Do the Shuffle
Stemmer invented his version of directed evolution, called “molecular breeding,” seven years ago, while working at Affymax, in Palo Alto, Calif. The brainchild of entrepreneur Alejandro Zaffaroni, Affymax was the first company to focus exclusively on discovering new drugs through combinatorial chemistry-a shotgun approach in which huge libraries of unique molecules are randomly generated, and then the useful ones are fished out through clever screening. Extending the combinatorial idea into the realm of protein development, Stemmer struck upon the idea of shuffling DNA. The basic concept: Start with a few different versions of the gene for a protein you’d like to improve, cut them up and mix the pieces together to generate a diverse pool of new versions of the gene, fish out the ones you like best, and start all over again. In 1997, Zaffaroni spun off Maxygen to take molecular breeding to market.
Larger companies, sniffing the commercial potential of evolution in a test tube, began hooking up with Maxygen. One of the first was the Danish firm Novo Nordisk, the largest industrial enzyme maker in the world, which had already been dabbling in directed evolution on its own. Reporting last year in the journal Nature Biotechnology, scientists from the two companies showed how directed evolution might churn out a new laundry detergent enzyme. The researchers began with the structural gene for Savinase, a stain-eating enzyme developed by Novo Nordisk. Next they collected DNA for the naturally occurring version of the enzyme, subtilisin, from 25 different strains of Bacillus bacteria. With scissors-like enzymes, the team chopped up Savinase and wild subtilisin DNA, shuffling it together to create a new generation of unique “daughter” genes. Then the researchers inserted each gene into a bacterial cell. Finally, they exposed those cells to different temperatures and pHs to see how the resulting proteins held up.
The results were intriguing. While Savinase works best in a limited range of conditions-cool water and a fairly alkaline environment-some of the new proteins produced by the experiment worked four times better, and under acidic conditions. Many of the daughter molecules also performed better than their parents when heated or dunked in organic solvents. For 35 years, Stemmer notes, teams of industry scientists have tried to rationally design an improved subtilisin enzyme. “Within a year, three of our scientists turned out a much better molecule,” he says.