Neurons in the brain that are rarely analyzed by scientists because of their chaotic signaling may be essential for most brain functions, according to a study by Stefano Fusi, PhD, and other neuroscientists at Columbia University Medical Center and MIT. The study was published in the May 30 issue of Nature.
The finding suggests that many studies may have to be reanalyzed to better understand how the brain processes multiple streams of information to make complex decisions that are common in day-to-day life.
“To make decisions we typically need to combine different types of information. For example, before deciding whether to buy something, we need to weigh both the price and our needs,” says Larry Abbott, PhD, co-director of Columbia’s Center for Theoretical Neuroscience, who was not involved in the study. “Fusi’s work shows us how this sort of combining is done in neural circuits, and reveals that errors arise when neurons fail to represent the required combinations.”
Typically, when neuroscientists study the brain, they focus on neurons that fire only in response to one feature, for example, the color of a visual stimulus. Imagine neurons that are activated only by the color red. These neurons may be used by multiple different processes, but their message always remains the same: red. These studies have led to notion that the brain is a patchwork of highly specialized cells, each with a single message.
But the majority of cells in the brain—particularly in the prefrontal cortex, where the most demanding tasks are processed—don’t behave in such an orderly fashion. These cells react to a wide variety of features in the environment, responding to color but in a way that depends on the shape or the movement of the object. So they may respond to a red circle but not to a red triangle.
The jumbled signals of these multi-tasking neurons have bewildered neuroscientists.
Using recent advances in machine learning, Fusi’s team devised a way to analyze the activity of these multi-tasking cells to see what role they played in a complex memory task. The researchers applied the method to a previously published experiment of memory in monkeys conducted by members of Earl Miller’s lab at MIT’s Picower Institute for Learning and Memory.
During the experiment, monkeys were presented with images and asked to pick out the ones they had seen before. The researchers recorded the activity of more than 200 neurons as the monkeys worked on the task.
Fusi’s analysis showed that the multi-tasking neurons appeared to be necessary for the monkey to correctly recall images, even though all the necessary information was also present in the highly specialized neurons.
“Our study is a strong indication that these cells are necessary for most processes in the brain.”
“When monkeys made a mistake, we could decode the identity of the correct images from the recordings. The information needed to make the correct choice is there in the highly specialized cells. But the monkey can’t make the right decision unless the information is also distributed across many multi-tasking cells,” Fusi says.
The study is the first to find a correlation between activity in the multi-tasking cells and behavior, “and it is a strong indication that these cells are necessary for most processes in the brain,” Fusi says. To determine if they are truly essential, the next step is to manipulate these multi-tasking neurons to see how changes affect decision-making.
If their conclusions are accurate, Fusi says, all neuroscientists need to start looking at multi-tasking cells.
“Until now, if researchers didn’t find specialized neurons in an area, they concluded that that area of the brain is not involved in processing that signal. That’s no longer true,” Fusi says.
“Scientists need to switch from recording the activity of single, highly specialized neurons to recording many multi-tasking neurons simultaneously. A lot of studies probably have to be redone or reanalyzed at this point.”
The Gatsby Foundation, the Swartz Foundation and the Kavli Foundation supported this research. Individual authors also received support from Swiss National Science Foundation grant PBSKP3-133357; the Janggen-Poehn Foundation; the McKnight Foundation; the McDonnell Foundation; NIMH grant R37-MH087027; the Picower Foundation; the Brain & Behavior Research Foundation; and a NARSAD Young Investigator grant.