Published online 19 December 2007 | Nature | doi:10.1038/news.2007.392

News

The power of a single neuron

Stimulating one brain cell can be enough to change behaviour.

Stimulating just one neuron can be enough to affect learning and behaviour, researchers have found. The results, published this week by Nature,conflict with the long-held notion that many neurons — in the order of thousands — are required to generate a behavioural reaction.

GFP neuron picFluorescent protein reveals neurons that can be activated by light.D. HUBER, JANELIA FARM RESEARCH CAMPUS, HOWARD HUGHES MEDICAL INSTITUTE

The findings lend support to the ‘sparse-coding’ hypothesis of neural networks, which suggests that only a few neurons need to fire to generate a response. That theory has been hotly debated, says Karel Svoboda, a neurobiologist at the Janelia Farm research campus at Howard Hughes Medical Institute in Ashburn, Virginia, and one of the study's authors. “There are lots of fights about whether or not neural codes are sparse,” he says.

Svoboda and his colleagues, as well as an independent group of researchers lead by Michael Brecht of the Humboldt University in Berlin, Germany, tackled the debate by studying the region in the rodent brain that receives sensory inputs from the whiskers. That region, called the barrel cortex, is made up of roughly two million neurons. Each whisker transmits signals to a group of cells clustered in the barrel-shaped arrangement that gives the region its name.

The size of the barrel cortex reflects its importance; rats can navigate almost as well with their whiskers as with their eyes. Brecht says that when he did experiments with blind and sighted rats, "You really had to label the cage with the blind rats because they moved so perfectly with their whiskers.”

Light touch

Both teams used techniques that allowed them to stimulate specific sets of neurons. Svoboda and his colleagues created transgenic mice that express a light-responsive protein specifically within the region of the barrel cortex associated with learning. The protein, naturally found in algae, responds to blue light by allowing ions to flow through cell membranes, creating an electric current.

Having implanted a glass window into the skulls of their mice, the researchers mounted a tiny light-emitting diode onto the animals' heads. The technique did not allow them to target individual neurons, but the researchers could vary the intensity of the effect on the cell membranes by dialling the intensity of the light up or down.

Svoboda and his colleagues rewarded their mice with a drink of water every time they correctly selected one of two ports in their cages after stimulation. They found that the mice learned to respond to pulses of light that activated as few as sixty neurons1.

Brecht’s group took a different tact. Working with rats, the researchers implanted electrodes designed to activate single neurons deep within the barrel cortex. They then trained their rats to interrupt a light beam with multiple tongue licks when they sensed the neural stimulation.

Using this method, the researchers found that on average, the rats responded to the stimulation of a single neuron 5% of the time2. But the extent of that response was highly dependent on which neuron was stimulated. Some neurons could provoke a response nearly 50% of the time.

These results could have a radical effect on how neurobiologists view neural networks, says Dirk Feldmeyer, a neurobiologist at Jülich Research Centre in Germany. “Increasing the activity by just a minute amount is actually changing the way the cortex perceives sensory stimuli,” he says. “It’s really changing the view that the cortex is responding to stimuli with massive activity.”

But Brecht and Svoboda readily acknowledge that their findings won’t end the sparse-coding debate. “These experiments show that animals can read out very sparse codes,” says Svoboda. “But they don’t tell you that coding during normal behaviour is sparse.”

To do so would require extremely sensitive imaging techniques to probe neural activity. “That’s not yet been done in a satisfying way,” says Svoboda. “There are technical issues that need to be overcome, but a lot of people are working on that.” 

  • References

    1. Huber, D. et al. Nature 451, 61-64 (2007).
    2. Houweling, A. R. & Brecht, M. Nature 451, 65-68 (2007).
Commenting is now closed.