By Shankar Vedantam Washington Post Staff Writer Tuesday, August 28, 2001; Page A03
Scientists for the first time have linked multiple brain cells with silicon chips to create a part-mechanical, part-living electronic circuit.
To construct the partially living electronic circuit, scientists at the Max Planck Institute for Biochemistry in Germany managed to affix multiple snail neurons onto tiny transistor chips and demonstrated that the cells communicated with each other and with the chips.
The advance is an important step toward a goal that is still more science fiction than science: to develop artificial retinas or prosthetic limbs that are extensions of the human nervous system. The idea is to combine the mechanical abilities of electronic circuits with the extraordinary complexity and intelligence of the human brain.
Such combinations of biology and technology may not only one day help the blind to see and the paralyzed to move objects with their thoughts, but also help to build computers that are as inventive and adaptable as our own nervous systems and a generation of robots that might truly deserve to be called intelligent.
Meshing nerve cells with electronics has become a hot new field in science -- and has long been a staple of science fiction. But what "Star Trek" accomplished in a stroke of the pen has proved harder to achieve in real life.
"The nervous system is quite different than a computer," said Eve Marder, a professor of neuroscience at Brandeis University who studies how the brain adapts to change. "Many functions that are physically separate in a computer are carried out by the same piece of tissue" in the brain and nervous system.
The greatest challenge has been in building the interface between biology and technology. Nerve cells in the brain find each other, strengthen connections and build patterns through complex chemical signaling that is driven in part by the environment. Slice away some neurons, for example, and others will leap in to replace their function. No one understands how the brain learns to adapt to change, but it is a process that is as sophisticated as it is messy.
Silicon chips, on the other hand, can perform specific functions with great reliability and speed, but have limited responsiveness to the environment and almost no ability to alter themselves according to need.
"Things are constantly changing . . . processes are growing, there are substances called neuromodulators that change the properties of nerve cells and the strength of connections," said Marder. "That's the challenge of making a silicon-brain interface -- the rules of computation are not the same."
The German researchers used micropipettes to lift individual cells from the snail brain and then puff them out onto silicon chips that were layered with a kind of glue. The snail neurons, according to biophysicist Peter Fromherz, are a little larger than human or rat neurons and were therefore easier to work with.
"They suck them out and then blow them onto the structure," said Astrid Prinz, a post-doctoral researcher at Brandeis University, who used to work with the German group. "It's a matter of practice to learn to handle individual cells. You have them in a little pipette with fluid. You blow them out and you can maneuver them. One guy in the lab made a little movie on how to blow cells."
Each cell was positioned over a Field Effect Transistor, a device that is capable of amplifying tiny voltages, and a stimulator to prod the cell into activity.
The process was repeated with some 20 cells over multiple transistors and stimulators. By using polymers, the German scientists built tiny picket fences around the neurons to keep them in place over the transistors -- one of the great difficulties in building such circuits is that nerve cells tend to wander around, as they do in the brain.
Neurons on this silicon base developed a connection between each other known as a synapse. When researchers stimulated one neuron, it released an electrical signal. That signal was detected by the transistor that the neuron sat on as well as the transistor beneath a second neuron -- showing that the electrical signal had passed from the chip to the first neuron, through a synapse to the second neuron and then converted back into electricity and the second transistor.
"It's very primitive, but it's the first time that a neural network was directly interfaced with a silicon chip," said Fromherz, who published the results in today's issue of the Proceedings of the National Academy of Science. "It's a proof of principle experiment."
The group, he said, was already working on linking greater numbers of neurons with more transistors. The real challenge, he said, lay in figuring out where exactly the neuron's synapse was relative to the transistor, and in developing techniques that could reliably construct larger circuits.
Fromherz said plans were underway to build a system with 15,000 neuron-transistor sites.
When the number gets large enough, researchers hope they will begin to see the early glimmers of what actually happens in the brain: neurons forming complex connections that transmute electrical activity into computation, thoughts and maybe consciousness itself.