Caltech Scientists Devise First Neurochip

NEW ORLEANS—Caltech researchers have invented a "neurochip" that connects a network of living brain cells wired together to electrodes incorporated into a silicon chip.

The neurochips are being unveiled today at the annual meeting of the Society for Neurobiology, which is being held in New Orleans the week of October 25-30. According to Dr. Jerome Pine, one of the five coinventors of the neurochip, the technology is a major step forward for studying the development of neural networks.

The neurons used in the network are harvested from the hippocampus of rat embryos. Once the cells have been separated out by a protein-eating enzyme, each is individually inserted into a well in the silicon chip that is about half the diameter of a human hair. The cell is spherical in shape when it is inserted and is slightly smaller in diameter than the silicon chip well. When it is set in place and fed nutrients, it grows dendrites and an axon that spread out of the well.

In doing so, each neuron remains close to a single recording and stimulating electrode within the well, and also links up with other dendrites and axons attached to other neurons in other nearby wells.

According to Michael Maher, one of the coinventors, the neurochip currently has room for 16 neurons, which appear to develop normal connections with each other. "When the axons meet dendrites, they make an electrical connection," says Maher, who left Caltech in September to assume a postdoctoral appointment at UC San Diego. "So when one neuron fires, information is transmitted to the next neuron."

The neurochip network will be useful in studying the ways in which neurons maintain and change the strengths of their connections, Maher adds. "It's believed that memory in the brain is stored in the strength of these connections.

"This is pretty much a small brain connected to a computer, so it will be useful in finding out how a neural network develops and what its properties are. It will also be useful for studying chemical reactions at the synapses for weeks at a time. With conventional technology, you can record directly from at most a few neurons for at most a couple of hours."

There are two challenges facing the researchers as they attempt to improve the neurochips. One is providing the set of growth factors and nutrients to keep the cells alive for long periods of time. At present, two weeks is the limit.

The second challenge is finding a way to insert the cells in the silicon wells in a less time-consuming way. At present, the technique is quite labor intensive and requires a highly skilled technician with considerable patience and dexterity.

Other than the sheer effort involved, however, there is no reason that millions of cells could not be linked together at present, Maher says.

The other Caltech coinventors of the neurochip are Hanna Dvorak-Carbone, a graduate student in biology; Yu-Chong Tai, an associate professor of electrical engineering; and Tai's student, John Wright. The latter two are responsible for the silicon fabrication.