A brain must constantly cope with the highly variable, fast-paced nature of the world when trying to judge what's going on around it. On one hand, it has to be open to whatever new sensory information may come its way, but on the other hand, just to keep up, it has to try to leverage prior experience to make predictions about what seems to be happening. In a new study in Science, MIT neuroscientists identify a circuit that links a sensory decision-making region with one that advises it on how much sensory information just changed.
"This circuit organizes a comparison between what has just happened versus what is happening now in the sensory world in a manner that can be used to act," said study senior author Mriganka Sur , Newton Professor in The Picower Institute for Learning and Memory and MIT's Department of Brain and Cognitive Sciences.
Study lead author Ning Leow, a former graduate student in Sur's lab who is now a postdoc at A*STAR in Singapore, said the study in mice sheds light on closely analogous circuitry in humans, in which an area of the prefrontal cortex (the anterior cingulate cortex, or ACC) makes sensory decisions. The new study shows it bases those decisions on advice about immediate past history from an area of the thalamus called the pulvinar (though in mice, it's called the lateral posterior thalamus, or LP).
"The brain does not evaluate each new event from scratch," Leow said. "The pulvinar has traditionally been studied for its role in attention and filtering visual information, but we found that it was also important for comparing present information with the immediate past and highlighting meaningful changes to influence whether we maintain or update a decision."
As part of Sur's longstanding interest in how the brain's cortex integrates sensory perception and learning to produce behavior, Leow and Sur began comprehensively mapping the copious inputs to the LP-ACC circuit, culminating in a paper in 2022 . It was clear from that study how the circuit would seem well positioned to help focus attention, which is what it was known for at the time.
But in thinking more deeply about what focused attention is for, and about how these well-connected regions seemed to sit at the center of not only attention but also perception and action, Sur and Leow hypothesized that they might also have a hand in guiding decisions based on sensory information. The new study presents multiple lines of evidence that it does.
The findings not only shed light on a fundamental function of the brain, Sur said, but could also be applicable to studies of autism, in which many patients show significant differences in the predictions they make about the sensory world. Often, this manifests as difficulty filtering out stimuli that neurotypical people are able to regard as recurring and therefore mundane.
Which way?
To conduct the study, the researchers trained lab mice to play a video game in which dots on a screen would drift around, but at least some would move together in the same direction (left or right). In each trial, the mice had to discern that trend. From one trial to the next, then, the sensory cue could vary not only by the direction of movement, but also by how what proportion of dots were participating. For instance, on one trial maybe 64 percent of the dots would move left and on the next trial maybe 16 percent of the dots would move right. In this way, the researchers could measure a whole continuum of differences from one trial to the next.
Meanwhile, as mice played the game, the scientists used a two-photon microscope to record the activity of the LP-ACC circuit and the response of neurons in the ACC. In some experiments, they used a technique called optogenetics to artificially activate the circuit.
By tracking how mice performed the task trial after trial, the researchers were able to see that the mice indeed factored in not only what they were seeing in the moment but also what they had just seen previously. For instance, when mice guessed right, they were very likely to repeat their guess if the new cue was very similar to the prior one and very unlikely to if the cue was very different. But if they guessed wrong, then the opposite was true: they wouldn't repeat that decision if the cue was similar to the last, but would if it looked very different.
Looking in the brain
Of course, behavioral observations only indicated that the mice indeed compared new cues to prior ones. Determining whether that was indeed because of the LP-ACC circuit required the researchers to use optogenetics to perturb it (by stimulating extra activity in the LP's inputs into the ACC). For instance, optogenetic perturbation of the circuit in the left brain hemisphere made mice less likely to guess that dots were moving right and perturbation in the right hemisphere made mice more likely to guess dots were moving to the right. But in both cases, the extent of these deviations from normal behavior was directly proportional to the difference between the current cue and the previous one. In other words, perturbing the circuit disrupted how mice used recent sensory history when evaluating new evidence, Leow said.
"That showed the pathway is causally involved in the comparison process that influences how current evidence is interpreted rather than merely carrying the information," Leow said.
Moreover, using the microscope imaging (which visualizes calcium levels in neurons, a close proxy of the electrical activity), the researchers extensively analyzed the activity patterns of the LP input into the ACC and how ACC neurons reacted to that input.
"The main takeaway is that the LP and ACC were performing different jobs," Leow said. "The pulvinar doesn't appear to be making the decision itself. Instead, it sends that history-referenced sensory comparison to the frontal cortex. The ACC then transforms that information into the neural activity that predicts the animal's final choice."
Essentially, the pulvinar advises the ACC on the degree of change so that the frontal cortex can consider whether it's time to change a guess. After all, if a mouse is guessing right and little is changing, why not keep on trucking? But if there's a big change, then it might make sense for the mouse to re-evaluate what it's thinking.
It turns out, the brain has this dedicated circuit for doing so.
In addition to Leow and Sur, the paper's other authors are Arundhati Natesan, Alexandria Barlowe, Sofie Ährlund-Richter, Tianyu (Cindy) Luo, and Mehrdad Jazayeri.
The National Institutes of Health, a MURI grant, the Simons Foundation Autism Research Initiative, A*STAR, and the Freedom Together Foundation funded the research.
