What color was Tyrannosaurus rex? Muddy brown with a tawny underbelly? Slate gray with a pinkish hue? Or powder blue with rainbow stripes?

Probably not the latter, but who really knows? Fossils tell us that T. rex roamed the Earth some 65 million years ago. But we're left with little more than our imaginations to fill in the details about the terrible lizard's looks and lifestyle - attributes that were not preserved in the fossil record.

The same goes for the RNA world. The discovery that RNA can catalyze a variety of enzymatic reactions has convinced many scientists that RNA was most likely the first macromolecule that could carry and copy genetic information. But what was the RNA world like? "I think we can never know," says Andrew Ellington of the University of Texas at Austin.

Still, in vitro selection experiments, conducted by Ellington and others, continue to call attention to the catalytic prowess of RNA. At a workshop held October 24-26, 1997, on molecular evolution - sponsored by the Center for Advanced Studies in the Space Life Sciences at the Marine Biological Laboratory in Woods Hole, Massachusetts - researchers presented data that suggest it may be relatively easy to persuade RNAs to perform simple chemistry or bind to selected molecules. Although the results support the theory that an RNA world once thrived, scientists are really no closer to realizing precisely which loops and folds gave ancient RNAs their power.

Further, because RNAs are so flexible, Ellington argues that RNAs could have reinvented the structures they needed to perform particular catalytic feats many times throughout the course of history. As a result, scientists should not view modern RNAs as well-preserved relics that offer a direct window to the past. "If we can't separate historical traits from modern inventions," says Ellington, "we may never know our origins."

At the center of the debate is a question that borders on the philosophical: What can in vitro selection experiments reveal about the RNA world? To determine how early RNAs might have behaved, scientists try to convince modern RNAs to catalyze the reactions that would have been needed to seed and sustain life. The experiments are technically simple: Researchers drop a large collection of random RNA molecules into a test tube and apply a "selective pressure." Those RNA sequences that can perform a particular task are isolated and amplified. By forcing RNAs to evolve - those that fail the selection do not survive to "reproduce" - scientists can see just what RNA can do. And what it might have done in the past.

The answer, it seems, is quite a lot. "The number of new things RNA can do has increased dramatically in the past five years," says Mike Yarus of the University of Colorado at Boulder. RNAs can join oligonucleotides, add nucleotides to a primer, bind amino acids, and catalyze peptide-transfer reactions. "Selection experiments are progressively adding new activities to RNA's catalytic palette, making it more plausible that there was an RNA world," says Yarus. "We're really only limited by the rate at which we can do the selections."

"I think many people are starting to believe that RNA can do anything," says Nancy Maizels of Yale University. "We now have a lot of examples of people asking whether RNA can do something - and it can."

But that doesn't mean that RNA once performed these catalytic functions. "Bluntly, the question is to what extent one confuses selection experiments with real life," says Yale's Alan Weiner. "It's interesting to see what contemporary RNA can do, but that's different from seeing what happened."

Of course, scientists can never recreate the exact conditions that were present on the prebiotic Earth, says Dave Bartel of the Whitehead Institute for Biomedical Research. "Partly because we don't know what those conditions were." And partly because of space limitations. "The early Earth was a much larger test tube than we can use," he says. Another difference: The original experiment took more than a few million years. "We try to rig things so they happen in a reasonable amount of time and in a small volume," says Bartel.

And the RNAs do the rest. "Selection is so pleasant, it's amazing you can get paid for doing it," says Yarus. "You ask the universe a question and it always answers," he says. The catch? "You have to figure out what the results have to do with the question you asked," Yarus responds.

The way Ellington sees it, the fact that he can easily generate a plethora of RNAs with custom-made characteristics indicates that the properties of modern RNAs are not necessarily remnants of the ancient past. Instead, they could be recent evolutionary additions that reveal little about our RNA ancestors.

For example, modern ribosomal RNAs can bind aminoglycosides - antibiotics that inhibit protein translation. Because the skill is shared by many RNAs, some scientists have argued that the aminoglycoside binding site has been conserved from antiquity. But Ellington and his colleagues have found that they can isolate scores of RNA sequences capable of binding aminoglycosides. "We get a bestiary of sequences that don't look a thing like one another," he says. The results - published in Chemistry and Biology in 1995 - lead Ellington to conclude that RNAs could have evolved the ability to bind to aminoglycosides many times throughout history. "If I can generate a million or a billion sequences that can bind aminoglycosides, I'd say there's nothing particularly special about that structure in the ribosome."

The results are not peculiar to aminoglycosides. Ellington and his colleagues find that they can do similar selection experiments and generate a slew of RNAs that can bind to the amino acid arginine. And they can produce RNAs that interact with tRNA synthetases, the enzymes that present amino acids to modern-day tRNAs.

But in vitro selection does not necessarily reflect on evolution, says Weiner. "In these selection experiments, you set up a particular hoop and isolate molecules that can leap through in a single bound," he says. "Some RNA almost always does. The question is: How likely is it that such RNAs arose during the course of evolution? We don't know - we weren't there."

The variety of RNA products may reflect, in part, the simplicity of the selection, say Weiner and Maizels. For example, Ellington selected RNA sequences that could bind to tRNA synthetases. But modern tRNAs also have to interact with ribosomes, amino acids, exonucleases, and other proteins. "How often can I evolve a structure that will interact with any one of those molecules?" asks Weiner. "It can be as easy as falling out of bed."

And deriving a variety of structures shouldn't be a surprise, considering the starting material, says Maizels. "In selection experiments, you may start sequences that represent more potential variation than we have in our genomes."

"You can't write history with selection experiments - and the RNA world has been dead for three-and-a-half billion years," says Yarus. "But as RNA acquires more and more catalytic capabilities, it fleshes out our idea of what RNA creatures could have done." Yarus and his colleagues have used in vitro selection to generate RNAs that bind to a transition-state analog that mimics the structure that amino acids adopt as they form a peptide bond. The resulting RNAs, described in the June 3, 1997 issue of Biochemistry, look something like the peptidyl transferase loop that binds to amino acids in modern 23S ribosomal RNA - a result Maizels finds "tremendously exciting."

"You start out with a random pool of oligonucleotides, and the selection brings you home to the center of the ribosome," she says.

But before RNAs turned to synthesizing proteins, they would have had to figure out how to replicate themselves - a trick that Bartel has gone some distance toward recreating in a test tube. He and his colleagues have used in vitro selection experiments to generate RNAs that use an oligonucleotide template to add nucleotides to an RNA primer, a result they published in the July 25, 1996 issue of Nature. At the October workshop, the researchers presented evidence suggesting that ribozymes may also be able to synthesize one of the nucleotides that form the building blocks of RNA.

What does this tell us about an RNA world? "Dave's work lends credence to the notion that the RNA world really could have existed in a catalytically complex form," says Ellington, whose lab is now using in vitro selection to examine how ribozymes might have evolved a dependence on cofactors, such as adenosine triphosphate (ATP) or DNA, for their catalytic activity.

So RNAs were probably self-replicating, and it appears that ancient RNAs could have handled protein translation. Does that reveal what the RNA world was like? Ellington thinks not. "Don't get me wrong," he says. "I strongly believe in the existence of an RNA world. I'm sure my ancestors were nucleic-acid replicators going all the way back." But Ellington believes we will never be able to recapitulate our history. "We started as pathetic little replicators in slimy little micelles or on the surface of a rock," he says. "But what did the ribozyme that replicated the first genome look like? We have no idea."

Bartel agrees. "The RNA world is gone forever and we'll never be able to prove it existed," he says. "But we are having a lot of fun seeing what RNA can do."