Gerald F. Joyce admits that when he saw the results of the experiment, he was tempted to halt further work and publish the results immediately. After years of trying, he and his student Tracey Lincoln had finally found a couple of short but powerful RNA sequences that when mixed together along with a slurry of simpler RNA building blocks will double in number again and again, expanding 10-fold in a few hours and continuing to replicate as long as they have space and raw material.

But Joyce was not fully satisfied. A professor and dean at the Scripps Research Institute in La Jolla, Calif., the 53-year-old molecular chemist is one of the founding champions of the "RNA world" hypothesis. That is the notion that perhaps life as we know it life based on DNA and enzymatic proteins, with RNA acting for the most part as a mere courier of genetic information evolved out of a simpler, prebiotic chemical system based mostly or even solely on RNA. Of course, the idea is plausible only if RNA can support evolution on its own. Maybe, Joyce thought, his synthetic RNA could help prove that possible. So he and Lincoln spent another year working with the molecules, mutating them and setting up competitions in which only the fittest would survive.

In January, one month before the bicentenary of Charles Darwin's birth, they announced the results in Science. Their little test-tube system did indeed manifest nearly all the essential characteristics of Darwinian evolution. The starting 24 RNA variants reproduced, some faster than others depending on the environmental conditions. Each molecular species competed with the others for the common pool of building blocks. And the reproduction process was imperfect, so new mutants Joyce calls them recombinants soon appeared and even thrived.

"We let it run for 100 hours," Joyce recalls, "during which we saw an overall amplification in the number of replicator molecules by 1023. Pretty soon the original replicator types died out, and the recombinants began to take over the population." None of the recombinants, however, could do something new that is, something that none of its ancestors could perform.

That crucial missing ingredient still separates artificial evolution from true Darwinian evolution. "This is not alive," Joyce emphasizes. "In life, novel function can be invented out of whole cloth. We don't have that. Our goal is to make life in the lab, but to get there we need to increase the complexity of the system so that it can start inventing new function, rather than just optimizing the function we've designed into it."

That goal clearly seems possible, because the RNA replicators in Joyce's lab were relatively simple: each has only two genelike sections that can vary. Each of those "genes" is a short building block of RNA. A replicator, being an RNA enzyme, can gather the two genes and link them together to create an enzyme that is the replicator's "mate." The mate is set free and gathers two loose genes, which it assembles into a clone of the original replicator. Recombinants appear when a mate is unfaithful and links up genes that were never meant for each other. Recombinants did not, however, create genes. It may be possible to engineer a system that does, or to add complexity by giving each replicator more genes with which to work.

Scott K. Silverman, a chemist at the University of Illinois who has done pioneering work with DNA enzymes, hopes that "by capturing Darwinian evolution in new molecules, we might be able to better understand the basic principles of biological evolution," much of which is still somewhat mysterious at the molecular level. Joyce and Lincoln, for example, noticed in their postmortem examination of the experiment that the three most successful recombinants had formed a clique. Whenever any clique member made a reproduction error, the result was one of the other two peers.

The next big step toward the creation of life in the lab, Joyce says, will be to engineer (or evolve) a set of synthetic molecules that can perform metabolism as well as replication. Geneticist Jack W. Szostak of Harvard Medical School has developed nonbiological proteins that bind ATP, an energy-carrying chemical crucial to metabolism. Szostak's lab is also attempting to fashion protocells that encase RNA within tiny spheres of fatty acids, called micelles, that can form, merge and replicate spontaneously.

Even if biochemists do manage to cobble RNA and other basic compounds into some form of synthetic life, the engineered system will probably be so complex at first that it will hardly prove that natural life began in some similar way, four billion years ago. Joyce's replicators consist of a mere 50 chemical letters, but the odds of such a sequence appearing by chance are roughly one in 1030, he notes. "If it were six or even 10 letters long, then I'd say we might be in the realm of plausibility, where one could imagine them assembling spontaneously" in the primordial soup.

From Test-Tube Life to Diagnostic Tools

Creating life in the laboratory would be a momentous occasion for humanity, even if it is more molecular than Frankensteinian. But there may be more mundane uses for such chemistry. A paper in press at Nature Biotechnology, Gerald F. Joyce says, describes how his lab at the Scripps Research Institute in La Jolla, Calif., has modified RNA replicators so that they must perform a biochemical function to reproduce. The winners of that evolutionary race will be good candidates for a medical diagnostic, he thinks. Scott K. Silverman of the University of Illinois says the idea has merit: "Suppose you need to do detection in a dirty environment with lots of different chemicals present say you want to find Salmonella inside peanut butter. That's hard to do without purification steps. It would be useful to be able to evolve the diagnostic system so that it still finds the signal despite all the noise." W.W.G.