Published online 5 April 2001 | Nature | doi:10.1038/news010404-15


Found in sequence space

If you test enough random proteins, you'll find one to do what you want.

DNA: stick or twist.DNA: stick or twist.

If you wanted to make a working protein, but didn't know where to start, how many rolls of the biochemical dice would it take to get lucky? One hundred billion, say Anthony Keefe and Jack Szostak, of the Massachusetts General Hospital in Boston, who've tried it to hunt out proteins to do a predetermined job from a vast number of random genes1.

These sort of odds make buying a lottery ticket seem like a sound investment. They suggest, says Ronald Breaker, a molecular biologist at Yale University, New Haven, Connecticut, that you'd have to strain a sizeable quantity of primordial soup before you found something that evolution could get its teeth into.

But for human beings who want to design a protein to do a job, 1011 to 1 isn't such a bad bet. "These experiments show that finding functional molecules is not all that hard," says Peter Stadler, a theoretical chemist at the University of Vienna. "You can fit that number of proteins in a test tube."

Keefe and Szostak began with a library of DNA sequences corresponding to 6 million million different proteins, all 80 amino acids long. Amino acids are the molecular building blocks from which proteins are made.

The information in DNA sequences is translated into proteins via an intermediate molecule, like DNA, called RNA. Using a nifty technique of their own invention (called messenger RNA display), the duo made molecules consisting of the RNA corresponding to each protein attached to the protein itself. They then tested how well these random proteins stuck to ATP, the molecule that stores energy in cells.

The sequences for the stickiest proteins got copied, the losers were washed away, and the whole process was repeated. Eight rounds later, the descendants of four of the original molecules had come to dominate the population. These all bound ATP in different ways -- ways unlike those found in nature.

So it seems that life as we know it is not inevitable or exclusive. Rather, it is a subset of a larger universe of biochemical possibilities. Stadler's theoretical explorations of 'sequence space' bear this out. "You find spots of molecules that do the same thing, but have no sequence similarity whatsoever," he says.

It's 20 years since Eric Drexler, one of the prophets of nanotechnology, suggested that proteins could be engineered, and that molecular machines could be used in computing or medicine2. But protein design has proved damnably difficult, because of our inability to predict how a linear sequence of amino acids will fold up into a three-dimensional protein. An evolutionary approach might sidestep this problem.

"[Messenger RNA display] really removes the last barriers that prevented access to the untapped potential of protein sequence space," says Breaker. "The future is very clear: use this test-tube evolution process to make proteins that bind everything of interest. These could find use in the clinic and in biotechnology." 

  • References

    1. Keefe,A. D.& Szostak, J. W. Functional proteins from a random-sequence library. Nature 410,715- 7182001. | Article | PubMed | ISI | ChemPort |
    2. Drexler,E. Molecular engineering: an approach to the development of general capabilities for molecular manipulation. Proceedings of the National Academy of Sciences USA 78,5275- 52781981.