a technique known as the polymerase chain reaction (PCR) to make millions of copies of the RNA sequences that bound strongly to the protein.

After four successive rounds of se­lection and "reproduction," Tuerk and Gold were left with just two RNA se­quences that bound strongly. One was identical to a naturally occurring RNA known to bind to the target. The oth­er sequence was unknown, and the re­searchers say it would never have been suspected of binding to the target. If such a sequence had been valuable as a pharmaceutical, the technique would have found it even though no chemist had thought of it.

The other three groups used a differ­ent technique to produce random se­quences. They spliced random nucleic acids into the genes of bacteriophag­es-viruslike agents that infect bacte­ria. The genes give rise to proteins on the outside of the bacteriophage, so many millions of random protein frag­ments were expressed on the phage surfaces.

The workers then allowed the phages to come into contact with immobilized target molecules and washed away the phages that did not bind. The few phag­es that did were then recovered and al-

lowed to reproduce in bacteria. All the groups found that after several rounds of selection and reproduction, only a few of the initial millions of protein se­quences were present.

A group headed by Steven E. Cwirla and William J Dower at Affymax Re­search Institute in Palo Alto, Calif. , used sequences of six amino acids. Writing in the Proceedings of the National Acade­my of Sciences, the group reports that it quickly found several sequences that bound to the chosen target molecule, a monoclonal antibody. All the sequen­ces bore a strong similarity to a portion of a known natural protein (beta en­dorphin) that is bound strongly by the antibody. (The exact natural sequence, though, was not found .)

Jamie K. Scott and George P. Smith of the University of Missouri at Co­lumbia screened about 40 million ran­dom protein sequences six amino acids long for the ability to bind to two monoclonal antibodies. They also quick­ly "evolved " sequences similar to natu­rally occurring ones. In addition, they found an unknown sequence. A team headed by James J Devlin at Cetus Cor­poration used sequences of 15 amino acids selected for the ability to bind to a substrate. Several binding sequences

were quickly found, all containing a particular short stretch of amino acids that was presumably responsible for the binding ability.

Although it is too early to predict how valuable directed evolution will turn out to be, some researchers are al­ready eyeing commercial applications. The Cetus group, which was racing Scott and Smith to publication ( both reports were in the same issue of Sci­ence), has applied for a patent on its technique as a means of drug discov­ery. Gold hopes to adapt his technique, which he has named SELEX, to select nucleic acids directly for the type of proteins they encode. "We're looking at a quantum jump" in the ability to identify biologically active molecules, Dower says.

Aside from new drugs, the technique will teach investigators much about how proteins and nucleic acids behave. Directed-evolution experiments are also likely to fuel speculation about how life evolved, even though the experiments are carried out under unnatural con­ditions. The experiments, Gold says, provide hard data for "people who wonder what DNA and RNA are ca­pable of." And that is just about all biologists. - Tim Beardsley

A Latin American caterpillar calls ants to a free meal

U lterior motives often underlie friendships. That is certainly the case with ants and caterpillars. Cat­erpillars commonly feed ants a secretion rich in

sugar and amino acids. in return, the ants refrain from devouring the caterpillars and even protect them from wasps and other predators.

Entomologists have observed ants and caterpillars throughout the world engaged in this odd alliance. Re­searchers have also suspected that many caterpillars

24 SCIENTIFIC AMERICAN October 1990

somehow "call" the ants to them. But how, exactly? Philip J. DeVries, an entomologist at the University of Texas at Austin, has answered that question-at least in the case of a Latin American caterpillar known to biologists as Thisbe irenea.

The caterpillar, which metamorphoses into a small brown and white butterfly with iridescent blue trim on its lower wings, has a pair of ridged, rod-shaped organs lo­cated just behind its head. The caterpillar "plays" these

tiny organs, which are called pa­pillae, by scratching its bumpy head against them, according to DeVries.

The papillae generate a high­pitched, rhythmic chirping that is too subtle for humans to hear, but ants get the message in vibrations that travel through the ground or leaves. DeVries was able to record the sound by holding a microphone to the ground.

Devries says the papillae re­mind him of a Latin American per­cussion instrument called the guei­ro, a gourd encircled with grooves that one plays by bumping a stick along its length. Next question: Can Thisbe irenea play "La Cuca­racha"7 -Ho/ger Wittekindt

© 1990 SCIENTIFIC AMERICAN, INC