Scientists have been unable to map the relationship between brain activity and behavior for all the neurons in a brain. This has proven to be a difficult task. After inventing new methods and technologies for this purpose, a group of scientists at The Picower Institute for Learning and Memory, MIT, has created a detailed account of all the neurons within the tiny brain of a C. elegans worm. This includes mapping how the brain cells encode essential behaviors, such as feeding and movement.
The team published in the journal Cell new brain-wide recordings and a mathematical model which accurately predicts how neurons can represent worms’ behaviors. The team applied the model to every cell and produced an atlas that shows how the majority of cells, as well as the circuits in which they participate, encode animal actions. The Atlas reveals how the worm’s brain produces a complex and flexible repertoire even when its environment changes.
This study shows how an animal’s nervous system is organized to control its behavior, said Steven Flavell, Senior Author and Associate Professor at MIT’s Department of Brain and Cognitive Sciences. It shows how the nervous system nodes are organized to encode specific behavioral features. This depends on the animal’s current state and recent experiences.
The study’s lead authors are Jungsoo Kim and Adam Atanas, graduate students. Both earned their PhDs in the spring of this year for their research. The researchers have also made their data and findings from their Atlas and model freely available for other researchers on a website named WormWideWeb.
Models and microscopes
Flavell’s lab developed a new software and microscope system to track almost all behaviors (movement of the worm, eating, sleeping, laying eggs, etc.). The lab also measured the activity of each neuron within its head. (The cells are designed to flash when calcium levels increase). Writing custom software using the latest machine learning tools was necessary to reliably distinguish and track separate neurons while the worm wriggled around and bent. Scientists report that the system was 99.7 percent accurate at sampling neuron activity, with a much-improved signal-to-noise ratio compared to other systems.
The team recorded simultaneous behavior and neuronal data from over 60 worms while they roamed around their dishes doing whatever they liked.
The data analysis revealed three new observations about the neural activity of the worm. Neurons not only track the behavior of the moment but also of recent moments; they tune the encoding behaviors, such as movement, based on an unexpected variety of factors.
While wriggling in a small laboratory dish may seem simple, neurons represent factors like speed, steering, and whether or not the worm is consuming food. They sometimes represented an animal’s movement going back a few minutes. These neurons encode recent motion rather than current direction, which could help the worm calculate how past actions affected its current outcome. Many neurons combine behavioral information to perform more complex maneuvers. In the same way that a driver has to remember to turn the steering wheel in the opposite direction when driving in reverse as opposed to forward, specific neurons within the worm brain integrate the direction of movement and steering direction.
Scientists developed the C. elegans probabilistic neural encoding model after carefully analyzing patterns in how neural activity correlated to behaviors. This simple equation model accounts for how each neuron represents different factors to predict if accurately and how neural activity reflects behavior. Nearly 60% of the neurons within the head of the worm were responsible for at least one conduct.
The research team fitted the model using a probabilistic approach, letting them know their confidence in each model parameter. This approach was pioneered by Vikash Msinghka, a principal researcher scientist at MIT who heads the Probabilistic Computing Project.
The team collected neuron data but did not track the cell’s identity. This was done to create a model to predict and quantify how a brain cell would represent behavioral behavior. The study’s key objective is to determine how each cell contributes to behavior. To apply the model to the specific neurons of each worm, which had all been previously mapped, the next step for the team was to correlate neural activity with behavior in each cell. To do this, each neuron had to be given a unique color to link its action with its identity. The team performed this experiment on dozens of animals that were free to move, and they obtained information about how the behavior of each animal was related to almost all the neurons defined in the head of the worm.
This Atlas, which resulted from this research, revealed many insights. It more accurately mapped the neural circuits controlling each animal’s behavior. Flavell explained that these new findings would allow a better understanding of the control mechanisms for animal behavior.
He said, “It enabled us to complete circuits.” Our colleagues studying aspects of neural circuit functions can refer to this Atlas to get a reasonably comprehensive view of the critical neuron involved.
A significant finding of the team was that, while most neurons obeyed predictions from the model, only a small group of neurons–about 30% of those neurons that encode behavior in the worm brain–could adapt their behavior encoding. They could remap it and take on new tasks. These neurons were consistently similar in all animals and well-connected in the worm’s synaptic diagram.
These remappings could happen for a variety of reasons. So the team conducted further experiments to determine if they could cause neurons to map. The researchers used a laser to heat the agar surrounding the head of the worms as they wriggled in their dishes. The heat was harmless but enough to irritate the worms and cause a behavior change that lasted minutes. The team could see from these recordings that neurons were remapped to encode new behaviors as the animals changed their behavior states.
The authors noted, “Behavioral information is expressed in the brain across many forms, with different tunings, timescales, and levels of flexibility, which map onto the neuron classes defined by the C.elegans connectome.”
The paper is also co-authored by Ziyu Wan, Eric Bueno, and McCoy Becker. Other authors include Talya Kremer, Flossie Wan, Saba Basoylu, Ugur Dagh, Elpiniki Kalogeropoulou, and Matthew Gomes. Cassi Estrem and Netta Cohl are also included.
The research is funded by the National Institutes of Health (NIH), the National Science Foundation (NSF), The McKnight Foundation (McKnight Foundation), The Alfred P. Sloan Foundation (The Alfred P. Sloan Foundation), The Picower Institute for Learning and Memory and The JPB Foundation.
