It is unclear how regulatory genes establish neural circuits that compose sex-specific behaviors. The Drosophila melanogaster male courtship song provides a powerful model to study this problem. Courting males vibrate a wing to sing bouts of pulses and hums, called pulse and sine song, respectively. We report the discovery of male-specific thoracic interneurons—the TN1A neurons—that are required specifically for sine song. The TN1A neurons can drive the activity of a sex-non-specific wing motoneuron, hg1, which is also required for sine song. The male-specific connection between the TN1A neurons and the hg1 motoneuron is regulated by the sexual differentiation gene doublesex. We find that doublesex is required in the TN1A neurons during development to increase the density of the TN1A arbors that interact with dendrites of the hg1motoneuron. Our findings demonstrate how a sexual differentiation gene can build a sex-specific circuit motif by modulating neuronal arborization.

•Doublesex-expressing TN1 neurons are necessary and sufficient for the male sine song•A subclass of TN1 neurons, TN1A, contributes to the sine song•TN1A neurons are functionally coupled to a sine song motoneuron, hg1•Doublesex regulates the connectivity between the TN1A and hg1 neurons

It is unclear how developmental regulatory genes specify sex-specific behaviors. Shirangi et al. demonstrate that the Drosophila sexual differentiation gene doublesex encodes a sex-specific behavior—male song—by promoting the connectivity between the male-specific TN1A neurons and the sex-non-specific hg1 neurons, which are required for production of the song.

Multi-modal image registration is a challenging task that is vital to fuse complementary signals for subsequent analyses. Despite much research into cost functions addressing this challenge, there exist cases in which these are ineffective. In this work, we show that (1) this is true for the registration of in-vivo Drosophila brain volumes visualizing genetically encoded calcium indicators to an nc82 atlas and (2) that machine learning based contrast synthesis can yield improvements. More specifically, the number of subjects for which the registration outright failed was greatly reduced (from 40% to 15%) by using a synthesized image.