Summary: Researchers have discovered that the mouse brain can simultaneously encode multiple hypotheses about their spatial location while sailing environments with ambiguous reference points. In a complex task that requires mice to distinguish between identical signals, neurons in the retrosple cortex (RSC) exhibited different activity patterns that reflect different possibilities of where the mouse could be.
These patterns collapsed in a single “correct” once more information available, which shows that the brain not only stores ideas in competition, but uses them to make decisions. This is the first time that navigation coding based on brain hypothesis is directly observed.
Key facts:
Multiple coded hypotheses: RSC neurons maintain separate neuronal representations for different possible locations. Traking in action: These representations are not only passive: they guide the mouse towards the correct objective once the ambiguity is resolved.
Source: MIT
By navigating a place with which we are only familiar, we often trust unique reference points to help us make our way. However, if we are looking for an office in a brick building, and there are many brick buildings along our route, we could use a rule such as looking for the second building in a street, instead of depending on distinguishing the building itself.
Until this ambiguity is resolved, we must take into account that there are multiple possibilities (or hypothesis) where we are in relation to our destiny. In a mice study, MIT’s neuroscientists have discovered that these hypotheses are explicitly represented in the brain by different neuronal activity patterns.
This is the first time that neuronal activity patterns have been observed that encode simultaneous hypotheses in the brain. The researchers found that these representations, which were observed in the retrosplenial cortex of the brain (RSC), not only encode hypotheses, but also could be used by animals to choose the correct way to do so.
“As far as we know, no one has demonstrated in a complex reasoning task that there is an area in the association cortex that takes into account two hypotheses and then uses one of those hypotheses, once you get more information, to complete the task,” says Mark Harnett, associate professor of brain and cognitive sciences, a member of the MC Gough Institute of the MIT for brain research and the senior author of the study of the study.
Jakob Voigts PhD ’17, Ex Postdocs in the Harnett laboratory and now a group leader at the Janelia research campus of the Howard Hughes Medical Institute, is the main author of the article, which appears today in the neuroscience of nature.
Ambiguous reference points
The CSR receives information from the visual cortex, the formation of the hippocampus and the anterior thalamus, which integrates to help guide navigation.
In a 2020 article, Harnett’s laboratory discovered that the CSR uses visual and spatial information to encode reference points used for navigation. In that study, the researchers showed that neurons in the RSC of mice integrate visual information about the surrounding environment with the spatial feedback of the position of the mice along a track, which allows them to learn where to find a reward based on reference points they saw.
In their new study, researchers wanted to deepen how the CSR uses space information and situational context to guide navigation decision making. To do that, the researchers devised a much more complicated navigation task than the one typically used in mice studies. They configured a large round sand, with 16 small openings, or ports, along the side walls.
One of these openings would give mice a reward when they put their nose through it. In the first set of experiments, the researchers trained the mice to go to different reward ports indicated by light points on the floor that were only visible when the mice approach them.
Once the mice learned to perform this relatively simple task, the researchers added a second point. The two points were always at the same distance from each other and from the center of the sand. But now the mice had to go to the port next to the antihorarium point to obtain the reward.
Because the points were identical and only made visible at close distances, the mice could never see both points at the same time and could not immediately determine what point it was which.
To solve this task, the mice had to remember where they expected a point, integrating their own body position, the direction in which they were heading and the path they took to discover what point of reference is which.
When measuring the RSC activity as mice were approaching the ambiguous reference points, researchers could determine if the CSR encodes hypotheses about spatial location. The task was carefully designed to require mice to use visual reference points to obtain rewards, instead of other strategies such as smell signs or dead calculations.
“The important thing about the behavior in this case is that mice need to remember something and then use that to interpret future contributions,” says Voigts, who worked on this study while it was a postdoc in Harnett’s laboratory.
“It’s not just remembering something, but remember in such a way that you can act accordingly.”
The researchers found that, as mice accumulated information about what DOT could be whose, RSC neurons populations showed different activity patterns for incomplete information. Each of these patterns seems to correspond to a hypothesis about where the mouse thought about the reward.
When the mice approached enough to discover what point the reward port indicated, these patterns collapsed in which it represents the correct hypothesis. The results suggest that these patterns not only storpted passively hypothesis, but can also be used to calculate how to get to the correct location, researchers say.
“We show that RSC has the information required to use this short -term memory to distinguish the ambiguous reference points. And we show that this type of hypothesis is coded and processed in a way that allows the RSC to use it to solve the calculation,” says Voigts.
Interconnected neurons
When analyzing their initial results, Harnett and Voigts consulted with the MIT teacher, Ila Fiete, who had conducted a study about 10 years using an artificial neuronal network to perform a similar navigation task.
That study, previously published in Biorxiv, showed that the neuronal network showed activity patterns that were conceptually similar to those observed in animal studies administered by the Harnett laboratory. The neurons of the artificial neuronal network ended up forming lowly interconnected networks, such as RSC neurons.
“That interconnectivity seems, in forms that we still do not understand, to be key to how these dynamics arise and how they are controlled. And it is a key feature of how RSC takes into account these two hypotheses at the same time,” says Harnett.
In his laboratory in Janelia, Voigts now plans to investigate how other brain areas involved in navigation, such as prefrontal cortex, commit themselves as mice explore and forge in a more naturalistic way, without being trained in a specific task.
“We are investigating whether there are general principles by which tasks are learned,” says Voigts. “We have a lot of knowledge in neuroscience about how the brains operate once the animal has learned a task, but in comparison we know extremely little about how mice learn tasks or what they choose to learn when they are given freedom to behave naturally.”
Funds:
The investigation was financed, in part, by the National Health Institutes, a Simons center for the social brain of the Postdoctoral MIT Felowship, the National Institute of General Medical Sciences and the center of brains, minds and machines in the MIT, financed by the National Science Foundation.
On this neuroscience research news
Author: Sarah McDonnell
Source: MIT
Contact: Sarah McDonnell – Mit
Image: The image is accredited to Neuroscience News
Original research: open access.
“Spatial reasoning through the recurring neuronal dynamics in the retrosplenial cortex” by Mark Harnett et al. Nature neuroscience
Abstract
Spatial reasoning through recurrent neuronal dynamics in the retrospplenial cortex of the mouse
From visual perception to language, sensory stimuli change their meaning depending on previous experience.
Recurrent neuronal dynamics can interpret stimuli based on the context of external indication, but it is unknown if they can calculate and use internal hypotheses to solve ambiguities.
Here we show that the retrosplenial mouse cortex (RSC) can form several hypotheses over time and perform spatial reasoning through recurrent dynamics.
In our task, the mice sailed using ambiguous reference points that are identified through their mutual spatial relationship, which requires a sequential refinement of hypotheses.
RSC neurons and artificial neuronal networks encoded mixtures of hypothesis, location and sensory information, and were limited by a low dimension robust dynamic.
RSC encoded hypotheses as locations in the activity space with divergent trajectories for identical sensory entries, allowing its correct interpretation.
Our results indicate that the interactions between internal hypotheses and external sensory data in recurrent circuits can provide a substrate for complex sequential cognitive reasoning.
