Abstract

To navigate, animal brains must continuously estimate the body's heading in space and compare it with internal goals to guide movement. In the fruit fly Drosophila melanogaster a neural circuit in the central complex of the brain serves as internal compass by functioning as a ring attractor network. While this region is highly conserved and involved in navigation in many insects, it remains unclear whether the fly circuit represents a general blueprint for head direction computation, or whether different ecologies have driven distinct circuit solutions. Using synaptic-resolution circuit mapping, we identified homologous head direction networks in bees and ants and compared them to the fly circuit. We show that the insect head direction network is conserved across at least 300 million years of evolution. At the level of cell types and projection patterns all studied species share a nearly identical neural layout, both qualitatively and quantitatively. At the synaptic level, however, the fly and bee circuits differed fundamentally. The distinct wiring principles of homologous neurons expose highly evolvable elements within this otherwise stable circuit. Using these differences in circuit architecture to constrain computational models, we show that both the bee and fly circuits can effectively function as ring attractors with similar, yet distinct, properties. These results demonstrate that complex neural circuits can remain stable over many hundreds of millions of years, while still offering access points for evolution to flexibly adjust neural computations to changing ecological demands - illustrating how evolution balances stability and flexibility in brain circuits.