While examining a sennight-old zebrafish larva, researchers at Weill Cornell Medicine and their colleagues solved how a network of neurons in the zebrafish brainstem guides the fish’s gaze. A simplified artificial circuit based on the neuronal system’s architecture, revealed by the study published on Nov. 22 in Nature Neuroscience, can predict activity in the network.
The findings also shed some light on how the brain deals with short-term memory and may provide the basis for novel strategies for treatments for eye movement disorders. There’s a cluster of sensory information that organisms take in all the time about environmental changes happening from moment to moment. The brain must cherish these informational nuggets long enough to form a complete picture: for example, forming sentences by linking together their words or permitting an animal to keep its eyes focused on a field of interest.
Understanding dynamical systems in neuroscience:
Neuroscientists use tools of dynamical systems, that build mathematical models describing how the state of the system changes progressively, the current state determines the next future states according to a set of norms. An example is a short-term memory circuit that will stay in one preferred state until a new stimulus causes it to settle down into a new activity state. Each of these states can store memory in the visual-motor system about where an animal should be looking.
What parameters could be used to establish that type of dynamical system? One possibility is the anatomy of the circuit: how many connections each neuron makes and the connections that form between them. Another probable possibility is the physiological strength of those connections, depending on many factors including how much neurotransmitter (or other signalling molecule) is being released at the synapse, the kind of receptors on the synapse, and the concentration of those receptors.
Dr. Emre Aksay and his collaborators then examined larval zebrafish to help understand what proportion of synaptic potency can be attributed to circuit anatomy. Since the fishlets are five days old, they swim around and hunt prey, an ability that requires sustained visual attention. What’s important for the research team is that the brain area governing animal eye movement in fish and mammals is very similar in structure. However, the zebrafish system only has 500 neurons. “We can analyze the entire circuit, microscopically, and functionally,” Dr. Aksay said. “It’s very difficult to do in other vertebrates.”
Dr. Aksay and colleagues then used an array of advanced imaging techniques to reveal which neurons in the animals control their gaze and how they are wired together. It turned out that the system contains two important feedback loops, each containing the three clusters of cells tightly connected.
This distinctive architecture was used to build a computational model. The activity patterns of the zebrafish circuit were accurately predicted from the artificial network, which was validated by comparison of findings to physiological data. Dr. Aksay describes himself first and foremost as a physiologist. The researchers will then investigate how the cells within each cluster participate in the circuit’s dynamics, and whether the neurons in a cluster are genetically different. Clinicians could therapeutically target such cells if they malfunction in their role of eye movement disorders. The findings also provide a blueprint for studying more complex computational systems in the brain that depend on short-term memory, such as visual scene deciphering or speech understanding.
The authors summarize key factors relevant to rigor and reproducibility, detailed elsewhere in the manuscript. Electron microscopy (EM) data were acquired from one reimaged animal, with no statistical method used to predetermine sample size. Reconstructions, seeded by 22 VPNI neurons identified through functional imaging, were semiautomated, validated, and corrected by experts. Synapses were segmented with a convolutional network (95% accuracy), with a 2% error correction step. Connectivity analyses assumed normal distributions but did not formally test this. Randomization was used extensively to determine modular organization, though investigators were not blinded during analyses.
Reference: Vishwanathan A, Sood A, Wu J, et al. Predicting modular functions and neural coding of behavior from a synaptic wiring diagram. Nat Neurosci. 2024;27(12):2443-2454. doi:10.1038/s41593-024-01784-3‌
