For a neuroscientist, the opportunity to record single neurons in people doesn’t knock every day. It is so rare, in fact, that after 14 years of waiting by the door, Florian Mormann says he has recruited just 110 participants—all with intractable epilepsy. All participants had electrodes temporarily implanted in their brains to monitor their seizures.
But the slow work to build this cohort is starting to pay off for Mormann, a group leader at the University of Bonn, and for other researchers taking a similar approach, according to a flurry of studies published in the past year. For instance, certain neurons selectively respond not only to particular scents but also to the words and images associated with them, Mormann and his colleagues reported in October.
Other neurons help to encode stimuli, form memories and construct representations of the world, recent work from other teams reveals. Cortical neurons encode specific information about the phonetics of speech, two independent teams reported last year. Hippocampal cells contribute to working memory and map out time in novel ways, two other teams discovered last year, and some cells in the region encode information related to a person’s changing knowledge about the world, a study published in August found.
These studies offer the chance to answer questions about human brain function that remain challenging to answer using animal models, says Ziv Williams, associate professor of neurosurgery at Harvard Medical School, who led one of the teams that worked on speech phonetics. “Concept cells,” he notes by way of example, such as those Mormann identified, or the “Jennifer Aniston” neurons famously described in a 2005 study, have proved elusive in the monkey brain.
Some researchers say they are worried about maintaining the momentum for meaningful single-neuron studies because the opportunity may soon knock even less often. Mormann, for example, has only about 10 participants per year from whom he and his colleagues can reliably record—and that population may shrink as noninvasive methods for tracking and treating epilepsy improve, he says.
But as findings from these studies start to pile up, researchers “are becoming less afraid to ask very difficult questions that you can uniquely ask in people,” Williams says. Their success so far opens the door to understanding additional brain functions, such as the encoding and production of language, and more abstract representations such as cognitive maps, he adds. Ultimately, conducting single-neuron studies “will help us start understanding how the human brain works.”
A
mong the recent revelations from single-neuron studies is a more granular understanding of the mechanisms of memory.The systems that enable working memory and long-term memory, for example, are typically thought to be distinct from each other, based mainly on evidence from animal studies. But the neural mechanisms used for the two systems overlap more than previously thought and are linked functionally in the human brain, according to findings published in December in Neuron.
The participants in that study completed a test of working memory in which they viewed an image, such as a lake or a cat, and then, after a two-second delay, had to say whether a new picture matched the first. After 10 to 30 minutes, the participants viewed a string of images, some of which had been included in the first part of the task, and they had to say whether any of the images had been repeated—a test of their long-term memory.
A set of neurons in the hippocampus turned on and stayed on when a participant held an image in their working memory, and this persistent activity predicted whether they later recognized the item, the team found. This shows that short- and long-term memory exist “in a continuum,” says Tansu Celikel, professor of psychology at the Georgia Institute of Technology, who was not involved in the work.
The persistently active neurons in the hippocampus function as an intermediate between long-term and working memory, “like a bridge between the two systems,” says Ueli Rutishauser, professor of neurosurgery at Cedars-Sinai, who led the study. Those findings challenge previous ideas about working memory, he adds. “If you open the textbook, [the hippocampus] is not considered to be part of the network that supports working memory,” he says. “And that’s not entirely true.”
A
nother mechanism of human memory may differ from that in animals, other single-neuron studies suggest.In particular, people might differ from animals in the neural mechanisms through which they use existing knowledge about their world to make inferences about new situations—a process that also relies on the activity of cells in the hippocampus, Rutishauser and his colleagues reported in August. “This is more than memory: It’s really about abstracting across a lot of memories to have knowledge,” he says.
Participants in that study viewed a sequence of pictures and had to respond to each image by pressing one of two buttons, only one of which was considered to be a “correct” response and was associated with a small monetary reward. They learned the correct responses (which were based on a specific rule that the participants were not privy to) through trial and error, but partway through the task, the rule changed—as did the correct responses. Once participants figured out that a switch had happened, they could then infer the correct responses for pictures they had not yet seen.
As the participants learned these implicit rules, the activity of neurons across multiple areas changed, Rutishauser and his team found. And once participants figured out that the rules had changed, the activity in only the hippocampus became reformatted—suggesting that these cells had encoded the new rules in an abstract format.
Previous approaches, such as functional MRI, had failed to tackle this problem: “There are many who have tried to see these abstract structures in the hippocampus and how they change as a function of learning,” Rutishauser says. “So far, nobody has figured out how to do that. We have only seen that kind of representational change by looking at single neurons.”
Similar studies in monkeys, on the other hand, had shown that these changes are widely observed throughout the brain, and that abstract representations of inferred information are associated with the frontal lobe in addition to the hippocampus.
That points to the importance of performing these experiments in people, Mormann says. “Now we have access to the human brain, and we find that there are actually striking differences,” he says. “And these differences, my impression is, tend to be ignored.”
O
f course, not all brain mechanisms differ between animals and humans. People are just like macaques and rodents when it comes to the hippocampus encoding the “when” of a memory, according to a study published in September.Just as the brain maps out a person’s physical surroundings, it can also form maps of abstract spaces, including concepts, relationships and time—abilities thought to arise from the activity of hippocampal cells. “There’s quite a lot of evidence that this is the case in humans, but it was coming from neuroimaging methods that don’t have this type of resolution,” says Pawel Tacikowski, assistant investigator at Coimbra University, who worked on the study as a postdoctoral researcher in Itzhak Fried’s lab at the University of California, Los Angeles.
Tacikowski and his colleagues arranged six different photos of people on a triangle—positioning each at one of the triangle’s corners or along its sides—and then presented the images in a sequence to 17 participants, who had to indicate if the pictured person was male or female. The sequence related to the images’ position on the triangle: Only those that were directly connected to each other by lines running along the sides or through the triangle’s center were shown one after the other. As the participants viewed this coordinated set of images, neurons in the hippocampus modified their activity to encode the temporal structure of the sequence.
The activity of single cells coordinated to create temporal representations, which could help the brain predict what image would come next. Put another way, through that changing activity, the cells time-stamped the experience, Williams says. Macaques similarly create and use cognitive maps in this way, but this temporal encoding had not been previously recorded in humans.
Numerous other open questions could be answered through single-neuron recordings. But because of improved methods for studying epilepsy, it’s possible that people soon won’t need to be implanted with electrodes, cutting off researchers’ access to a person’s neuronal activity. “As noninvasive methods get better and better, especially structural MRI imaging, there might be fewer and fewer patients,” Mormann says.
To combat the scarcity of data from single-neuron recordings, researchers need to make more of an effort to centralize what exists and make it accessible, Celikel says. “We should push for a giant database where all this data can be combined together,” he says. “Together we can make maximum utilization of this data in a way that the field can advance all together.”
Because of the potential decrease in access to epilepsy patients, Mormann says he hopes researchers put that limited opportunity to good use. “I hope that enough researchers resist the temptation of dedicating themselves to replicating rodent findings,” he says. “We don’t know for sure how long this window will remain open.”
