Molecular switch makes memory stick.

How unstable molecules can store information stably has long baffled scientists. Human long-term memory may depend on such a mechanism. Recently, a biologist at Brandeis University in Waltham, Mass., proposed a theoretical model for a molecular "switch," made up of two enzymes with opposite actions. The system could store information indefinitely despite continual turnover of the individual molecules that make up the switch.

John E. Lisman describes in the May (Number 9) PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES a hypothetical chemical switch made of a kinase (an enzyme that catalyzes phosphate group transfer to form triphosphates like ATP) and a phosphatase (an enzyme that cleaves a phosphate group from a molecule). The switch is bistable, Lisman says, because it can be either "on" or "off" but is never in between. It is local, he says, because it could be located in specific compartments of neurons -- such as dendrites, the spiny processes that conduct impulses to other neurons -- rather than centralized in DNA of the cell nucleus.

In his paper, Lisman discusses how this molecular switch could be turned on permanently by an external stimulus. In the "off" state, the kinase, which Lisman calls kinase-1, is unphosphorylated and inactive. A stimulus such as a protein binding to a receptor on a cell membrane, Lisman says, could activate a second kinase, kinase-2, which phosphorylates kinase-1. Kinase-1 is activated briefly until the phosphatase removes a phosphate group and returns kinase-1 to its inactive state.

When kinase-2 stimulation reaches a critical level, kinase-1 becomes activated permanently. Phosphorylation of kinase-1 happens faster than phosphatase can reverse it, and kinase-1 starts to phosphorylate and activate other kinase-1 molecules. Thus, even if the stimulus is removed and kinase-2 activity is eliminated, kinase-1 activity continues. The natural replacement of kinase-1 molecules over time does not inactivate the system. "Thus long-term information can be stored by this switch," Lisman says, "even though the molecules that make up the switch turn over rapidly and completely."

Lisman's model and a similar one proposed last year by Francis Crick of the Salk Institute for Biological Studies in La Jolla, Calif., differ from previous models for memory storage. According to these earlier models, Lisman says, an elementary bit of information is stored in a neuron by phosphorylating a key enzyme and turning it on. The "on" state lasts as long as the phosphatase that can dephosphorylate the enzyme is not present. Even in the absence of phosphatase, he says, the phosphorylated enzyme molecules would be gradually replaced by new unphosphorylated enzyme. Such a switch would be "on" only for the lifetime of the enzyme.

Because of these limitations, the previous models hypothesized that the switch would be in a cell's DNA. However, Lisman says, "It is hard to see how nuclear DNA could independently control (or be controlled by) events in subcellular compartments, such as the thousands of spines that constitute the input regions of neuronal dendrites." In contrast, he says, kinase molecules localized in dendritic spines could easily phosphorylate other kinase molecules within the same spine "and thereby form local bistable switches that are independent of the switches in other spines."

Lisman cautions that the molecular switch model is theoretical and has not been observed in any living systems. "But the surprise of the model," he says, "is how you can take known types of biochemical reactions and get something as complicated as memory storage."