The challenge
Head-direction (HD) neurons act as an internal compass, but the signal they carry is only useful if it stays aligned with the outside world. During navigation, landmarks can reset drift, while optic flow and other self-motion cues update heading during turns. In mammals, this visual anchoring is often attributed to interactions between cortical visual regions and the HD network. Larval zebrafish, however, navigate with a tiny brain that lacks an elaborate visual telencephalon, yet they still show sophisticated visually guided behaviours. How can such a compact vertebrate circuit combine landmarks and optic flow, learn a stable landmark-to-heading mapping, and keep it flexible when visual scenes change?
Our approach
We combined two-photon calcium imaging of GABAergic anterior hindbrain HD cells with a panoramic, closed-loop virtual reality arena. By manipulating scenes (landmark jumps, featureless motion, symmetric landmarks) and performing unilateral laser ablations of visual habenula axons, we tested how visual cues and specific pathways shape the HD activity “bump”.
Our findings
HD cells reliably aligned their population activity bump to multiple scenes, with a bias toward distal cues in the upper visual field. They maintained alignment when landmarks jumped (landmark-only) and partially followed rotations in spatial noise (optic-flow-only), indicating dual visual mechanisms. Symmetric “double-sun” training induced an unexpected remapping in which 180° of visual space expanded to occupy the full 360° HD ring. Disrupting the visual habenula projection abolished landmark anchoring while leaving the bump itself detectable. This points to an underlying neuronal architecture, that of a ring attractor, as being responsible for encoding heading.
The implications
A conserved habenula–interpeduncular pathway can implement experience-dependent landmark anchoring of a vertebrate compass—suggesting an evolutionarily ancient solution that does not require cortical visual circuitry.
Creating SyNergies
This work brought together neuroengineering, optical physiology, and theory: building a compact panoramic VR system, imaging population dynamics, modelling Hebbian landmark anchoring, and using targeted laser ablation to test causality. The result illustrates how cross-disciplinary methods can expose conserved navigation principles across species and brain architectures.
