000 times more tightly to its target in just four rounds of directed evolution. Researchers like Wittrup hope better-binding antibodies might someday be used to fight cancer: Attach a cell-killing agent to an antibody that binds specifically and tightly to molecules found only on cancer cells, the logic goes, and you can wipe out the cancer without damaging healthy tissues. And it's not just antibodies but a wide range of potentially useful proteins-from industrial enzymes to protein pharmaceuticals-that stand to benefit. "With just a few evolutionary cycles, we can dramatically improve molecules that have frustrated companies for years," says Stemmer.