A Neuronal Circuit Motif for Leaky Vector Integration

Author

Abby Janke

Date of Award

2024

Document Type

Thesis

Degree Name

Doctor of Philosophy (PhD)

Thesis Advisor

Gaby Maimon

Abstract

To efficiently navigate their environments, animals ranging from insects to humans can keep track of their trajectory by continuously summing their direction and distance traveled over time. It is not clear how brains perform this fundamental computation––known as path integration––even though neural correlates and computer models_1–4 abound. In arthropods, spatial-cognition related processes such as path integration are thought to be implemented in the central complex^2,5,6. Experimenters have shown that contiguously arrayed populations of neurons in the Drosophila fan-shaped body, a substructure of the central complex, display sinusoidally shaped bumps of calcium activity that signal vectors relevant for spatial navigation. The phase of the bump signal represents the angle of a vector, and the amplitude represents the length. To date, the sinusoidal bumps discovered in the fan-shaped body signal real-time variables, such as a fly’s traveling^7,8 or goal directions⁹, rather than a vector that is integrated over time. Here we characterize a class of neurons in the fan-shaped body, called h∆G cells, whose activity drops dramatically when a fly encounters a food source (i.e., a sugar drop). After this reset, the mean activity of h∆G neurons rises over many minutes and the amplitude of a sinusoidally shaped bump signal grows and shrinks on the seconds timescale. We show that the amplitude of the bump tracks the distance the fly has walked in the past few seconds. We further argue––based on calcium imaging and optogenetic perturbations––that the bump signal in hΔG neurons is built via integration of synaptic input from v∆E cells. Based on these data, we developed a formal model for the v∆E-hΔG integration process that combines (1) a minutes-long rise in baseline activity, (2) a vΔE-driven slow-down of this rise and (3) a continuous leak of the h∆G signal. This leak helps preserve the sinusoidal shape of the bump signal but also causes the h∆G signal to not reflect a perfect (i.e., leak-free) path integral. Our experimental results focus on only vΔE and hΔG neurons, but the fly’s connectome^10–12 reveals multiple vΔ-hΔ circuits, spanning different layers of the fan-shaped body. By varying leak rates and ramp rates across different vΔ-hΔ circuits, the fly brain could build a repertoire of path-integrated signals with spatiotemporal dynamics suitable for a range of navigational tasks.

Comments

A Thesis Presented to the Faculty of The Rockefeller University in Partial Fulfillment of the Requirements for the degree of Doctor of Philosophy

Recommended Citation

Janke, Abby, "A Neuronal Circuit Motif for Leaky Vector Integration" (2024). Student Theses and Dissertations. 808.
https://digitalcommons.rockefeller.edu/student_theses_and_dissertations/808