State-switching navigation strategies in C. elegans are beneficial for chemotaxis

TL;DR

This study reveals that C. elegans employs a hierarchical, state-switching navigation strategy driven by sensory input, which enhances its ability to perform gradient climbing and challenges the idea of static, purely sensory-driven behaviors.

Contribution

It introduces a novel statistical model of state-dependent navigation and demonstrates the functional benefits of internal state switching in worm chemotaxis.

Findings

Worm navigation is governed by two persistent internal states.

Transitions between states are causally driven by sensory input.

State-switching improves gradient-climbing performance.

Abstract

Animals employ different strategies for relating sensory input to behavioral output to navigate sensory environments, but what strategy to use, when to switch and why remain unclear. In C. elegans, navigation is composed of 'steering' and 'turns', corresponding to small heading changes and large reorientation events, respectively. It is unclear whether transitions between these elements are driven solely by sensory input or are influenced by internal states that persist over time. It also remains unknown how worms accomplish seemingly surprising feats of navigation--for example, worms appear to exit turns correctly oriented toward a goal, despite their presumed lack of spatial awareness during the turn. Here, we resolve these questions using detailed measurements of sensory-guided navigation and a novel statistical model of state-dependent navigation. We show that the worm's navigation is well described by a sensory-driven state-switching model with two distinct states, each persisting over many seconds and producing different mixtures of sensorimotor relations. One state is enriched for steering, while the other is enriched for turning. This hierarchical, temporal organization of strategies challenges the previous assumption that strategies are static over time and driven solely by immediate sensory input. Sensory input causally drives transitions between these persistent internal states, and creates the appearance of 'directed turns.' Genetic perturbations and a data-constrained reinforcement learning model demonstrate that state-switching enhances gradient-climbing performance. By combining measurement, perturbation, and modeling, we show that state-switching plays a functionally beneficial role in organizing behavior over time--a principle likely to generalize across species and contexts.