Computational Biology Group/Intra-University Cooperative LaboratoriesToyoshima Laboratory
(Systems Biology for Neuroscience, BS)

Systems biology, Information processing, Whole-brain imaging, Animal behavior, Bioimage informatics, Mathematical modeling, C. elegans
How Does Nervous System Perform Computations?

The animal brain is the most complex information processing system in any living organism. To understand how real nervous systems perform computations is one of the fundamental goals of neuroscience. Neural activation with stimulation should propagate through the neural circuits and induce behavioral responses. We want to understand how the environmental information is encoded in the neural activities and interacts with the internal states of the neural circuits. Measuring whole-brain activity and behavior will be important in revealing information processing mechanisms.
Animals move to their preferred environment based on the smells and sounds of their surroundings. Such navigation behaviors are suitable for clarifying the mechanism of information processing in neural circuits because the input-output relationship is clear: the organism senses the external environment, selects the necessary information, and outputs the processing result as a behavior.
We use the nematode Caenorhabditis elegans in our research. The worms remember salt concentration at which they were cultivated with food, and they migrate to the region of the specific concentration in an environment with a salt concentration gradient. The neurons and their connections in the worm have been studied in detail. However, even in the organism with the most extensive fundamental information in neuroscience, how each neuron processes information to produce behavior is not well understood.

Whole-Brain Activity Imaging

We thought that simultaneous measurement of whole-brain activity with single-cell resolution will be the key-techniques in revealing information processing mechanisms. Since the body of the nematode is small and transparent, we can measure the neural activity of living animals under the fluorescent microscopes with microfluidic devices. We have developed experimental techniques such as 4D microscope (Fig. 1), which allows us to simultaneously observe the all neurons in the head region of the worm, and image analysis methods to extract neural activity from volumetric time-lapse movies. We have also developed a cell identification method that maps individual cells in the movies to known nerves in order to compare the extracted neural activities among individuals and map them to neural circuits. By integrating these techniques, we have achieved the whole-brain activity imaging.

  • Fig.1: 4D microscope

Simultaneous Measurement of Neural Activity and Behavior

Independent component analysis of time-series data of whole-brain neural activity revealed that external stimuli affect only a few neurons, while most other neurons show spontaneous activity correlated with the movement of the worm. Therefore, we have constructed a tracking 4D microscope by combining a 4D microscope with a motorized stage to measure the whole brain neural activity of free-moving animals while automatically tracking them (Fig. 2).
We can also measure neural activity while inducing specific postures and behaviors by using a microfluidic channel that we developed. By measuring neural activity and behavior at the same time, we can identify patterns of neural activity related to behavior and investigate the mechanisms that generate the patterns.

  • Fig.2: Whole-brain activity imaging of free-moving

Mathematical Modeling and Simulation

Neurons are connected to each other to form a complex network. Neural activity generates posture and behavior through muscle contraction, and the results affect the nervous system by priproceptive feedback and changes in the external environment. Quantitative mathematical modeling and simulation are important to understand the dynamics of such multi-level and multi-element systems.
The worms migrate to a region of the specific oxygen concentration in an environment with a gradient in oxygen concentration. We have succeeded developing a new mathematical model that reproduces the neural activity of oxygen-sensing neurons over a wide range of oxygen concentrations. Based on this neural model and known neuroscience knowledge, we have developed a comprehensive mathematical model that quantitatively reproduces the experimental result of the aerotaxis behavior (Fig. 3).
We succeeded to develop a deep learning model that predict the posture sequences of the worms. By combining this model with optogenetic perturbation and reinforcement learning, we also succeeded to automatically learn strategies to control the behavior of the worms.

  • Fig.3: Comprehensive mathematical model of aerotaxis behavior

Page Top