Visual object recognition in the fly: the internal representation of a predator drives defensive behaviour

Sachira Denagamage, Merrin Foltz, Valentin M. Dalstein, Rudy Behnia, Gwyneth M. Card, Richard Axel & Kevin M. Cury*

*corresponding author

bioRxiv (2026), doi.org/10.64898/2026.11.09.765114

pdf

Animals recognize visual objects and respond to them appropriately to survive. This process is thought to unfold in stages: the visual scene is first deconstructed into constituent features, which are then reconstructed into representations of object identity and relayed to motor circuits. Here we define a neural circuit in Drosophila that underlies the visual recognition of a natural predator, the parasitoid wasp, and orchestrates defence. We found that flies exhibit a repertoire of directed defensive actions structured by the wasp’s position within the fly’s field of view. Expression of these behaviours is driven by visual detection of a conspicuous morphological cue — the wasp’s elongated antennae. A small subset of lobula columnar (LC) visual projection neuron types are required for defence and are activated by constituent visual features of the wasp. Three LC types respond selectively to the wasp’s antennae, and relay this information to LC9, which sums antennal- and body-feature signals to form a composite representation of the wasp. Spatial filtering of this representation onto a descending premotor pathway translates the wasp’s location in visual space into directed defence. Feature combination followed by spatial filtering may constitute a general mechanism for coupling object identity to context-appropriate action.

Fly–male wasp interactions in a rectangular chamber. Two representative video recordings each showing the initial 4 min of the lights-on session of an interaction between a female D. melanogaster fly and a male L. heterotoma parasitoid wasp with intact (MWasp intact; TOP video clip) or fully removed antennae (MWasp ablated; BOTTOM video clip). Video playback speed is 10× real time.

Fly–fly interactions in a rectangular chamber. Two representative video recordings each showing 6 min of an interaction between a female D. melanogaster fly and a male heterospecific fly (D. micromelanica) without (2ndFly; TOP video clip) or with (2ndFly+filament; BOTTOM video clip) a synthetic filament affixed to its head. Video playback speed is 20× real time.

Next
Next

The TRP-channel painless mediates substrate stiffness sensing in the legs during Drosophila oviposition