The central dogma of modern biology holds that genetic information is inherited in the form of DNA, copied into RNA and expressed as protein; pride of place goes to DNA. But the spectacular discovery that a species of plant can summon up genes its parents have lost highlights biologists' increasing recognition of RNA as a more versatile and important molecule in its own right.

RNA already has a special place among biological molecules. It can store genetic information, as DNA does, but it can also adopt complex three-dimensional shapes and catalyze chemical reactions on itself, as proteins do. "RNA is DNA on steroids," says Robert Reenan, a geneticist at the University of Connecticut. "It can do just about anything." Life probably began as an "RNA world," in which concatenations of RNA molecules pulled double duty as genetic template and reproductive machinery.

The mustard plant Arabidopsis thaliana may be revealing another way in which life exploits RNA's capacity for genetic storage. Susan J. Lolle and Robert E. Pruitt of Purdue University study Arabidopsis whose petals are fused. Such plants have two mutant copies of a gene called hothead, which differ from the normal gene by a single base pair. Strangely, in a few percent of the offspring of Lolle and Pruitt's mutants, one copy of hothead spontaneously reverted to the normal version, repairing its point mutation. Even one such event is statistically unlikely outside of rapidly reproducing bacterial colonies. The investigators systematically ruled out mundane explanations, such as cross-pollination of a mutant plant by a normal one, an extremely high mutation rate or the presence of another, hidden copy of hothead.

Hothead mutants contained changes in other parts of their DNA, too, all of which matched the sequences of the plants' grandparents or great-grandparents but not their parents. This match suggested that a backup copy of the ancestral plants' genome was somehow being passed down, the researchers reported in the March 24 Nature. If true, such leapfrogging would circumvent the normal rules of genetics established by Gregor Mendel in 1865. Because the investigators could find no DNA to play the role, they have proposed that the backup template is double-stranded RNA (which ordinarily has just a single strand). "Double-stranded RNA is hot because that's what's needed for RNA interference," a common way of deactivating genes, says Richard Jorgensen, a plant scientist at the University of Arizona, "but there's no reason it couldn't be a DNA molecule either, and there's no reason it has to be double-stranded."

RNA would be a convenient mechanism, however, because researchers have uncovered several ways in which it modifies the expression or structure of DNA, and it might explain the mysterious production of RNA molecules that do not result in proteins. Several species, including Arabidopsis, rice, mice and humans, copy a surprising amount of RNA from the "wrong" DNA strand--that is, the strand opposite the one that specifies a protein. "Maybe this is where some of that template is coming from," says Joseph Ecker, a plant biologist at the Salk Institute for Biological Studies in La Jolla, Calif. Plants have many enzymes capable of duplicating RNA, Ecker notes, as well as a system for transporting the chemical between cells.

The Purdue group speculates that a separate genetic archive may serve as a hedge against hard times, such as an extended drought, by allowing a plant to access genes that helped its ancestors persist. In this sense, it would bear some resemblance to another strange property of RNA, called recoding.

A next step is determining how widespread the effects are. Unexplained cases of spontaneous reversions also appear in human genetic diseases, although the natural frequency of such events is unclear. Pruitt, for one, would be surprised if the mechanism were exclusive to plants: "It's hard to believe something that general wouldn't persist in other organisms."