The protein family notorious for causing neurogenerative diseases such as Creutzfeldt-Jakob—not to mention mad cow—appears to play an important role in healthy cells.* “Do you think God created prions just to kill?” muses Eric R. Kandel of Columbia University. “These things have evolved initially to have a physiological function.”

Kandel's work on memory helped to reveal that animals make and use prions in their nervous systems as part of an essential function: stabilizing the synapses involved with forming long-term memories. These natural prions are not infectious, but on a molecular level they chain up exactly the same way as their disease-causing brethren. (Some researchers call them “prionlike” to avoid confusion.) Now neuroscientist Kausik Si of the Stowers Institute for Medical Research in Kansas City, Mo., one of Kandel's former students, has shown that the prion's action is tightly controlled by the cell and can be turned on when a new long-term memory needs to be formed.

Once the prion's chain reaction gets started, it is self-perpetuating, and thus the synapse—where neurons connect—can be maintained after the initial trigger is gone, perhaps for a lifetime. But that still does not explain how the first prion is triggered or why it happens at only certain of the synapses, which play a crucial role in forming memories. Si's work, published February 11 in PLOS Biology, traces the biochemistry of this protein-preservation process in fruit flies, showing how the cell turns on the machinery responsible for the persistence of memory—and how the memory can be stabilized at just the right time and in the right place.

Si and his colleagues focused on a protein called Orb2A—its human equivalent is CPEB—that functions as a prion in the flies. A series of molecular interactions results in a phosphate becoming attached to Orb2A but only when an electrical impulse is targeted to a particular synapse among the multitude that can populate a neuron. The specificity allows the prion chain reaction to turn on at the specific time and place needed, stabilizing some synapses but not others—and perhaps explaining why some of our memories fade.

*Clarification (7/2/14): This sentence was edited after its print publication. It originally stated that Parkinson's is part of the same protein family, but the link between prions and Parkinson's has not been conclusively proved.